CN115803815A - Block position and/or rotation based weight set selection for base detection - Google Patents

Abstract

The present invention discloses a system for base calling that includes a memory that stores a topology of a neural network, a plurality of weight sets, and sensor data for a series of sensing cycles. Sequencing events span the temporal progression of the base calling operation through a subseries of sensing cycles and the spatial progression of the base calling operation through locations on the biosensor. The configurable processor is configured to: load the topology on the configurable processor, select the set of weights based on the subject sub-series of sensing cycles and/or the subject's position on the biosensor, load the processing element with subject sensor data for the subject subseries of the sensing cycle and the subject location, configure the topology using the selected set of weights, and cause the neural network to process the subject sensor data to generate Base calling categorical data for subseries and subject positions.

Description

Block position and/or cycle based weight set selection for base calling

priority application

This application claims U.S. Provisional Patent Application No. 63/161,880, filed March 16, 2021, entitled "Tile Location and/or Cycle BasedWeight Set Selection for Base Calling" (Attorney Docket No. ILLM 1019-1/IP-1861- PRV); U.S. Provisional Patent Application No. 63/161,896, filed March 16, 2021, entitled "Neural NetworkParameter Quantization for Base Calling" (Attorney Docket No. ILLM 1019-2/IP-2049-PRV); March 2022 U.S. Nonprovisional Patent Application No. 17/687,551 (Attorney Docket ILLM 1019-3/IP-1861-US), entitled "Tile Location and/orCycle Based Weight Set Selection for Base Calling," filed May 4; 2022 Benefit of U.S. Nonprovisional Patent Application No. 17,687,583 (Attorney Docket ILLM 1019-4/IP-2049-US), filed March 4, entitled "NeuralNetwork Parameter Quantization for Base Calling." The priority application is hereby incorporated herein by reference for all purposes.

technical field

The technology disclosed in the present invention relates to artificial intelligence type computers and digital data processing systems and corresponding data processing methods and products for simulating intelligence (that is, knowledge-based systems, inference systems, and knowledge acquisition systems); Inference systems (eg, fuzzy logic systems), adaptive systems, machine learning systems, and artificial neural networks. In particular, the disclosed techniques relate to the use of deep neural networks, such as deep convolutional neural networks, to analyze data and the selective use of weight sets.

Literature incorporated

The following documents are incorporated by reference as if fully set forth herein:

U.S. Provisional Patent Application No. 62/979,384, entitled "ARTIFICIAL INTELLIGENCE-BASED BASE CALLINGOF INDEX SEQUENCES," filed February 20, 2020 (Attorney Docket No. ILLM 1015-1/IP-1857-PRV);

U.S. Provisional Patent Application No. 62/979,414, entitled "ARTIFICIAL INTELLIGENCE-BASED MANY-TO-MANYBASE CALLING," filed February 20, 2020 (Attorney Docket No. ILLM 1016-1/IP-1858-PRV);

U.S. Provisional Patent Application No. 62/979,385, entitled "KNOWLEDGE DISTILLATION-BASED COMPRESSION OFARTIFICIAL INTELLIGENCE-BASED BASE CALLER," filed February 20, 2020 (Attorney Docket No. ILLM 1017-1/IP-1859-PRV);

U.S. Provisional Patent Application No. 63/072,032, entitled "DETECTING AND FILTERING CLUSTERS BASED ONARTIFICIAL INTELLIGENCE-PREDICTED BASE CALLS," filed August 28, 2020 (Attorney Docket No. ILLM 1018-1/IP-1860-PRV);

U.S. Provisional Patent Application No. 62/979,411, entitled "DATA COMPRESSION FOR ARTIFICIALINTELLIGENCE-BASED BASE CALLING," filed February 20, 2020 (Attorney Docket No. ILLM 1029-1/IP-1964-PRV);

U.S. Provisional Patent Application No. 62/979,399, entitled "SQUEEZING LAYER FOR ARTIFICIALINTELLIGENCE-BASED BASE CALLING," filed February 20, 2020 (Attorney Docket No. ILLM 1030-1/IP-1982-PRV);

U.S. Nonprovisional Patent Application No. 16/825,987, entitled "TRAINING DATA GENERATION FOR ARTIFICIALINTELLIGENCE-BASED SEQUENCING," filed March 20, 2020 (Attorney Docket No. ILLM 1008-16/IP-1693-US);

U.S. Nonprovisional Patent Application No. 16/825,991, entitled "ARTIFICIAL INTELLIGENCE-BASED GENERATION OFSEQUENCING METADATA," filed March 20, 2020 (Attorney Docket No. ILLM 1008-17/IP-1741-US);

U.S. Nonprovisional Patent Application No. 16/826,126, entitled "ARTIFICIAL INTELLIGENCE-BASED BASE CALLING," filed March 20, 2020 (Attorney Docket No. ILLM 1008-18/IP-1744-US);

U.S. Nonprovisional Patent Application No. 16/826,134, entitled "ARTIFICIAL INTELLIGENCE-BASED QUALITYSCORING," filed March 20, 2020 (Attorney Docket No. ILLM 1008-19/IP-1747-US);

U.S. Nonprovisional Patent Application No. 16/826,168, entitled "ARTIFICIAL INTELLIGENCE-BASED SEQUENCING," filed March 21, 2020 (Attorney Docket No. ILLM 1008-20/IP-1752-US);

U.S. Nonprovisional Patent Application No. 16/874,599, entitled "Systems and Devices for Characterization and Performance Analysis of Pixel-Based Sequencing," filed May 14, 2020 (Attorney Docket No. ILLM 1011-4/IP-1750-US) ;as well as

U.S. Nonprovisional Patent Application No. 17/176,147, entitled "HARDWARE EXECUTION AND ACCELERATION OFARTIFICIAL INTELLIGENCE-BASED BASE CALLER," filed February 15, 2021 (Attorney Docket No. ILLM1020-2/IP-1866-US).

Background technique

Subject matter discussed in this section should not be admitted to be prior art merely by virtue of its mention in this section. Similarly, issues mentioned in this section or in connection with subject matter provided as background should not be admitted to have been previously recognized in the prior art. The subject matter in this section merely represents different approaches, which may themselves correspond to specific implementations of the claimed technology.

In recent years, the rapid increase in computing power has allowed deep convolutional neural networks (CNNs) to achieve great success in many computer vision tasks with significantly improved accuracy. In the inference stage, many applications require low-latency processing of an image with strict power consumption requirements, which reduces the efficiency of graphics processing units (GPUs) and other general-purpose platforms. By customizing digital circuits dedicated to deep learning algorithm inference, Opportunities arise for specific acceleration hardware such as Field Programmable Gate Arrays (FPGAs). However, it is still challenging to deploy CNNs on portable and embedded systems due to large data volumes, intensive computations, changing algorithmic structures, and frequent memory accesses.

Since convolution contributes most of the operations in CNN, the convolution acceleration scheme significantly affects the efficiency and performance of hardware CNN accelerators. Convolution involves multiply and accumulate (MAC) operations with four recurrent stages sliding along the kernel and feature maps. The first loop stage computes the MAC of the pixels within the kernel window. The second recurrent stage accumulates the sum of products of the MACs across different input feature maps. After completion of the first and second cyclic stages, the final output pixels are obtained by adding a bias. The third recurrent stage slides a kernel window within the input feature map. The fourth recurrent stage generates different output feature maps.

FPGA due to its (1) high reconfigurability, (2) faster development time compared to application-specific integrated circuit (ASIC) to keep up with the rapid development of CNN, (3) good performance, and (4) integration with GPU Compared to superior energy efficiency, it has gained increasing attention and popularity, especially in accelerating inference tasks. The high performance and efficiency of FPGAs can be achieved by synthesizing circuits tailored for specific computations to directly process billions of operations with custom memory systems. For example, hundreds to thousands of digital signal processing (DSP) blocks on modern FPGAs support core convolution operations, eg, multiplication and addition, with high parallelism. A dedicated data buffer between the external on-chip memory and the on-chip processing engine (PE) can be designed to enable optimal data flow by configuring tens of megabytes of on-chip block random access memory (BRAM) on the FPGA chip.

Efficient data flow and CNN-accelerated hardware architecture are required to minimize data communication while maximizing resource utilization to achieve high performance. Therefore, there is an opportunity to design methods and frameworks to accelerate the inference process of various CNN algorithms on accelerated hardware with high performance, high efficiency, and high flexibility.

Description of drawings

In the drawings, like reference characters generally refer to like parts throughout the different views. Additionally, the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the disclosed technology. In the following description, various implementations of the disclosed technology are described with reference to the following figures, in which:

Figure 1 shows a cross-section of a biosensor that can be used in various embodiments.

Figure 2 shows an implementation of a flow cell containing clusters in its blocks.

Figure 3 shows an exemplary flow cell with eight channels, and also shows an enlarged view of a block and its clusters and their surrounding context.

4 is a simplified block diagram of a system for analyzing sensor data from a sequencing system, such as base calling sensor output.

5 is a simplified diagram illustrating aspects of a base calling operation, including the functionality of a runtime program executed by a host processor.

6 is a simplified diagram of the configuration of a configurable processor, such as the configurable processor of FIG. 4 .

7 is a diagram of a neural network architecture that may be implemented using a configurable or reconfigurable array configured as described herein.

FIG. 8A is a simplified illustration of the organization of blocks of sensor data used by a neural network architecture like FIG. 7 .

FIG. 8B is a simplified illustration of a patch of blocks of sensor data used by a neural network architecture like FIG. 7 .

Figure 9 shows a portion of the configuration of a neural network like Figure 7 on a configurable or reconfigurable array, such as a Field Programmable Gate Array (FPGA).

10 is a diagram of another alternative neural network architecture that may be implemented using a configurable or reconfigurable array configured as described herein.

FIG. 11 shows a specific implementation of a specialized architecture of a neural network-based base caller for isolating the processing of data for different sequencing cycles.

Figure 12 shows one implementation of isolation layers, each of which may include convolutions.

Figure 13A shows one implementation of combined layers, each of which may include convolutions.

Figure 13B shows another implementation of combined layers, each of which may include convolutions.

Figures 14, 15, and 16 illustrate various exemplary tile position-based weight selection schemes for base calling.

Figure 17A shows an example of fading in which signal strength decreases with cycle number of a sequencing run as a base calling operation.

Figure 17B conceptually illustrates the decreasing signal-to-noise ratio as the sequencing cycle progresses.

Figure 18 illustrates an exemplary base calling cycle number based weight selection scheme for base calling.

Figures 19, 20, 21A and 21B illustrate various exemplary weight selection schemes based on (i) temporal progression of base calling cycle numbers and (ii) spatial location of blocks.

FIG. 22 shows an implementation of a base calling operation, in which a set of weights for base calling is selected based on spatial block information and temporal sensing cycle subseries information.

Figure 23A shows various sets of weights for various classes of blocks and for various sensing cycles, each set of weights including corresponding spatial weights and corresponding temporal weights.

Figure 23B shows various sets of weights for various classes of tiles and for various cycles, where the different sets of weights for a particular class of tiles include common spatial weights and different temporal weights.

Figure 23C illustrates a system for selecting a set of weights based on one or more sequencing run parameters.

Figure 24 is a block diagram of a base calling system according to one implementation.

FIG. 25 is a block diagram of a system controller that may be used in the system of FIG. 24 .

26 is a simplified block diagram of a computer system that can be used to implement the disclosed techniques.

Detailed ways

Embodiments described herein can be used in various biological or chemical processes and systems for academic or commercial analysis. More specifically, the embodiments described herein can be used in various processes and systems where it is desirable to detect an event, property, quality or characteristic indicative of a desired response. For example, embodiments described herein include cartridges, biosensors and components thereof, and bioassay systems that operate with the cartridges and biosensors. In certain embodiments, the cartridge and biosensor include a flow cell and one or more sensors, pixels, photodetectors, or photodiodes coupled together in a substantially unitary structure.

The following detailed descriptions of certain embodiments are better understood when read in conjunction with the accompanying figures. To the extent that figures show diagrams of the functional blocks of various embodiments, the functional blocks are not necessarily indicative of the division between hardware circuitry. Thus, for example, one or more of the functional blocks (eg, a processor or memory) may be implemented in a single piece of hardware (eg, a general-purpose signal processor or random access memory, hard disk, etc.). Similarly, a program may be a stand-alone program, may be incorporated as a subroutine in an operating system, may be a function in an installed software package, and so on. It should be understood that the various embodiments are not limited to the arrangements and instrumentalities shown in the drawings.

As used herein, an element or step recited in the singular and preceded by the word "a" or "an" should be understood as not excluding plural of said elements or steps, unless such exclusion is explicitly stated. Furthermore, references to "one embodiment" are not intended to be interpreted as excluding the existence of additional embodiments that also incorporate the recited features. Furthermore, an embodiment that "comprises" or "has" or "comprises" one or more elements having a particular attribute may include additional elements whether or not they have that attribute, unless expressly stated to the contrary.

As used herein, a "desired response" includes a change in at least one of the chemical, electrical, physical or optical properties (or qualities) of an analyte of interest. In certain embodiments, the desired response is a positive binding event (eg, binding of a fluorescently labeled biomolecule to an analyte of interest). More generally, the desired reaction may be a chemical transformation, chemical change, or chemical interaction. The desired response may also be a change in electrical properties. For example, the desired response may be a change in the concentration of ions in the solution. Exemplary reactions include, but are not limited to, chemical reactions such as reduction, oxidation, addition, elimination, rearrangement, esterification, amidation, etherification, cyclization, or substitution; binding interactions in which a first chemical interacts with a second chemical Binding; dissociation reactions, in which two or more chemicals are separated from each other; fluorescence; luminescence; bioluminescence; chemiluminescence; and biological reactions, such as nucleic acid replication, nucleic acid amplification, nucleic acid hybridization, nucleic acid ligation, phosphorylation, enzymatic catalysis , receptor binding or ligand binding. The desired reaction may also be the addition or removal of protons, for example detectable as a change in pH of the surrounding solution or environment. An additional desired reaction can detect the flow of ions across a membrane (eg, a natural or synthetic bilayer membrane), eg, when ions flow through the membrane, the current is interrupted and the interruption can be detected.

)共振能量转移(FRET)，通过分离供体荧光团和受体荧光团来降低FRET，通过分离淬灭基团与荧光团来增加荧光，或通过共定位淬灭基团和荧光团来减少荧光。In certain embodiments, the desired reaction comprises binding a fluorescently labeled molecule to the analyte. Analytes can be oligonucleotides, and fluorescently labeled molecules can be nucleotides. When excitation light is directed at the oligonucleotide with labeled nucleotides, and the fluorophore emits a detectable fluorescent signal, the desired reaction is detected. In alternative embodiments, the detected fluorescence is the result of chemiluminescence or bioluminescence. The desired reaction can also increase fluorescence (or

) resonance energy transfer (FRET), which reduces FRET by separating the donor and acceptor fluorophores, increases fluorescence by separating the quencher from the fluorophore, or decreases fluorescence by colocalizing the quencher and fluorophore .

As used herein, "reaction component" or "reactant" includes any substance that can be used to obtain a desired reaction. For example, reaction components include reagents, enzymes, samples, other biomolecules, and buffers. Typically the reaction components are delivered to and/or immobilized at the reaction site in solution. A reaction component may directly or indirectly interact with another species, such as an analyte of interest.

As used herein, the term "reaction site" is a localized area where a desired reaction can occur. A reaction site may comprise a support surface of a substrate on which a substance may be immobilized. For example, a reaction site may comprise a substantially planar surface in a channel of a flow cell having a population of nucleic acids thereon. Usually, but not always, the nucleic acids in a population have the same sequence, eg, are clonal copies of a single- or double-stranded template. However, in some embodiments, a reactive site may comprise only a single nucleic acid molecule, eg, in single- or double-stranded form. In addition, the plurality of reaction sites may be unevenly distributed along the support surface or arranged in a predetermined pattern (eg, arranged side-by-side in a matrix, such as in a microarray). A reaction site may also include a reaction chamber (or well) that at least partially defines a spatial region or volume configured to separate desired reactions.

This application uses the terms "reaction chamber" and "well" interchangeably. As used herein, the term "reaction chamber" or "well" includes a region of space that is in fluid communication with a flow channel. The reaction chamber can be at least partially isolated from the surrounding environment or other spatial regions. For example, multiple reaction chambers may be separated from each other by a common wall. As a more specific example, the reaction chamber can include a cavity defined by the interior surface of the bore and have an opening or orifice such that the cavity can be in fluid communication with the flow channel. Biosensors comprising such reaction chambers are described in more detail in International Application No. PCT/US2011/057111, filed October 20, 2011, which is incorporated herein by reference in its entirety.

In some embodiments, the reaction chamber is sized and shaped relative to the solid (including semi-solid) such that the solid can be fully or partially inserted therein. For example, the reaction chamber can be sized and shaped to accommodate only one capture bead. The capture beads may have clonally amplified DNA or other material thereon. Alternatively, the reaction chamber can be sized and shaped to receive an approximate number of beads or solid substrate. As another example, the reaction chamber may also be filled with a porous gel or substance configured to control diffusion or filter fluids that may flow into the reaction chamber.

In some embodiments, sensors (eg, photodetectors, photodiodes) are associated with corresponding pixel regions of the sample surface of the biosensor. Thus, a pixel area is a geometrical configuration representing the area on the sample surface of a biosensor for one sensor (or pixel). A sensor associated with a pixel area detects light emission collected from the associated pixel area when a desired reaction occurs at a reaction site or reaction chamber covering the associated pixel area. In flat surface implementations, the pixel areas may overlap. In some cases, multiple sensors may be associated with a single reaction site or a single reaction chamber. In other cases, a single sensor can be associated with a set of reaction sites or a set of reaction chambers.

As used herein, "biosensor" includes structures having multiple reaction sites and/or reaction chambers (or wells). A biosensor may include a solid-state imaging device (eg, a CCD or CMOS imaging device) and optionally a flow cell mounted thereto. A flow cell may comprise at least one flow channel in fluid communication with a reaction site and/or a reaction chamber. As a specific example, a biosensor is configured to be fluidly and electrically coupled to a bioassay system. A bioassay system can deliver reactants to reaction sites and/or reaction chambers and perform multiple imaging events according to a predetermined protocol (eg, sequencing by synthesis). For example, a bioassay system can direct the flow of a reaction solution along a reaction site and/or a reaction chamber. At least one of the solutions may contain four types of nucleotides with the same or different fluorescent labels. Nucleotides can bind to corresponding oligonucleotides located at reaction sites and/or reaction chambers. The bioassay system can then illuminate the reaction site and/or reaction chamber using an excitation light source (eg, a solid state light source such as a light emitting diode (LED)). The excitation light may have a predetermined wavelength or wavelengths, including a range of wavelengths. The excited fluorescent label provides an emission signal that can be captured by the sensor.

In alternative embodiments, a biosensor may include electrodes or other types of sensors configured to detect other identifiable properties. For example, a sensor may be configured to detect changes in ion concentration. In another example, the sensor can be configured to detect ionic current flow across the membrane.

As used herein, a "cluster" is a population of similar or identical molecules or nucleotide sequences or DNA strands. For example, a cluster can be an amplified oligonucleotide or any other set of polynucleotides or polypeptides having the same or similar sequence. In other embodiments, a cluster can be any element or group of elements that occupies a physical area on the sample surface. In embodiments, clusters are immobilized to reaction sites and/or reaction chambers during base calling cycles.

As used herein, the term "immobilized" when used in reference to a biomolecule or biological or chemical substance includes substantially attaching the biomolecule or biological or chemical substance to a surface at the molecular level. For example, biomolecules or biological or chemical substances can be immobilized to the surface of the substrate material using adsorption techniques including non-covalent interactions (e.g., electrostatic forces, van der Waals forces, and dehydration of hydrophobic interfaces) and covalent bonding techniques, Among them, functional groups or linkers facilitate the attachment of biomolecules to surfaces. The immobilization of biomolecules or biological or chemical substances to the surface of the substrate material may be based on the properties of the substrate surface, the liquid medium carrying the biomolecules or biological or chemical substances, and the properties of the biomolecules or biological or chemical substances themselves. In some cases, the substrate surface can be functionalized (eg, chemically or physically modified) to facilitate the immobilization of biomolecules (or biological or chemical species) to the substrate surface. The surface of the substrate may first be modified to bind functional groups to the surface. The functional groups can then be combined with biomolecules or biological or chemical substances to immobilize the functional groups thereon. Substances can be immobilized to surfaces via gels, for example, as described in US Patent Publication No. US 2011/0059865 Al, which is incorporated herein by reference.

In some embodiments, nucleic acids can be attached to surfaces and amplified using bridge amplification. Useful bridge amplification methods are described, for example, in U.S. Patent No. 5,641,658, WO 2007/010251, U.S. Patent No. 6,090,592, U.S. Patent Publication No. 2002/0055100 Al, U.S. Patent No. 7,115,400, U.S. Patent Publication No. 2004/0096853 Al, U.S. Patent No. Described in Publication No. 2004/0002090 Al, US Patent Publication No. 2007/0128624 Al, or US Patent Publication No. 2008/0009420 Al, each of which is incorporated herein in its entirety. Another useful method for amplifying nucleic acids on surfaces is rolling circle amplification (RCA), eg, using the method described in further detail below. In some embodiments, nucleic acids can be attached to a surface and amplified using one or more primer pairs. For example, one of the primers can be in solution and the other primer can be immobilized on the surface (eg, 5'-attached). By way of example, a nucleic acid molecule can be hybridized to one of the primers on the surface, followed by fixed primer extension to produce a first copy of the nucleic acid. The primer in solution then hybridizes to the first copy of the nucleic acid, which primer can be extended using the first copy of the nucleic acid as a template. Optionally, after the first copy of the nucleic acid is produced, the original nucleic acid molecule can be hybridized to a second immobilized primer on the surface, and can be extended concurrently with or subsequent to primer extension in solution. In any embodiment, repeated extension (eg, amplification) rounds using immobilized primers and primers in solution provide multiple copies of the nucleic acid.

In certain embodiments, assay protocols performed by the systems and methods described herein include the use of natural nucleotides and enzymes configured to interact with natural nucleotides. Natural nucleotides include, for example, ribonucleotides (RNA) or deoxyribonucleotides (DNA). Natural nucleotides can be in monophosphate, diphosphate or triphosphate form and can have a (C) base. It should be understood, however, that non-natural nucleotides, modified nucleotides, or analogs of the foregoing may be used. Some examples of useful unnatural nucleotides are set forth below with respect to reversible terminator-based sequencing-by-synthesis approaches.

In embodiments that include a reaction chamber, objects or solid matter (including semi-solid matter) may be disposed within the reaction chamber. When positioned, items or solids may be physically held or secured within the reaction chamber by interference fit, adhesion or entrapment. Examples of items or solids that may be disposed within the reaction chamber include polymeric beads, pellets, sepharose, powders, quantum dots, or other solids that may be compressed and/or held within the reaction chamber. In particular embodiments, nucleic acid superstructures, such as DNA spheres, may be disposed in or at the reaction chamber, eg, by attaching to an inner surface of the reaction chamber or by residing in a liquid within the reaction chamber. DNA spheres or other nucleic acid superstructures can be made and then placed in or at the reaction chamber. Alternatively, DNA spheres can be synthesized in situ at the reaction chamber. DNA spheres can be synthesized by rolling circle amplification to produce concatemers of specific nucleic acid sequences, and the concatemers can be treated with conditions that form relatively compact spheres. DNA spheres and methods of their synthesis are described, for example, in US Patent Publication Nos. 2008/0242560 Al or 2008/0234136 Al, each of which is incorporated herein in its entirety. The substance held or disposed in the reaction chamber may be solid, liquid or gaseous.

As used herein, "base calling" identifies nucleotide bases in a nucleic acid sequence. Base calling refers to the process of determining the base call (A, C, G, T) for each cluster at a specific cycle. For example, base calling can be performed using the four-pass, two-pass or one-pass methods and systems described in the incorporated material of US Patent Application Publication No. 2013/0079232. In certain embodiments, a base calling cycle is referred to as a "sampling event." In one-dye and two-channel sequencing protocols, sampling events consist of two illumination phases in time series such that pixel signals are generated in each phase. The first illumination stage induces illumination from a given cluster, indicative of nucleotide bases A and T in the AT pixel signal, and the second illumination stage induces illumination from a given cluster, indicative of the nucleotide base in the CT pixel signal Base C and T.

biological sensor

Figure 1 shows a cross-section of a biosensor 100 that can be used in various embodiments. Biosensor 100 has pixel regions 106', 108', 110', 112', and 114' that can each hold more than one cluster (eg, 2 clusters per pixel region) during a base calling cycle. As shown, biosensor 100 may include a flow cell 102 mounted to a sampling device 104 . In the illustrated embodiment, the flow cell 102 is directly attached to the sampling device 104 . However, in alternative embodiments, flow cell 102 may be removably coupled to sampling device 104 . Sampling device 104 has a sample surface 134 that can be functionalized (eg, chemically or physically modified in a manner appropriate to conduct a desired reaction). For example, sample surface 134 may be functionalized and may include a plurality of pixel regions 106', 108', 110', 112', and 114', which may each hold more than one cluster during a base calling cycle. (eg, each pixel region has a corresponding cluster pair 106A, 106B; 108A, 108B; 110A, 110B; 112A, 112B; and 114A, 114B affixed thereto). Each pixel area is associated with a corresponding sensor (or pixel or photodiode) 106, 108, 110, 112, and 114 such that light received by the pixel area is captured by the corresponding sensor. Pixel regions 106 ′ may also be associated with corresponding reaction sites 106 ″ holding cluster pairs on sample surface 134 such that light emitted from reaction sites 106 ″ is received by pixel regions 106 ′ and captured by corresponding sensors 106 . Due to this sensing structure, the pixel signal in the base-calling cycle carries information based on all of the two or more clusters in the case where during a base-calling cycle, at a specific There are two or more clusters in a pixel area of the sensor (eg, each pixel area has a corresponding pair of clusters). Thus, signal processing as described herein is used to distinguish each cluster where there are more clusters than pixel signals in a given sampling event of a particular base calling cycle.

In the illustrated embodiment, the flow cell 102 includes side walls 138 , 125 and a flow shield 136 supported by the side walls 138 , 125 . Sidewalls 138 , 125 are coupled to sample surface 134 and extend between flow shield 136 and sidewalls 138 , 125 . In some embodiments, sidewalls 138 , 125 are formed from a curable adhesive layer that bonds flow shield 136 to sampling device 104 .

The sidewalls 138 , 125 are sized and shaped such that a flow channel 144 exists between the flow shield 136 and the sampling device 104 . Flow shield 136 may include a material that is transparent to excitation light 101 propagating into flow channel 144 from the exterior of biosensor 100 . In an example, the excitation light 101 approaches the flow mask 136 at a non-orthogonal angle.

Also as shown, the flow shield 136 may include inlet and outlet ports 142, 146 configured to fluidly engage other ports (not shown). For example, other ports can come from cartridges or workstations. Flow channel 144 is sized and shaped to direct fluid along sample surface 134 . The height H ₁ and other dimensions of the flow channel 144 may be configured to maintain a substantially uniform flow of fluid along the sample surface 134 . The dimensions of the flow channel 144 may also be configured to control bubble formation.

By way of example, flow shield 136 (or flow cell 102 ) may comprise a transparent material such as glass or plastic. The flow shield 136 may constitute a substantially rectangular block having a planar outer surface and a planar inner surface defining the flow channel 144 . The block can be mounted to the side walls 138,125. Alternatively, flow cell 102 may be etched to define flow shield 136 and side walls 138 , 125 . For example, grooves may be etched into the transparent material. The grooves may become flow channels 144 when the etching material is installed into the sampling device 104 .

The sampling device 104 may be similar to, for example, an integrated circuit including a plurality of stacked substrate layers 120-126. Substrate layers 120 to 126 may include a base substrate 120 , a solid state imaging device 122 (eg, a CMOS image sensor), a filter or light management layer 124 , and a passivation layer 126 . It should be noted that the above is illustrative only, and other embodiments may include fewer or additional layers. Additionally, each of the substrate layers 120-126 may include multiple sub-layers. Sampling device 104 may be fabricated using processes similar to those used in fabricating integrated circuits such as CMOS image sensors and CCDs. For example, substrate layers 120 - 126 , or portions thereof, may be grown, deposited, etched, etc., to form sampling device 104 .

Passivation layer 126 is configured to shield filter layer 124 from the fluidic environment of flow channel 144 . In some cases, passivation layer 126 is also configured to provide a solid surface (ie, sample surface 134 ) that allows biomolecules or other analytes of interest to immobilize thereon. For example, each reaction site may include a cluster of biomolecules immobilized to the sample surface 134 . Accordingly, the passivation layer 126 may be formed of a material that allows reaction sites to be fixed thereto. Passivation layer 126 may also include a material that is transparent to at least the desired fluorescence. By way of example, passivation layer 126 may include silicon nitride (Si ₂ N ₄ ) and/or silicon dioxide (SiO ₂ ). However, other suitable materials may be used. In the illustrated embodiment, passivation layer 126 may be substantially planar. However, in alternative embodiments, the passivation layer 126 may include recesses, such as pits, holes, grooves, and the like. In the illustrated embodiment, passivation layer 126 has a thickness of about 150 nm to 200 nm, and more specifically about 170 nm.

Filter layer 124 may include various features that affect the transmission of light. In some implementations, filter layer 124 may perform multiple functions. For example, filter layer 124 may be configured to (a) filter unwanted light signals, such as light signals from excitation light sources; (b) direct emission signals from reaction sites to corresponding sensors 106, 108, 110, 112 and 114, the sensors are configured to detect emission signals from reaction sites; or (c) block or prevent detection of unwanted emission signals from adjacent reaction sites. Accordingly, filter layer 124 may also be referred to as a light management layer. In the illustrated embodiment, filter layer 124 has a thickness of about 1 μm to 5 μm, more specifically about 2 μm to 4 μm. In alternative embodiments, filter layer 124 may include an array of microlenses or other optical elements. Each microlens can be configured to direct an emission signal from an associated reaction site to a sensor.

In some embodiments, solid-state imaging device 122 and base substrate 120 may be provided together as a previously constructed solid-state imaging device (eg, a CMOS chip). For example, the base substrate 120 may be a silicon wafer, and the solid-state imaging device 122 may be mounted thereon. Solid state imaging device 122 includes a layer of semiconductor material (eg, silicon) and sensors 106 , 108 , 110 , 112 , and 114 . In the illustrated embodiment, the sensor is a photodiode configured to detect light. In other embodiments, the sensor includes a light detector. The solid-state imaging device 122 can be manufactured as a single chip by a CMOS-based manufacturing process.

Solid-state imaging device 122 may include a dense array of sensors 106 , 108 , 110 , 112 , and 114 configured to detect activity indicative of a desired response from within or along flow channel 144 . In some embodiments, each sensor has a pixel area (or detection area) of about 1 to 2 square microns (μm ² ). Arrays may include half a million sensors, five million sensors, ten million sensors, or even 120 million sensors. Sensors 106, 108, 110, 112, and 114 may be configured to detect light of a predetermined wavelength indicative of a desired response.

In some embodiments, sampling device 104 includes a microcircuit arrangement, such as that described in US Patent No. 7,595,882, which is incorporated herein by reference in its entirety. More specifically, sampling device 104 may include an integrated circuit having a planar array of sensors 106 , 108 , 110 , 112 , and 114 . Circuitry formed within sampling device 104 may be configured for at least one of signal amplification, digitization, storage, and processing. Circuitry may collect and analyze the detected fluorescence and generate pixel signals (or detection signals) for communicating the detection data to a signal processor. The circuitry may also perform additional analog and/or digital signal processing in the sampling device 104 . Sampling device 104 may include conductive vias 130 that perform signal routing (eg, transmit pixel signals to a signal processor). Pixel signals may also be transmitted through the electrical contacts 132 of the sampling device 104 .

Relative to U.S. Nonprovisional Patent Application No. 16/874,599, entitled "Systems and Devices for Characterization and Performance Analysis of Pixel-Based Sequencing," filed May 14, 2020 (Attorney Docket No. ILLM 1011-4/IP-1750- US) discusses the sampling device 104 in further detail, which patent application is incorporated herein by reference as if fully set forth herein. Sampling device 104 is not limited to the aforementioned configurations or uses as described above. In alternative embodiments, sampling device 104 may take other forms. For example, sampling device 104 may include a CCD device, such as a CCD camera, that is coupled to a flow cell or that moves to interact with a flow cell having reaction sites therein.

Figure 2 shows one implementation of a flow cell 200 that includes clusters in its blocks. Flow cell 200 corresponds to flow cell 102 of FIG. 1 , eg, without flow shield 136 . Furthermore, the depiction of flow cell 200 is symbolic in nature, and flow cell 200 symbolically depicts the various channels and blocks therein without showing various other components therein. FIG. 2 shows a top view of flow cell 200 .

In one embodiment, flow cell 200 is divided or partitioned into a plurality of channels, such as channels 202a, 202b, . . . , 202P, ie, P channels. In the example of Fig. 2, the flow cell 200 is shown as comprising 8 channels, ie P = 8 in this example, but the number of channels within the flow cell is implementation specific.

In one embodiment, each channel 202 is further partitioned into non-overlapping regions referred to as “blocks” 212 . For example, FIG. 2 shows an enlarged view of section 208 of an exemplary channel. Section 208 is shown to include a plurality of tiles 212 .

In an example, each slot 202 includes one or more columns of blocks. For example, in FIG. 2 , each channel 202 includes two corresponding block columns 212 , as shown in enlarged section 208 . The number of blocks in each block column within each slot is implementation specific, and in one example there may be 50 blocks, 60 blocks, 100 blocks or another appropriate number of blocks.

Each block includes a corresponding number of clusters. During the sequencing process, the clusters on the block and their surrounding background are imaged. For example, FIG. 2 shows an example cluster 216 within an example block.

Figure 3 shows an exemplary Illumina GA-IIx ^™ flow cell with eight channels, and also shows a magnified view of a block and its clusters and their surrounding background. For example, there are one hundred tiles per lane in the Illumina Genome Analyzer II, and sixty-eight tiles per lane in the Illumina HiSeq2000. Blocks 212 hold hundreds of thousands to millions of clusters. In FIG. 3 , an image generated from a block with clusters shown as bright spots is shown at 308 (eg, 308 is an enlarged image view of the block), with exemplary cluster 304 labeled. Cluster 304 includes about a thousand identical copies of the template molecule, but the clusters vary in size and shape. Clusters are generated from template molecules by bridge amplification of the input library prior to a sequencing run. The purpose of amplification and cluster growth is to increase the intensity of the emitted signal, since imaging devices cannot reliably sense individual fluorophores. However, the physical distance of the DNA fragments within a cluster 304 is small, so the imaging device perceives the cluster of fragments as a single point 304 .

Further Details Relative to U.S. Nonprovisional Patent Application No. 16/825,987, Filed March 20, 2020, entitled "TRAINING DATA GENERATION FORARTIFICIAL INTELLIGENCE-BASED SEQUENCING," (Attorney Docket No. ILLM 1008-16/IP-1693-US) clusters and blocks are discussed;

4 is a simplified block diagram of a system for analyzing sensor data from a sequencing system, such as base calling sensor output (see, eg, FIG. 1 ). In the example of FIG. 4 , the system includes a sequencing machine 400 and a configurable processor 450 . The configurable processor 450 may execute the neural network-based base caller in coordination with a runtime program executed by a host processor, such as a central processing unit (CPU) 402 . Sequencing machine 400 includes a base calling sensor and flow cell 401 (eg, as discussed with respect to FIGS. 1-3 ). The flow cell may comprise one or more blocks in which clusters of genetic material are exposed to a sequence of analyte streams that are used to elicit reactions in the clusters to identify bases in the genetic material, as described with respect to FIGS. Figure 3 is discussed. A sensor senses the response of each cycle of the sequence in each block of the flow cell to provide block data. Examples of this technique are described in more detail below. Genetic sequencing is a data-intensive operation that converts base-calling sensor data into a base-calling sequence for each cluster of genetic material sensed during a base-calling operation.

The system in this example includes a CPU 402 that executes a runtime program to coordinate a base calling operation, a memory 403 for storing the sequence of the block data array, the base calling reads resulting from the base calling operation, and the base Additional information used in the base checkout operation. Additionally, in this illustration, the system includes memory 404 for storing a configuration file (or files) such as an FPGA bitfile and a model of a neural network for configuring and reconfiguring a configurable processor 450 and executing a neural network parameter. Sequencing machine 400 may include a program for configuring a configurable processor, and in some embodiments a reconfigurable processor, to execute a neural network.

Sequencing machine 400 is coupled to configurable processor 450 via bus 405 . Bus 405 may be implemented using high throughput technology, such as in one example, a bus technology compatible with the PCIe standard (Peripheral Component Interconnect Express) currently maintained and developed by PCI-SIG (PCI Special Interest Group). Additionally, memory 460 is coupled to configurable processor 450 via bus 461 in this example. Memory 460 may be an on-board memory provided on a circuit board with configurable processor 450 . Memory 460 is used for high-speed access by configurable processor 450 to working data used in base calling operations. Bus 461 may also be implemented using high-throughput technology such as a bus technology compatible with the PCIe standard.

Configurable processors, including field-programmable gate arrays (FPGAs), coarse-grained reconfigurable arrays (CGRAs), and other configurable and reconfigurable devices, that can be configured to perform better than is possible with a general-purpose processor executing a computer program more efficient or faster implementation of various functions. Configuration of a configurable processor involves compiling a functional description to produce a configuration file, sometimes called a bitstream or bitfile, and distributing the configuration file to configurable elements on the processor.

The configuration file is configured by configuring the circuit to set the data flow mode, use of distributed memory and other on-chip memory resources, look-up table content, configurable logic blocks, and configurable execution other elements of the array) to define the logical functions to be performed by the configurable processor. A configurable processor is reconfigurable if the configuration file can be changed in the field by changing the loaded configuration file. For example, configuration files may be stored in volatile SRAM elements, in non-volatile read-write memory elements, and combinations thereof, distributed among arrays of configurable elements on a configurable or reconfigurable processor. A variety of commercially available configurable processors are suitable for base calling operations as described herein. Examples include commercially available products such as Xilinx Alveo ^™ U200, Xilinx Alveo ^™ U250, Xilinx Alveo ^™ U280, Intel/Altera Stratix ^™ GX2800, Intel/Altera Stratix ^™ GX2800, and Intel Stratix ^™ GX10M. In some examples, the host CPU may be implemented on the same integrated circuit as the configurable processor.

Embodiments described herein use a configurable processor 450 to implement a multi-cycle neural network. A configuration file for a configurable processor may be implemented by specifying the logical functions to be performed using a high-level description language (HDL) or register-transfer level (RTL) language specification. The specification may be compiled using resources designed for the selected configurable processor to generate the configuration file. To generate a design for an ASIC, which may not be a configurable processor, the same or a similar specification may be compiled.

Accordingly, in all embodiments described herein, alternatives to configurable processors include configured processors comprising application-specific ASICs or application-specific integrated circuits or groups of integrated circuits, or system-on-chip (SOC) devices , the configured processor configured to perform a neural network based base calling operation as described herein.

In general, configurable processors and configured processors as described herein that are configured to perform the operations of a neural network are referred to herein as neural network processors.

In this example, configurable processor 450 is configured by using a configuration file loaded by a program executed by CPU 402 or other source that configures an array of configurable elements on configurable processor 454 to perform base calling Function. In this example, the configuration includes data flow logic 451 that is coupled to bus 405 and bus 461 and that performs functions for distributing data and control parameters among elements used in a base calling operation.

Additionally, the configurable processor 450 is configured with base calling execution logic 452 to execute a multi-cycle neural network. Logic 452 includes a plurality of multi-cycle execution clusters (eg, 453 ), including multi-cycle cluster 1 through multi-cycle cluster X in this example. The number of multi-cycle clusters can be chosen according to a trade-off involving the desired throughput of the operation and the resources available on the configurable processor.

The multi-cycle cluster is coupled to dataflow logic 451 through a dataflow path 454 implemented using configurable interconnect and memory resources on a configurable processor. In addition, multi-cycle clusters are coupled to dataflow logic 451 through control paths 455 implemented using, for example, configurable interconnects and memory resources on configurable processors, which provide indication of available clusters, readiness to provide available clusters for An input unit that executes a run of the neural network, a control signal ready to provide trained parameters for the neural network, an output patch ready to provide base calling classification data, and other control data for executing the neural network.

The configurable processor is configured to perform an operation of the multi-cycle neural network using the trained parameters to generate classification data for a sensing cycle of the baseflow operation. An operation of the neural network is performed to generate classification data for a subject sensing cycle of a base calling operation. A run of the neural network operates on a sequence (number N arrays comprising block data from corresponding ones of the N sensing cycles, in the examples described herein for each of the time series One base position for one operation provides sensor data for a different base calling operation. Optionally, some of the N sensing cycles may be out of sequence, if desired, depending on the particular neural network model being executed. The number N can be any number greater than one. In some examples described herein, a sensing cycle of the N sensing cycles represents at least one sensing cycle preceding the subject sensing cycle and at least one sensing cycle following the subject cycle in the time series. A set of sensing cycles for the sensing cycle. Examples are described herein where the number N is an integer equal to or greater than five.

Dataflow logic 451 is configured to move the tile data and at least some trained parameters of the model from memory 460 to the A configurable processor for running neural networks. Input cells can be moved by a direct memory access operation in one DMA operation, or in smaller cells that move in coordination with the execution of the deployed neural network during available time slots.

Block data for a sensing cycle as described herein may include an array of sensor data having one or more characteristics. For example, sensor data may include two images that are analyzed to identify one of four bases at a base position in a genetic sequence of DNA, RNA, or other genetic material. Block data may also include metadata about images and sensors. For example, in an embodiment of a base calling operation, the tile data may include information about the alignment of the image to the cluster, such as a distance from the center indicating the distance between each pixel in the sensor data array and the image on the tile. The distance from the center of the cluster of genetic material.

During the execution of the multi-cycle neural network as described below, the block data may also include data generated during the execution of the multi-cycle neural network, called intermediate data, which can be reused during the operation of the multi-cycle neural network without Not a recalculation. For example, during execution of a multi-cycle neural network, dataflow logic may write intermediate data to memory 460 in place of sensor data for a given patch of the tile data array. Embodiments similar to this are described in more detail below.

As shown, a system for analyzing base calling sensor output is described that includes a memory (e.g., 460) accessible by a runtime program that stores block data including Sensor data for a block of a sensing cycle of operation is detected. In addition, the system includes a neural network processor, such as a memory-accessible configurable processor 450 . The neural network processor is configured to perform an operation of the neural network using the trained parameters to generate classification data for the sensing cycle. As described herein, the execution of the neural network operates on a sequence of N arrays of block data from corresponding ones of the N sensing cycles (including the subject cycle) to produce categorical data for the subject cycle . Dataflow logic 451 is provided to move block data and trained parameters from memory to neural network processing using an input unit comprising data from spatially aligned patches of N arrays of corresponding ones of N sensing cycles for the operation of neural networks.

Additionally, a system is described in which a neural network processor has access to memory and includes a plurality of execution clusters, clusters of execution logic in the plurality of execution clusters configured to execute a neural network. The data flow logic is capable of accessing memory and execution clusters of the plurality of execution clusters to provide input units of block data to available execution clusters of the plurality of execution clusters, the input units comprising number N spatially aligned patches of the block data array of the or sensing loop), and cause the execution cluster to apply the N spatially aligned patches to the neural network to produce a spatially aligned patch of the subject sensing loop The output patch of the categorical data, where N is greater than 1.

5 is a simplified diagram illustrating aspects of a base calling operation, including the functionality of a runtime program executed by a host processor. In this figure, the output from an image sensor of a flow cell, such as that shown in FIGS. resampling, alignment, and placement in the sensor data array of the flow cell, and can be used by the process of computing a block cluster mask for each block in the flow cell, which identifies the genetic material on the corresponding block of the flow cell. Clusters correspond to pixels in the sensor data array. To compute the cluster mask, an exemplary algorithm is based on a procedure for detecting clusters that are unreliable in early sequencing cycles using metrics derived from the softmax output, then discarding data from those wells/clusters and not targeting those clusters Generate output data. For example, the process may identify clusters with high reliability during the first N1 (eg, 25) base calls and reject other clusters. Rejected clusters may be polyclonal or very weak in intensity or obscured by fiducials. The program is executable on the host CPU. In an alternative embodiment, this information would potentially be used to identify the necessary clusters of interest to pass back to the CPU, thereby limiting the storage required for intermediate data.

Depending on the status of the base-calling operation, the output of image processing thread 501 is provided on line 502 to scheduling logic 510 in the CPU, which routes the tile data array on high-speed bus 503 to data cache 504, or on high-speed Routed on bus 505 to multi-cluster neural network processor hardware 520 , such as the configurable processor of FIG. 4 . Hardware 520 returns the classified data output by the neural network to dispatch logic 510, which passes the information to data cache 504, or on thread 511 to thread 502 that performs base calling and quality score calculations using the classified data, And data for base calling reads can be laid out in a standard format. The output of thread 502, which performs base calling and quality score calculations, is provided on line 512 to thread 503, which aggregates the base calling reads, performs other operations such as data compression, and writes the resulting base calling output to Enter the designated destination for customers to use.

In some embodiments, the host computer may include threads (not shown) that perform final processing of the output of the hardware 520 to support the neural network. For example, hardware 520 may provide an output of classification data from the final layer of a multi-cluster neural network. The host processor can perform an output activation function, such as a softmax function, on the classification data to configure the data for use by the base calling and quality scoring thread 502 . Additionally, the host processor may perform input operations (not shown), such as resampling, batch normalization, or other adjustments to the tile data prior to input to hardware 520 .

6 is a simplified diagram of the configuration of a configurable processor, such as the configurable processor of FIG. 4 . In Figure 6, the configurable processor includes an FPGA with multiple high-speed PCIe interfaces. The FPGA is configured with a wrapper 600 that includes the data flow logic described with reference to FIG. 1 . Wrapper 600 manages the interface and coordination with the runtime program in the CPU through CPU communication link 609 and manages communication with on-board DRAM 602 (eg, memory 460 ) via DRAM communication link 610 . Data flow logic in encapsulator 600 provides patch data retrieved by traversing the number N cycles of block data arrays on onboard DRAM 602 to cluster 601 and retrieves process data 615 from cluster 601 for delivery back to the board Load DRAM602. Encapsulator 600 also manages data transfer between on-board DRAM 602 and host memory, for both input arrays of block data and output patches of sorted data. The wrapper transfers the patch data on line 613 to the allocated cluster 601 . The wrapper provides trained parameters such as weights and biases to the cluster 601 retrieved from the on-board DRAM 602 on line 612 . The wrapper provides configuration and control data to cluster 601 on line 611 , which is provided from or generated in response to a runtime program on the host via CPU communication link 609 . The cluster may also provide status signals to the encapsulator 600 on line 616, which are used in conjunction with control signals from the host to manage the traversal of the block data array to provide spatially aligned patch data and use the cluster 601's Resource to perform multi-cycle neural networks on patch data.

As noted above, there may be multiple clusters on a single configurable processor managed by wrapper 600 configured for execution on corresponding ones of the multiple patches of tile data. Each cluster can be configured to provide categorical data for base calls in a sensing cycle of a subject using the block data of a plurality of sensing cycles as described herein.

In an example of a system, model data (including kernel data such as filter weights and biases) can be sent from a host CPU to a configurable processor so that the model can be updated according to the number of cycles. As a representative example, a base calling operation may include on the order of hundreds of sensing cycles. In some embodiments, a base calling operation may include paired-end reads. For example, model training parameters may be updated every 20 cycles (or other number of cycles), or according to an update pattern implemented for a particular system and neural network model. In some embodiments that include paired-end reads, wherein the sequence of a given string in a genetic cluster on a block includes a first portion extending down (or up) the string from a first end and a first portion extending from a second end A second portion extending upwards (or downwards) along the string, the trained parameters may be updated in transition from the first portion to the second portion.

In some examples, image data for multiple cycles of the sensing data for a tile may be sent from the CPU to the wrapper 600 . Encapsulator 600 may optionally do some pre-processing and transformation of the sensed data and write the information to on-board DRAM 602 . The input block data for each sensing cycle may comprise an array of sensor data comprising approximately 4000 x 3000 pixels or more per block per sensing cycle, where the two features represent the color of the two images of the block, And one or two bytes per pixel per feature. For an embodiment where the number N is three sensing cycles to be used in each run of the multi-cycle neural network, the array of block data for each run of the multi-cycle neural network can consume approximately Hundreds of megabytes. In some embodiments of the system, the block data also includes an array of DFC data stored once per block, or other types of metadata about the sensor data and the block.

In operation, when a multicycle cluster is available, the wrapper assigns patches to the cluster. The wrapper fetches the next patch of block data in the block's traversal and sends it to the allocated cluster along with appropriate control and configuration information. Clusters can be configured with sufficient memory on configurable processors to hold patches of data that include patches from multiple cycles in some systems and are being processed in-place, and when used in various embodiments The data patch that will be processed when the ping-pong buffering technique or the raster scanning technique completes the processing of the current patch.

The allocated cluster will signal the wrapper when it has completed its run of the neural network for the current patch and produced an output patch. The wrapper will read output patches from the allocated cluster, or alternatively the allocated cluster will push data to the wrapper. The encapsulator will then assemble the output patch to the processed block in DRAM 602 . When the processing of the entire block has completed and the output patch of data has been transferred to DRAM, the encapsulator sends the processed output array of the block back to the host/CPU in the specified format. In some embodiments, on-board DRAM 602 is managed by memory management logic in wrapper 600 . A runtime program controls the sequencing operation to complete the analysis of all arrays of block data for all cycles in a run in a continuous stream, providing real-time analysis.

7 is a diagram of a multi-cycle neural network model that can be implemented using the systems described herein. The example shown in Figure 7 may be referred to as a five-cycle input, one-cycle output neural network. The input to the multi-cycle neural network model included five spatially aligned patches (eg, 700) from the block data array for five sensing cycles for a given block. A spatially aligned patch has the same aligned row and column dimensions (x,y) as the other patches in the set, so that the information relates to the same clusters of genetic material on blocks in the sequence cycle. In this example, the subject patch is the patch from the block data array for cycle K. A set of five spatially aligned patches included patches from cycle K-2 two cycles before the subject's patch, patches from cycle K-1 one cycle before the subject's patch, patches from The patch from cycle K+1 one cycle after the patch from the subject cycle, and the patch from cycle K+2 two cycles after the patch from the subject cycle.

The model includes an isolated stack 701 of layers of the neural network for each of the input patches. Thus, stack 701 receives as input the tile data from the patch of cycle K+2, and is isolated from stacks 702, 703, 704, and 705 such that they do not share input or intermediate data. In some implementations, all of the stacks 710-705 may have the same model and the same trained parameters. In other embodiments, the model and trained parameters may be different in different stacks. Stack 702 receives as input the tile data from the patch of cycle K+1. Stack 703 receives as input the block data from the patches of cycle K. Stack 704 receives as input the tile data from the patch of cycle K-1. Stack 705 receives as input the tile data from the patch of cycle K-2. The layers of the isolation stack each perform a convolution operation of a kernel comprising multiple filters on the layer's input data. As in the above example, patch 700 may include three features. The output of layer 710 may include more features, such as 10 to 20 features. Likewise, the output of each of layers 711-716 may include any number of features suitable for a particular implementation. The parameters of the filter are the trained parameters of the neural network, such as weights and biases. The output feature sets (intermediate data) from each of the stacks 701 to 705 are provided as input to the inverse hierarchy 720 of the temporal combination layer, where intermediate data from multiple cycles are combined. In the illustrated example, the reverse hierarchy 720 includes a first layer comprising three combined layers 721, 722, 723 each receiving intermediate data from three of the isolated stacks and the final layer, which includes a composite layer 730 that receives intermediate data from the three temporal layers 721 , 722 , 723 .

The output of the final combination layer 730 is an output patch of classification data for the clusters located in the corresponding patch from the block of cycle K. The output patches may be assembled into output array sort data for blocks of cycle K. In some implementations, the output patch may have a different size and dimensions than the input patch. In some implementations, the output patch can include pixel-by-pixel data that can be host filtered to select cluster data.

Depending on the particular implementation, the output classification data can then be applied to a softmax function 740 (or other output activation function), optionally executed by the host or on a configurable processor. An output function other than softmax may be used (eg, base calling output parameters are generated from the max output, then base quality is given using a learned non-linear mapping using the context/network output).

Finally, the output of the softmax function 740 can be provided as the base call probability for cycle K (750) and stored in host memory for use in subsequent processing. Other systems may use another function for the output probability calculation, eg, another nonlinear model.

The neural network can be implemented using a configurable processor with multiple execution clusters to complete the evaluation of one block cycle in a duration equal to or close to the time interval of one sensing cycle, effectively providing output data in real time. The dataflow logic may be configured to distribute input units of block data and trained parameters to execution clusters, and distribute output patches for aggregation in memory.

An input unit of data of a five-cycle input, one-cycle output neural network as in FIG. 7 for a base calling operation using two-channel sensor data is described with reference to FIGS. 8A and 8B . For example, for a given base in a genetic sequence, a base calling operation can perform two analysis streams and two reactions that generate two channels of signals (such as images) that can be processed to identify four Which of the bases is located at the current position in the genetic sequence of each cluster of genetic material. In other systems, different numbers of channels of sensed data may be utilized. For example, base calling can be performed using one-pass methods and systems. The incorporated material of US Patent Application Publication No. 2013/0079232 discusses base calling using various numbers of lanes, such as one, two, or four lanes.

FIG. 8A shows a five-cycle block data array for a given block (block M) used for the purpose of executing a five-cycle input, one-cycle output neural network. The five-cycle input block data in this example can be written to on-board DRAM or other memory in the system that is accessible by the dataflow logic, and includes array 801 for channel 1 and array 801 for channel 2 for cycle K-2. Array 811 comprising array 802 for channel 1 and array 812 for channel 2 for cycle K-1, array 803 for channel 1 and array 813 for channel 2 for cycle K, and array 813 for channel 2 for cycle K+1 An array 804 for channel 1 and an array 814 for channel 2 are included, and for cycle K+2 an array 805 for channel 1 and an array 815 for channel 2 are included. Additionally, the array 820 of metadata for the blocks may be written once in memory, in which case a DFC file is included to be used as input to the neural network along with each cycle.

Although FIG. 8A discusses a two-lane base calling operation, the use of two lanes is merely an example, and any other suitable number of lanes may be used to perform base calling. For example, the incorporated material of US Patent Application Publication No. 2013/0079232 discusses base calling using various numbers of lanes, such as one lane, two lanes, or four lanes, or another suitable number of lanes.

The data flow logic constitutes the input units of the block data, which can be understood with reference to FIG. The input patch performs a run of the neural network. The input unit for the allocated execution cluster consists of data flow logic by reading a spatially aligned patch ( For example, 851, 852, 861, 862, 870), and deliver them via a data path (illustratively, 850) to memory on a configurable processor configured for use by the allocated execution cluster. The assigned execution cluster performs a run of the five-cycle-in/one-cycle-out neural network and delivers, for subject cycle K, an output patch of the classification data of the same patch of the block in subject cycle K.

FIG. 9 is a simplified representation of a stack of neural networks that may be used in a system like FIG. 7 (eg, 701 and 720 ). In this example, some functions of the neural network (eg, 900, 902) are executed on the host computer, and other parts of the neural network (eg, 901) are executed on the configurable processor.

For example, the first function may be batch normalization formed on the CPU (layer 910). However, in another example, batch normalization as a function may be fused into one or more layers, and there is no separate batch normalization layer.

As discussed above with respect to the configurable processor, multiple spatially isolated convolutional layers are implemented as the first set of convolutional layers of the neural network. In this example, the first set of convolutional layers apply 2D convolutions spatially.

As shown in FIG. 9 , for the number L/2 (L is described with reference to FIG. 7 ) spatially isolated neural network layers in each stack, a first spatial convolution 921 is performed, followed by a second spatial convolution 922 , after which a third spatial convolution 923 is performed, and so on. As indicated at 923A, the number of spatial layers may be any practical number, which for context may range from a few to more than 20 in different implementations.

For SP_CONV_0, the kernel weights are for example stored in a (1,6,6,3,L) structure, since there are 3 input channels for this layer. In this example, the "6" in the structure is due to storing the coefficients in the transformed Winograd domain (the kernel size is 3x3 in the spatial domain, but expanded in the transform domain).

For this example, for the other SP_CONV layers, the kernel weights are stored in a (1,6,6L) structure, since for each of these layers there are K (=L) inputs and outputs.

The output of the stack of spatial layers is provided to the temporal layers, including convolutional layers 924, 925 executed on the FPGA. Layers 924 and 925 may be convolutional layers that apply 1D convolutions across cycles. As indicated at 924A, the number of temporal layers may be any practical number, which for context may range from a few to more than 20 in different implementations.

The first temporal layer TEMP_CONV_0 layer 824 reduces the number of loop channels from 5 to 3, as shown in FIG. 7 . The second temporal layer (Layer 925) reduces the number of recurrent channels from 3 to 1, as shown in Figure 7, and reduces the number of feature maps to four outputs for each pixel, representing each base call confidence in .

The output of the temporal layers is accumulated in an output patch and delivered to the host CPU to apply, for example, a softmax function 930 or other function to normalize the base call probability.

Figure 10 shows an alternative implementation showing a 10-input, six-output neural network that may be implemented for a base calling operation. In this example, tile data from the spatially aligned input patches of cycles 0 to 9 is applied to an isolated stack of spatial layers, such as stack 1001 of cycle 9 . The output of the isolation stack is applied to the inverse hierarchical arrangement of the time stack 1020 with outputs 1035(2) through 1035(7) to provide base call classification data for cycles 2 through 7 of the subject.

FIG. 11 illustrates one implementation of a specialized architecture of a neural network-based base caller (eg, FIG. 7 ) for isolating the processing of data for different sequencing cycles. First describe the motivation for using a specialized architecture.

The neural network-based base caller processes data for the current sequencing cycle, one or more previous sequencing cycles, and one or more subsequent sequencing cycles. Data from additional sequencing cycles provide sequence-specific context. A neural network-based base caller learns a sequence-specific context during training, and performs base calling on that sequence-specific context. In addition, data from pre- and post-sequencing cycles provide second-order contributions of pre- and phased signals to the current sequencing cycle.

Images captured at different sequencing cycles and in different image channels were misaligned relative to each other and had residual registration errors. To account for this misalignment, the specialized architecture includes spatial convolutional layers that do not mix information between sequencing cycles and only mix information within a sequencing cycle.

The spatial convolutional layers use so-called "isolated convolutions" that achieve isolation by independently processing the data for each of the multiple sequencing cycles via a "dedicated non-shared" convolutional sequence. This isolated convolution convolves only the data and resulting feature maps of a given sequencing cycle (ie, within a cycle), and not the data and resulting feature maps of any other sequencing cycles.

For example, consider that the input data includes (i) the current data for the current (time t) sequencing cycle to be base called, (ii) the previous data for the previous (time t-1) sequencing cycle, and (iii) the previous (time t) sequencing cycle t+1) Subsequent data of the sequencing cycle. Then, the specialized architecture initiates three separate data processing pipelines (or convolutional pipelines), namely the current data processing pipeline, the previous data processing pipeline and the subsequent data processing pipeline. The current data processing pipeline receives as input the current data of the current (time t) sequencing cycle and independently processes this current data through multiple spatial convolutional layers to produce a so-called "current spatial convolutional representation" as the final spatial convolutional layer Output. The previous data processing pipeline receives as input the previous data of the previous (time t-1) sequencing cycle and independently processes this previous data through multiple spatial convolution layers to produce the so-called "previous spatial convolution representation" as the final spatial convolution The output of the stack. The subsequent data processing pipeline receives as input the subsequent data of subsequent (time t+1) sequencing cycles and independently processes this subsequent data through multiple spatial convolution layers to produce a so-called "subsequent spatial convolution representation" as the final spatial convolution layer output.

In some implementations, the current pipeline, one or more previous pipelines, and one or more subsequent processing pipelines execute in parallel.

In some implementations, the spatial convolutional layer is part of a spatial convolutional network (or sub-network) within a specialized architecture.

The neural network-based base caller also includes a temporal convolutional layer that mixes information between sequencing cycles (ie, between cycles). A temporal convolutional layer receives its input from a spatial convolutional network and operates on the spatial convolutional representation produced by the final spatial convolutional layer of the corresponding data processing pipeline.

The inter-loop operability freedom of temporal convolutional layers stems from the fact that misalignment properties are cleaned from spatial convolutional representations by stacking or concatenation of isolated convolutions performed by sequences of spatial convolutional layers, which misalignment Attributes exist in the image data fed as input to a spatial convolutional network.

Temporal convolutional layers use so-called "combined convolutions" that convolve the input channels in subsequent inputs group by group on a sliding window basis. In one implementation, these subsequent inputs are subsequent outputs produced by previous spatial convolutional layers or previous temporal convolutional layers.

In some implementations, the temporal convolutional layer is part of a temporal convolutional network (or sub-network) within a specialized architecture. A temporal convolutional network receives its input from a spatial convolutional network. In one implementation, the first temporal convolutional layer of the temporal convolutional network combines the spatial convolutional representations between sequencing cycles group-by-group. In another implementation, subsequent temporal convolutional layers of the temporal convolutional network combine subsequent outputs of previous temporal convolutional layers.

The output of the final temporal convolutional layer is fed to the output layer which produces the output. The output is used to base call one or more clusters at one or more sequencing cycles.

During forward propagation, specialized architectures process information from multiple inputs in two stages. In the first stage, isolated convolutions are used to prevent information mixing between inputs. In the second stage, combinatorial convolutions are used to mix information between inputs. The results from the second stage are used to make a single inference over the multiple inputs.

This differs from batch-mode techniques where convolutional layers process multiple inputs in a batch simultaneously and make corresponding inferences for each input in the batch. In contrast, a specialized architecture maps the multiple inputs to this single inference. This single inference may include more than one prediction, such as a classification score for each of the four bases (A, C, T, and G).

In one implementation, the inputs are temporally ordered such that each input is generated at a different time step and has multiple input channels. For example, the plurality of inputs may include the following three inputs: the current input generated by the current sequencing cycle at time step (t), the previous input generated by the previous sequencing cycle at time step (t-1), and the input generated by the previous sequencing cycle at time step (t-1). Subsequent inputs generated by subsequent sequencing cycles at time step (t+1). In another implementation, each input is derived from the current output, previous output and subsequent output produced by one or more previous convolutional layers, respectively, and includes k feature maps.

In one specific implementation, each input may include the following five input channels: a red image channel, a red distance channel, a green image channel, a green distance channel, and a scaling channel. In another implementation, each input may include k feature maps produced by previous convolutional layers, and each feature map is considered as an input channel. In yet another example, each input may have only one channel, two channels, or another different number of channels. The incorporated material of US Patent Application Publication No. 2013/0079232 discusses base calling using various numbers of lanes, such as one, two, or four lanes.

Figure 12 shows one implementation of isolation layers, each of which may include convolutions. Isolated convolutions process this multiple inputs by synchronously applying convolution filters once to each input. With isolated convolution, a convolution filter combines input channels in the same input and does not combine input channels in different inputs. In one implementation, the same convolutional filter is applied to each input synchronously. In another implementation, different convolution filters are applied to each input synchronously. In some implementations, each spatial convolutional layer includes a set of k convolutional filters, where each convolutional filter is applied synchronously to each input.

Figure 13A shows one implementation of combined layers, each of which may include convolutions. Figure 13B shows another implementation of combined layers, each of which may include convolutions. Combined convolution mixes information between different inputs by grouping their corresponding input channels and applying a convolution filter to each grouping. The grouping of these corresponding input channels and the application of convolutional filters occurs on a sliding window basis. In this context, a window spans two or more subsequent input lanes, representing eg the output of two subsequent sequencing cycles. Since the window is a sliding window, most input channels are used in two or more windows.

In some implementations, the different inputs originate from output sequences produced by previous spatial convolutional layers or previous temporal convolutional layers. In this output sequence, these different inputs are arranged as subsequent outputs and are thus considered subsequent inputs by subsequent temporal convolutional layers. Then, in the subsequent temporal convolutional layers, the combined convolutions apply convolution filters to corresponding sets of input channels in these subsequent inputs.

In one implementation, these subsequent inputs have a temporal order such that the current input is generated by the current sequencing cycle at time step (t), the previous input is generated by the previous sequencing cycle at time step (t-1), and the subsequent Inputs are generated by subsequent sequencing cycles at time steps (t+1). In another specific implementation, each subsequent input is respectively derived from the current output, previous output and subsequent output produced by one or more previous convolutional layers, and includes k feature maps.

In one specific implementation, each input may include the following five input channels: a red image channel, a red distance channel, a green image channel, a green distance channel, and a scaling channel. In another implementation, each input may include k feature maps produced by previous convolutional layers, and each feature map is considered as an input channel.

The depth B of the convolutional filter depends on the number of subsequent inputs whose corresponding input channels are convoluted by the convolutional filter group by group on a sliding window basis. In other words, the depth B is equal to the number and group size of subsequent inputs in each sliding window.

In FIG. 13A, corresponding input channels from two subsequent inputs are combined in each sliding window, and thus B=2. In Figure 13B, corresponding input channels from three subsequent inputs are combined in each sliding window, and thus B=3.

In one implementation, the sliding windows share the same convolutional filter. In another implementation, a different convolution filter is used for each sliding window. In some implementations, each temporal convolutional layer includes a set of k convolutional filters, where each convolutional filter is applied to subsequent inputs on a sliding window basis.

Further details of Figures 4 through 10 and variations thereof can be found in co-pending U.S. Nonprovisional Patent Application No. 17/176,147, filed February 15, 2021, entitled "HARDWARE EXECUTION AND ACCELERATION OF ARTIFICIAL INTELLIGENCE-BASED BASE CALLER" (Attorney Docket No. ILLM 1020-2/IP-1866-US), which patent application is incorporated herein by reference as if fully set forth herein.

FIG. 14 illustrates an exemplary tile position-based weight selection scheme for base calling. For example, shown in FIG. 14 is an exemplary flow cell 1400 comprising a plurality of channels 1450, each channel comprising a corresponding plurality of blocks (eg, as discussed with respect to FIGS. 1 and 2). The depiction of flow cell 1400 is symbolic in nature, and flow cell 1400 symbolically depicts the various channels and blocks within it, while various other components of flow cell 1400 are not shown. FIG. 14 shows a top view of a flow cell 1400 (eg, without the flow shield 136 of FIG. 1 ).

In one embodiment and also as discussed with respect to FIG. 2 , flow cell 1400 is divided or zoned into a plurality of channels, such as channels 1450a, 1450b, 1450c, . . . , 1450(P-2), 1450(P- 1) and 1450P, ie, P slots, where P is a positive integer. Also as discussed with respect to FIG. 2, in one embodiment, the individual channels 1450 are further partitioned into non-overlapping regions referred to as blocks. In an example, each slot 1450 includes one or more tile columns. For example, in FIG. 14, each channel 1450 includes two corresponding columns of blocks, where a single block in FIG. 14 is shown by a corresponding rectangular box. The number of blocks in each block column in each slot is implementation specific. Each block includes a corresponding number of clusters. During the sequencing process, the clusters on the block and their surrounding background are imaged. For example, Figures 2 and 3 show examples of clusters within a block.

In one embodiment, the sections of the flow cell 1400 are classified into various types, eg, based on the location of the sections. In the exemplary implementation of FIG. 14 , individual ones of the blocks of the flow cell 1400 are classified as edge blocks 1408 , near edge blocks 1410 , or non-edge (or central) blocks 1412 .

For example, blocks on the vertical edges (eg, along the Y-axis) and/or horizontal edges (eg, along the X-axis) of the flow cell 1400 are classified as edge blocks 1408 , as shown in FIG. 14 . Thus, the edge blocks 1408 are immediately adjacent to the corresponding edges of the flow cell 1400 .

Blocks that are close to (eg, immediately adjacent to) edge blocks are classified as near-edge blocks 1410 . For example, near-edge segment 1410 is a segment that is spaced from the edge of flow cell 1400 . Thus, the edge segment 1408 separates the corresponding near-edge segment 1410 from the corresponding edge of the flow cell 1400 .

Blocks that are not edge blocks or near-edge blocks are non-edge blocks 1412 , also referred to as central blocks 1412 . Thus, central block 1412 is relatively closer to the center of flow cell 1400 than, for example, edge block 1408 or near-edge block 1410 . For example, central block 1414 is separated from the edge of flow cell 1400 by edge block 1408 and near edge block 1410 .

Although in FIG. 14 the blocks of the flow cell 1400 are classified into three categories (such as edge, near edge, and central or non-edge), such classifications are merely examples, and different block location-based classifications may also be used. For example, in another implementation, tiles may be categorized as (i) edge or near-edge tiles, and (ii) central tiles (e.g., edge and near-edge tile categories may be combined into a single category) , resulting in two block categories.

As previously discussed, Figures 7 and 10 are exemplary multi-cycle neural network models that can be used for base calling, and Figure 9 is a simplification of a stack of neural networks that can be used in systems such as Figures 7 and 9 express. Bias and weights are used by various functions within a neural network model for base calling. For example, during a convolution operation, a filter comprising one or more kernels (eg, as shown in FIG. 12 ) has a corresponding plurality of weights that are trained during the training phase of the neural network model. For example, the weights are tuned using training data generated from one or more blocks, and these weights are used, for example, for base calling in the flow cell of FIG. 14 .

A base calling cycle is performed for the clusters in each block of the flow cell 1400 . In an example, parameters related to a basecalling operation for a block may be based on the relative position of the block. For example, the excitation light 101 discussed with respect to FIG. Different amounts of excitation light 101 are received. For example, if the light source emitting excitation light 101 is positioned vertically above flow cell 1400 , central block 1412 may receive a different amount of light than edge block 1408 and/or near edge block 1410 .

In another example, ambient or external light around the flowcell 1400 (eg, ambient light from outside the biosensor 100 ) can affect the amount and/or characteristics of the excitation light 101 received by various sections of the flowcell 1400 . By way of example only, edge block 1408 may receive excitation light 101 and some amount of ambient light from outside flow cell 1400 , while central block 1412 may primarily receive excitation light 101 .

In yet another example, individual sensors (or pixels or photodiodes) included in flow cell 1400 (eg, sensors 106, 108, 110, 112, and 114 shown in FIG. Metering, these positions are based on the position of the corresponding block. For example, the sensing operations performed by one or more sensors associated with edge block 1408 may be relatively sensitive compared to the effect of ambient light on the sensing operations of one or more other sensors associated with central block 1412. are more affected by ambient light (and excitation light 101 ).

In another example, the flow of reactants (e.g., including anything that can be used to obtain a desired response during base calling, such as reagents, enzymes, samples, other biomolecules, and buffer solutions) to the various blocks is also May be affected by block location. For example, a block closer to a source of a reactant may receive a larger amount of reactant than a block further from the source.

So, in other words, the parameters associated with base calling may be slightly different for different classes of blocks. Thus, in one embodiment, different sets of weights are used for different classes of blocks to compensate for the exemplary block position dependence of the base calling process discussed above.

For example, in the implementation of FIG. 14, three candidate weight sets are used: (i) edge weight set WeT 1418 for edge blocks, (ii) near-edge weight set WnT 1420 for near-edge blocks, and (iii) Central weight set WcT 1422 for central (or non-edge) edge blocks.

In an example, when training a neural network model for base calling, such as those discussed with respect to FIGS. train on (e.g., do not use training data generated from near-edge or central patches). The resulting weights are included in edge weight set WeT 1418 .

Subsequently, the neural network model is trained on image data generated only by near-edge blocks 1410 (eg, without using training data generated from edge or central blocks), and the resulting weights are included in near-edge weight set WnT 1420 . Finally, the neural network model is trained on image data generated only by the central block 1412 (eg, without using training data generated from edge or near-edge blocks), and the resulting weights are included in edge weight set WcT 1422 .

Accordingly, each set of weights includes a corresponding plurality of weights for configuring a neural network model for processing sensor data from a corresponding class of blocks. For example, as discussed with respect to FIGS. 7, 9, 10, and 11, the topology of the neural network model includes (i) one or more features that do not combine sensor data and resulting feature maps between successive sensing cycles. spatial layers, and (ii) a temporal layer combining the resulting feature maps between successive sensing cycles. Thus, each weight set includes corresponding spatial weights for the spatial layer and corresponding temporal weights for the temporal layer. For example, the set of edge weights WeT 1418 for an edge block includes a corresponding first one or more spatial weights for the spatial layer and a corresponding first one or more temporal weights for the temporal layer. Similarly, the central set of weights WcT 1422 for the central tile includes a corresponding second one or more spatial weights for the spatial layer and a corresponding second one or more temporal weights for the temporal layer.

During the inference phase when performing a base calling loop, when a base within a cluster of an edge block is to be called, the neural network model is configured with the edge weight set WeT 1418, and the sensor data from the edge block is used for the base Base checkout operation. Similarly, when bases are to be called within a cluster of near-margin blocks, the neural network model is configured with the near-margin weight set WnT 1420 and sensor data from the near-margin blocks are used for the base calling operation. Finally, when the bases within the clusters of the central block are to be called, the neural network model is configured with the central weight set WeT 1422, and the sensor data from the central block is used for the base calling operation.

FIG. 15 illustrates another exemplary tile location-based weight selection scheme for base calling. For example, shown in FIG. 15 is a flow cell 1400 comprising the plurality of channels 1450a, 1450b, 1450c, ..., 1450(P-2), 1450(P-1), and 1450P, wherein each channel comprises Corresponding multiple blocks.

In the example of FIG. 15, each segment of the flow cell 1400 is classified based on the location of the corresponding lane to which the segment belongs. For example, the top one or more channels of flow cell 1400 (such as channels 1450P and 1450(P-1)) are classified as top peripheral channels, and the bottom one or more channels of flow cell 1400 (such as channel 1450a and 1450b) are classified as bottom peripheral channels, and the central channel or channels of flowcell 1400, such as channels 1450c and 1450(P-2), are classified as central channels. Note that the number of channels belonging to each category is just an example and variations may be possible. For example, instead of two channels, each peripheral channel category may include one corresponding channel or three corresponding channels, etc.

The blocks within the top peripheral channel are classified as top peripheral channel blocks 1508a, the blocks within the bottom peripheral channel are classified as bottom peripheral channel blocks 1508b, and the blocks within the central channel are classified as central channel regions Block 1510.

For reasons discussed with respect to FIG. 14 , in one embodiment, different sets of weights may be assigned to the blocks within the various classes of channels in the flow cell of FIG. 15 . For example, in the implementation of FIG. 15, two candidate weight sets are used: (i) Perimeter weight set WpL 1504 for perimeter channel blocks 1508a, 1508b (e.g., blocks belonging to top and bottom perimeter channels) , and (ii) the central weight set WcL 1506 for the central channel block 1510.

For example, when training a neural network model for base calling, such as those discussed with respect to FIGS. The neural network model is trained on the image data (eg, without using the training data generated from the central channel block 1510). The resulting weights are included in the perimeter weight set WpL1504.

Subsequently, the neural network model is trained on the image data generated by only the central channel block 1510 (e.g., without using training data generated from the peripheral channel blocks 1508a, 1508b), and the resulting weights are included in the central weight set WcL1506 middle.

During the inference phase when performing a base calling loop, when a base within a cluster of the perimeter lane block 1508 is to be called, the neural network model is configured with weights from the perimeter weight set WpL 1504, and from the perimeter lane The block's sensor data 1508 is used for base calling operations. Similarly, when bases are to be called within clusters of central lane block 1510, the neural network model is configured with weights from central weight set WcL 1506, and sensor data from central lane block 1510 is used for base Checkout operation.

FIG. 16 illustrates yet another exemplary block location-based weight selection scheme for base calling. For example, shown in FIG. 16 is a flow cell 1400 that includes the plurality of channels 1450a, 1450b, 1450c, ..., 1450(P-2), 1450(P-1), and 1450P, wherein each channel includes Corresponding multiple blocks.

In the example of FIG. 16, the flow cell 1400 is divided into segments or sections based on the dashed lines 1603 (ie, the dashed lines 1603 are used for classification and are not actually present in the flow cell). For example, flow cell 1400 is divided into top left section 1610TL (weight set WTL), top central section 1610TC (weight set WTC), top right section 1610TR (weight set WTR), middle left section 1610ML ( weight set WML), central section 1610C (weight set WC), middle right section 1610MR (weight set WMR), bottom left section 1610BL (weight set WML), bottom central section 1610BC (weight set WBC), and Bottom left section 1610BL (weight set WBL). Each block of the flow cell 1400 is classified based on the section to which the block belongs.

For reasons similar to those discussed with respect to FIG. 14 , in one embodiment, blocks within the various sections of FIG. 16 are assigned corresponding sets of weights. For example, in the implementation of FIG. 16, tiles in top left section 1610TL are assigned top left weight set WTL, tiles in top central section 1610TC are assigned top central weight set WTC, and top right Blocks in section 1610TR are assigned top right weight set WTR, blocks in middle left section 1610ML are assigned middle left weight set WML, blocks in central section 1610C are assigned central weight set WC, and blocks in middle left section 1610C are assigned central weight set WC. Blocks in middle right section 1610MR are assigned middle right weight set WMR, blocks in bottom left section 1610BL are assigned bottom left weight set WML, blocks in bottom center section 1610BC are assigned bottom center Weight Set WBC, assigns the bottom left weight set WBL to the blocks in the bottom left section 1610BL.

For example, when training a neural network model for base calling, such as those discussed with respect to FIGS. The neural network model is trained on the generated sensor data (eg, without using sensor data from other classes of blocks), and the resulting weights are included in the top left weight set WTL. This process is repeated for blocks of various other sectors to generate various candidate weight sets, such as top central weight set WTC, top right weight set WTR, middle left weight set WML, central weight set WC, middle right weight set Weight set WMR, bottom left weight set WML, bottom center weight set WBC, and bottom right section weight set WBL.

During the inference phase when performing a base calling loop, when a base is to be called within a cluster of blocks within the top left segment 1610TL, the neuron is configured with weights within the corresponding top left weight set WTL The network model and sensor data from the blocks of the top left section 1610TL are used for base calling operations. This process is similarly repeated for blocks of various other sectors.

In Figure 16, the flow cell 1400 is partitioned into 9 different sections. However, the flow cell 1400 may be partitioned into a different number of sections, such as four sections including a top left quadrant, a top right quadrant, a bottom left quadrant, and a bottom right quadrant.

Figure 17A shows an example of fading in which signal strength decreases with cycle number of a sequencing run as a base calling operation. Fade is the exponential decay of fluorescence signal intensity with cycle number. As the sequencing run progresses, the analyte strands are excessively washed, exposed to laser radiation that produces reactive species, and subjected to harsh environmental conditions. All of this leads to a progressive loss of fragments within each analyte, reducing its fluorescence signal intensity. Fading is also known as dimming or signal decay. An example of fading 1700 is shown in FIG. 17A . In Figure 17A, the intensity values of analyte fragments with AC microsatellites exhibit an exponential decay.

Figure 17B conceptually illustrates the decreasing signal-to-noise ratio as the sequencing cycle progresses. For example, accurate base calling becomes increasingly difficult as sequencing progresses because signal strength decreases and noise increases, resulting in a significantly lower signal-to-noise ratio. Physically, it was observed that later synthesis steps attach labels at different positions relative to the sensor compared to earlier synthesis steps. When the sensor is positioned below the sequence being synthesized, the signal is attenuated due to tags being attached to strands farther from the sensor in later sequencing steps than in earlier steps. This results in signal attenuation as the sequencing cycle progresses. In some designs, where the sensor is located above the substrate holding the clusters, the signal may increase rather than decay as sequencing progresses.

In the flow cell designs studied, as the signal decays, the noise becomes louder. Physically, phasing and prephasing add noise as sequencing proceeds. Phasing refers to the step in sequencing where the tags fail to progress along the sequence. Prephasing refers to the sequencing step in which the index jumps forward two positions instead of one during the sequencing cycle. Phasing and prephasing are relatively infrequent, occurring once in about 500 to 1000 cycles. Phasing is slightly more frequent than pre-phasing. Phasing and prephasing affect the individual strands in the cluster from which the intensity data is generated, so that as sequencing progresses, the intensity noise distribution from the clusters accumulates into binomial, trinomial, quaternary, etc. expansions.

Further details of fading, signal attenuation, and signal-to-noise ratio reduction, and Figures 17A and 17B can be found in a U.S. nonprovisional patent entitled "Systems and Devices for Characterization and Performance Analysis of Pixel-Based Sequencing," filed May 14, 2020 Application No. 16/874,599 (Attorney Docket No. ILLM 1011-4/IP-1750-US), which is incorporated herein by reference as if fully set forth herein.

Thus, during base calling, the reliability or quality of a base call (eg, the probability that a base is called correctly) may be based on the number of base calling cycles for which the current base is being called. Thus, in addition to or instead of depending on the location of the tile (eg, as discussed with respect to FIGS. 14, 15, 16), the set of weights may also be based on the current cycle number for which the base calling operation is being performed. Figure 18 illustrates an exemplary base calling cycle number based weight selection scheme for base calling.

For example, FIG. 18 points to an exemplary block M base calling run. Assume that there are N base calling cycles during which strands in various clusters in exemplary block M will be identified. As discussed, due to the factors discussed with respect to FIGS. 17A and 17B and/or various other factors, the signal strengths detected by the biosensors (e.g., sensors 106, 108, 110, 112, and 114 of FIG. 1 ) Varies (eg, decays) with base calling cycle number. For example, assume that N base-calling sensing cycles are divided into three cycle subseries, such as (a) initial sensing cycles 1 to N1, (b) intermediate sensing cycles (N1+1) to N2, and (c ) The final sensing cycle (N2+1) to N, as shown in FIG. 18, wherein N>N2>N1, and N, N1, N2 are positive integers. Thus, the N sensing cycles are divided into three cycle sub-series, although the N sensing cycles may also be divided into a different number (such as 2, 4 or a larger number) of cycle sub-series.

It should be noted that the number of sensing cycles in each of the above three cycle sub-series may or may not be equal and is implementation specific. As an example only and without limiting the scope of the present disclosure, if N is 100, the 100 cycles may be divided into sub-series comprising 30 initial cycles, 30 intermediate cycles, and 40 final cycles. That is, N1=30 and N2=60 in this simple example.

As discussed with respect to FIGS. 17A and 17B , for example, the average level of signal strength received by the base caller from the biosensor in, for example, cycle number N1 may be different than that received by the base caller in cycle number N1. The average level of signal strength received from the biosensor. Therefore, a neural network model trained for, for example, the number of cycles N1 may not provide satisfactory results for the number of cycles N.

Accordingly, neural network models for base calling, such as those discussed with respect to Figures 7, 9, and 10, can be trained for a particular subset of cycles. For example, a neural network model is initially trained on sensor data generated only during sensing cycles 1 to N1, and the resulting weights are included in the first cycle subseries weight set W(1-N1) 1810a. Subsequently, the neural network model is trained on sensor data generated only during sensing cycles (N1+1) to N2, and the resulting weights are included in the second cycle sub-series weight set W(N1-N2) 1810b. Finally, the neural network model is trained on sensor data generated only during sensing cycles (N2+1) to N, and the resulting weights are included in the third cycle sub-series weight set W(N2-N) 1810c. Note that, for example, in the first cycle sub-series weight set W(1-N1) 1810a, the phrase (1-N1) is the cycle index, which means that the weight set is associated with sensing cycles 1 to N1. It may be noted that in the example of FIG. 18 , base calling is performed using sensor data from one or more lanes, such as one lane, two lanes, three lanes, four lanes, or a greater number of lanes. operation, and for a given loop, weights can be applied to sensor data from all such channels.

During the inference phase, when a base is to be called for rounds 1 to N1, the neural network model is configured with the first round subseries weight set W(1-N1) 1810a. Similarly, when base calls are to be made for the cycle (N1+1) to N2, the neural network model is configured with the second cycle subseries weight set W(N1-N2) 1810b. Finally, when bases are to be called for cycles N2 to N3, the neural network model is configured with the third cycle subseries weight set W(N2-N3) 1810c.

Figures 14, 15, 16 show various examples of weight set selection based on the location of the tile. Accordingly, these figures illustrate various examples of weight set selection based on the spatial progression of the base calling operation through the location of the tiles on the biosensor. FIG. 18 , on the other hand, shows an example of weight set selection based on the time progression of a base calling operation through a sub-series of sensing cycles in a series of sensing cycles 1 to N. FIG. 19 combines the concepts of weight set selection based on spatial block location (e.g., as discussed with respect to FIGS. discussed in Figure 18) combination. Accordingly, FIG. 19 shows an exemplary weight selection scheme based on (i) the temporal progression of base calling cycle numbers and (ii) the spatial location of the blocks.

For example, FIG. 19 shows a first block M1 and a second block M2. Assume that the block M1 is a block of the first type, and the block M2 is a block of the second type. By way of example only, tile M1 may be edge tile 1408 of FIG. 14 , and tile M2 may be central tile 1412 of FIG. 14 . Thus, the set of weights used to base call strands within clusters in block M1 will be different from the set of weights used to base call strands within clusters in block M2, e.g., as opposed to discussed in Figures 14, 15 and 16.

Similar to FIG. 18 , in FIG. 19 it is assumed that there are N base calling cycles during which strands in the various clusters in blocks M1 and M2 will be identified. Furthermore, similar to FIG. 18 , it is assumed in FIG. 19 that the N base calling sensing cycles are divided into three cycle subseries such as (a) initial sensing cycles 1 to N1, (b) intermediate sensing cycles ( N1+1) to N2 and (c) final sensing cycles (N2+1) to N, where N>N2>N1, and N, N1, N2 are positive integers, although in other examples N sensing cycles can also Divided into a different number of cyclic sub-series such as 2, 4 or more.

In an example, a neural network model for base calling, such as those discussed with respect to FIGS. 7, 9, and 10, can be trained for a specific subset of cycles and for a specific block. For example, a neural network model is initially trained on sensor data generated only during sensing cycles 1 to N1 and only for edge tiles 1408, and the resulting set of weights is denoted "weight set (eT,(1-N1))" . Note that the phrase "eT" in this weight set is a block class or block position index, which means that this weight set is specific to edge blocks 1408 . Also, the phrase "(1-N1)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles 1 to N1.

Similarly, the neural network model is then trained on sensor data generated only during sensing cycles (N1+1) to N2 and only for edge tiles 1408, and the resulting weight set is denoted "weight set (eT, (N1 -N2))". Here, the phrase "eT" is a block position or block class index, which means that this set of weights is specific to edge blocks 1408 . Similarly, the phrase "(N1-N2)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles (N1+1) to N2.

Similarly, the neural network model is initially trained on sensor data generated only during sensing cycles (N2+1) to N and only for edge tiles 1408, and the resulting weight set is denoted "weight set (eT, (N2 -N))". Here, the phrase "eT" is the tile position index, which means that this set of weights is specific to edge tiles 1408 . Similarly, the phrase "(N2-N)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles (N2+1) to N.

Furthermore, the neural network model is trained on sensor data generated only during sensing cycles 1 to N1 and only for the central block 1412, and the resulting set of weights is denoted "weight set (cT,(1-N1))". Note that the phrase "cT" in this weight set is a block position index, which means that this weight set is dedicated to the central block 1412 . Also, the phrase "(1-N1)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles 1 to N1.

Similarly, the neural network model is then trained on sensor data generated only during sensing cycles (N1+1) to N2 and only for the central block 1412, and the resulting weight set is denoted "weight set (cT,(N1 -N2))". Here, the phrase "cT" is the tile position index, which means that this set of weights is dedicated to the central tile 1412 . Similarly, the phrase "(N1-N2)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles (N1+1) to N2.

Similarly, the neural network model is initially trained on sensor data generated only during sensing cycles (N2+1) to N and only for the central block 1412, and the resulting weight set is denoted "weight set (cT, (N2 -N))". Here, the phrase "cT" is the tile position index, which means that this set of weights is dedicated to the central tile 1412 . Similarly, the phrase "(N2-N)" in this weight set is a cycle index, which means that this weight set is dedicated to sensing cycles (N2+1) to N.

During the inference phase, when a base is to be called for loops 1 through N1 and for block M1 (e.g., which is edge block 1408 in the example of FIG. 19 ), the set of weights (eT,(1-N1) ) to configure the neural network model. Similarly, when bases are to be called for the cycle (N1+1) to N2 and for the block M1, the neural network model is configured with the weight set (eT, (N1-N2)). In addition, when the base is to be detected for the cycle (N2+1) to N and for the block M1, the neural network model is configured with the weight set (eT, (N2-N)).

Similarly, when a base is to be called for rounds 1 to N1 and for block M2 (e.g., which is central block 1412 in the example of FIG. 19 ), configure with weight set (cT,(1-N1)) neural network model. Similarly, when bases are to be called for the cycle (N1+1) to N2 and for the block M2, the neural network model is configured with the weight set (cT, (N1-N2)). In addition, when the base is to be detected for the cycle (N2+1) to N and for the block M2, the neural network model is configured with the weight set (cT, (N2-N)).

FIG. 20 shows another exemplary weight selection scheme based on (i) temporal progression of base calling cycle numbers and (ii) spatial location of blocks. The block classification shown in FIG. 20 is similar to the block classification shown in FIG. 14 . For example, referring to FIGS. 14 and 20 , edge block 1408 is shown with diagonal lines therein, near-edge block 1410 is shown with cross-hatching therein, and central block 1412 is shown as having dots or shades of gray within it.

Also shown in FIG. 20 are three blocks 1908 , 1910 and 1912 . Referring to block 1908, a set of weights specific to the edge tiles 1408 and various sub-series of sensing cycles is shown. For example, the weight set (eT,(1-N1)) is dedicated to the edge block 1408 and sensing cycles 1 to N1. The set of weights (eT, (N1-N2)) is dedicated to the edge block 1408 and the sensing cycles (N1+1) to N2. The set of weights (eT, (N2-N)) is dedicated to edge blocks 1408 and sensing cycles (N2+1) to N.

Similarly, referring to block 1910, a set of weights specific to the near-edge block 1410 and various sub-series of sensing cycles is shown. For example, the weight set (nT,(1-N1)) is dedicated to near-edge block 1410 and sensing cycles 1 to N1. The set of weights (nT, (N1-N2)) is dedicated to near-edge block 1410 and sensing cycles (N1+1) to N2. The set of weights (nT, (N2-N)) is dedicated to near-edge block 1410 and sensing cycles (N2+1) to N.

Similarly, referring to block 1912, a set of weights specific to the central block 1412 and the various sub-series of sensing cycles is shown. For example, the weight set (cT,(1-N1)) is dedicated to the central block 1412 and sensing cycles 1 to N1. The set of weights (cT, (N1-N2)) is dedicated to the central block 1412 and sensing cycles (N1+1) to N2. The set of weights (cT, (N2-N)) is dedicated to the central block 1412 and sensing cycles (N2+1) to N.

FIG. 21A shows another exemplary weight selection scheme based on (i) temporal progression of base calling cycle number and (ii) spatial location of blocks. The block classification shown in FIG. 21A is similar to the block classification shown in FIG. 15 . For example, referring to FIGS. 15 and 21 , peripheral channel block 1508 (which is a combination of top peripheral channel block 1508 a and bottom peripheral channel block 1508 b of FIG. 15 ) is shown with diagonal lines therein. , and the central channel block 1510 is shown with dashed lines or gray shading.

Also shown in Figure 21A are two blocks 2110 and 2112. Referring to block 2110, a set of weights specific to the perimeter channel block 1508 and the various subseries of sensing cycles is shown. For example, the weight set (pl,(1-N1)) is dedicated to the perimeter channel block 1508 and sensing cycles 1 to N1. The set of weights (p1, (N1-N2)) is dedicated to the perimeter channel block 1508 and the sensing cycles (N1+1) to N2. The set of weights (p1, (N2-N)) is dedicated to the perimeter channel block 1508 and sensing cycles (N2+1) to N.

Similarly, referring to block 2112, a set of weights specific to the central channel block 1510 and various sub-series of sensing cycles is shown. For example, the weight set (cl,(1-N1)) is dedicated to the central channel block 1510 and sensing cycles 1 to N1. The set of weights (cl,(N1-N2)) is dedicated to the central channel block 1510 and the sensing cycles (N1+1) to N2. The weight set (cl,(N2-N)) is dedicated to the central channel block 1510 and the sensing cycles (N2+1) to N.

In one embodiment and as discussed above, weight set (pl, (1-N1)), weight set (pl, (N1-N2)), weight set (pl, (N2-N)), weight set Each of (cl, (1-N1)), weight set (cl, (N1-N2)), weight set (cl, (N2-N)) includes a corresponding weight. For example, the set of weights (pl,(1-N1)) includes a first plurality of weights for configuring a corresponding plurality of spatial and temporal layers (see, for example, FIGS. 7 and 9 for examples of such layers), weights The set (pl, (N1-N2)) includes a second plurality of weights for configuring the corresponding multiple spatial layers and temporal layers, and the weight set (pl, (N2-N)) includes the second multiple weights for configuring the corresponding multiple spatial layers The third plurality of weights of layers and time layers, the weight set (cl, (1-N1)) includes a fourth plurality of weights for configuring corresponding multiple spatial layers and time layers, and the weight set (cl, (N1- N2)) includes a fifth plurality of weights for configuring the corresponding plurality of spatial layers and temporal layers, and the weight set (cl, (N2-N)) includes a fifth plurality of weights for configuring the corresponding plurality of spatial layers and temporal layers More than six weights.

At least one weight of the first plurality of weights is different from a corresponding weight of the second plurality of weights (in some examples, both sets of weights may have one or more common or identical weights). At least one weight in the second plurality of weights is different from a corresponding weight in the third plurality of weights, at least one weight in the third plurality of weights is different from a corresponding weight in the fourth plurality of weights, and so on. In one embodiment, one or more weights in the various weight sets are quantized using different scaling factors.

Because various sets of weights are associated with corresponding sequencing cycles, in examples, the weights in the various sets of weights correspond to various sequencing chemistries, sequencing configurations, and/or sequencing assays, respectively. For example, weight set (pl,(1-N1)), weight set (pl,(N1-N2)) and weight set (pl,(N2-N)) correspond to the first sequencing chemistry, the second sequencing chemistry and A third sequencing chemistry (eg, they are used during sequencing cycles 1 to N1, (N1+1) to N2, and (N2+1) to N, respectively. Weight set (pl, (1-N1)), weight set (pl ,(N1-N2)) and weight set (pl,(N2-N)) correspond to the first sequencing determination, the second sequencing determination and the third sequencing determination respectively. Weight set (pl,(1-N1)), weight The set (pl, (N1-N2)) and the weight set (pl, (N2-N)) correspond to the first sequencing configuration, the second sequencing configuration, and the third sequencing configuration, respectively.

FIG. 21B shows yet another exemplary weight selection scheme based on (i) temporal progression of base calling cycle numbers and (ii) spatial location of blocks. The block classification shown in FIG. 21B is similar to the block classification shown in FIG. 16 . For example, referring to FIGS. 16 and 21B , the flow cell 1400 is divided into a top left section 1610TL, a top central section 1610TC, a top right section 1610TR, a middle left section 1610ML, a central section 1610C, a middle right section Section 1610MR, bottom left section 1610BL, bottom central section 1610BC, and bottom left section 1610BL. Each block of the flow cell 1400 is classified based on the section to which the block belongs.

FIG. 21B also shows a table 2150 that includes various weights for the blocks of the various sectors and for the various sub-series of sensing cycles 1-N. For example, referring to the first row of table 2150, the weight set (TL,(1-N1)) is dedicated to the blocks and sensing cycles 1 to N1 of the top left section 1610TL. The set of weights (TL,(N1-N2)) is specific to the block and sensing loops (N1+1) to N2 of the top left section 1610TL. The set of weights (TL,(N2-N)) is specific to the blocks and sensing cycles (N2+1) to N of the top left section 1610TL.

Similarly, referring to the second row of table 2150, the weight set (TC,(1-N1)) is specific to the block and sensing cycles 1 to N1 of the top central section 1610TC. The set of weights (TC, (N1-N2)) is specific to the blocks of the top center section 1610TC and the sensing cycles (N1+1) to N2. The set of weights (TC, (N2-N)) is specific to the block and sensing cycles (N2+1) to N of the top central section 1610TC. Similarly, various other rows of table 2150 include blocks for various other segments and weight sets for various subseries of sensing cycles, and will be apparent to those skilled in the art based on the above discussion. .

FIG. 22 illustrates an implementation of a base calling operation 2200 in which a set of weights for base calling is selected based on spatial block information and temporal sensing cycle subseries information.

For the base calling operation 2200 of Figure 22, assume that the blocks of the flow cell 1400 are sorted according to the examples of Figures 15 and 21A. Such block classification is not intended to limit the scope of the present disclosure, and the base calling operation 2200 may also be applied to any other type of block classification, such as discussed with respect to FIGS. 14, 16, 20, 21B Any block classification and/or any other block classification contemplated by those skilled in the art based on the teachings of this disclosure.

In addition, for the base calling operation 2200 of FIG. 22 , it is assumed that N sensing cycles are divided into three cycle subseries, including (a) cycles 1 to N1, (b) cycles (N1+1) to N2, and ( c) Cycle (N2+1) to N, as discussed with respect to Figures 18-21B. Again, such sensing cycle divisions are not intended to limit the scope of the present disclosure, and the base calling operation 2200 is also applicable to any other type of sensing cycle subdivisions that one skilled in the art can conceive based on the teachings of this disclosure. .

In FIG. 22, base calling operations 1a to 6a are dedicated to peripheral lane blocks and cycles 1 to N1. Similarly, base calling operations 1b to 6b are dedicated to the central lane block and cycles 1 to N1. Operations 1a to 6a and 1b to 6b can be repeated for cycles (N1+1) to N2, and can be further performed for cycles (N2+1) to N repeats, but such repeats are not shown in detail in FIG. 22 . Such repetitions for cycles (N1+1) to N2, and further for cycles (N2+1) to N, will be understood by those skilled in the art based on the discussion of operations 1a to 6a and 1b to 6b for cycles 1 to N1.

At act la, dataflow logic 451 (see, e.g., FIG. 4 ) receives cluster sensor data and weight set (pl,(1-N1)) for perimeter channel block 1508 and for cycle 1 to N1 (see Figure 21A). The cluster data includes a sequencing image that depicts the intensity emission of the clusters within the peripheral lane block 1508 at sequencing cycles 1 through N1 of the sequencing run, as described above. At action 2a, the dataflow logic 451 forwards the cluster data and weight set (p1,(1-N1)) for the perimeter slot block 1508 and for cycles 1 to N1 to be processed by the configurable processor 450 (e.g. , see FIG. 4) implemented neural network based base caller 2308 (eg, examples of which are shown in FIG. 7, FIG. 9, FIG. 10). The set of cluster data and weights (p1,(1-N1)) for the peripheral channel block 1508 and for round 1 to N1 are loaded in the neural network based base caller 2308. Also, although not shown in FIG. 22 , the topology of the neural network model is also loaded from memory to configurable processor 450 via data flow logic 451 .

At action 3a, the configurable processor 450 configures the topology of the neural network running on the configurable processor 450 with the loaded set of weights (pl,(1-N1)). The neural network based base caller 2308 configured with the loaded set of weights (pl,(1-N1)) generates a representation from the cluster data based on the loaded set of weights (pl,(1-N1)) (e.g., feature map) (e.g., by virtue of spatial and temporal convolutional layers configured through it to process cluster data), and based on the representation generate bases for multiple clusters within the perimeter lane block 1508 and for sequencing cycles 1 through N1 Base call classification data (eg, base call classification scores). For example, the neural network based base caller 2308 applies the loaded set of weights (pl,(1-N1)) to the cluster data to generate base calling classification data. In one implementation, the base call classification scores are not normalized, eg, they are not subjected to exponential normalization by a softmax function.

At act 4a, the configurable processor 450 sends to the dataflow logic 451 base calling sort data for clusters within the perimeter lane block 1508 and for cycles 1 to N1. At act 5a, the dataflow logic 451 provides the host processor 2304 with base call sort scores for clusters within the perimeter lane block 1508 and for rounds 1 through N1.

At act 6a, the host processor 2304 normalizes the unnormalized base call classification scores (e.g., by applying a softmax function, block 740 of FIG. 7 or block 930 of FIG. 9 ), and generates Strands within clusters of the channel block 1508 and used for normalized base call classification scores, ie, base calls, for rounds 1 to N1.

Thus, in operations 1a to 6a, using the set of weights (pl,(1-N1)) trained specifically for the peripheral channel block 1508 and for cycles 1 to N1, the system and base calling for cycles 1 to N1. Note that operations la through 6a depict advanced and simplified versions of base calling operations and may not illustrate one or more other operations that may be performed for base calling. Additional details of base calling operations can be found in U.S. Provisional Patent Application No. 63/072,032, filed August 28, 2020, entitled "DETECTING AND FILTERING CLUSTERS BASED ON ARTIFICIALINTELLIGENCE-PREDICTED BASE CALLS," (Attorney Docket No. ILLM 1018 -1/IP-1860-PRV), which patent application is incorporated herein by reference as if fully set forth herein.

Operations la through 6a are dedicated to base calling for strands within clusters of peripheral lane block 1508 and for cycles 1 through N1. These operations are repeated as operations lb-6b, but for the clusters within the central channel block 1510 and for rounds 1-N1. For example, at act 1b, the dataflow logic 451 receives the cluster data and weight set (cl,(1-N1)) for the central slot block 1510 and for cycles 1 to N1 (see FIG. 21A ). The cluster data includes a sequencing image that depicts the intensity emission of the clusters within the central lane block 1510 at sequencing cycles 1 through N1 of the sequencing run, as described above. At action 2b, the dataflow logic 451 forwards the cluster data and weight set (cl,(1-N1)) for the central slot block 1508 and for cycles 1 to N1 to the configurable processor 450 executing Neural network based base caller 2308. The set of cluster data and weights (cl,(1-N1)) for the central lane block 1510 and for round 1 to N1 are used to reconfigure the neural network based base caller 2308.

At act 3b, the reconfigurable neural network based base caller 2308 running on the configurable processor 450 generates an initial representation (e.g., a feature map) from the cluster data (e.g., by virtue of passing through its spatial and temporal volumes cluster data) and generate base call classification scores for multiple clusters within the central lane block 1510 and for sequencing cycles 1 through N1 based on the initial intermediate representation. In one implementation, the initial base call classification scores are not normalized, eg, they are not subjected to exponential normalization by a softmax function.

At act 4b, the configurable processor 450 sends to the dataflow logic 451 the base call sort scores for the clusters within the central lane block 1510 and for rounds 1 to N1. At act 5b, the dataflow logic 451 provides the host processor 2304 with base call sort scores for the clusters within the central lane block 1510 and for rounds 1 through N1.

At act 6b, the host processor 2304 normalizes the unnormalized base call classification scores (e.g., by applying a softmax function) and generates chains within clusters for the central lane block 1510 and uses Normalized base call classification scores at cycles 1 to N1, ie, base calls.

Thus, base calling operations la through 6a are dedicated to peripheral lane block 1508 and cycles 1 through N1. Similarly, base calling operations 1b through 6b are dedicated to central lane block 1510 and cycles 1 through N1. Operations 1a through 6a and 1b through 6b are repeated for cycles (N1+1) through N2, and further for cycles ( N2+1) to N repeats, as symbolically shown in FIG. 22 .

Referring back to FIG. 7 , the model shown includes isolated stacks 701 , 702 , 703 , 704 , 705 . Stack 701 receives as input the tile data from the patch of cycle K+2. Stack 702 receives as input the tile data from the patch of cycle K+1. Stack 703 receives as input the block data from the patches of cycle K. Stack 704 receives as input the tile data from the patch of cycle K-1. Stack 705 receives as input the tile data from the patch of cycle K-2. The layers of the isolation stack each perform a convolution operation of a kernel comprising multiple filters on the layer's input data. The output feature sets (intermediate data) from each of the stacks 701 to 705 are provided as input to the inverse hierarchy 720 of the temporal combination layer, where intermediate data from multiple cycles are combined.

Thus, as discussed with respect to FIGS. 7 , 9 and 11 , stacks 701 , . . . , 705 perform isolated spatial convolutions. There is no temporal mixing or interaction between inputs from the various loops within the various stacks 701 , . . . 705 . Finally, following the data processing in the stack 701 , . . . , 705 , in section 720 there is processing of the data from the various sequencing cycles. The various layers within the stack 701 , . . . , 705 are also referred to herein as spatial layers, and the weights of the kernels of the various filters within the stack 701 , . . . , 705 are referred to herein as spatial weights. Similarly, the various layers within section 720 are also referred to herein as temporal layers, and the weights of the kernels of the various filters within section 720 are also referred to herein as temporal weights. For example, the weights applied during the spatial convolutions 921 , 922 , 923 in FIG. 9 are spatial weights, while the weights applied during the temporal convolutions 924 , 925 in FIG. 9 are temporal weights.

Figure 23A shows various sets of weights for various classes of blocks and for various sensing cycles, each set of weights including corresponding spatial weights and corresponding temporal weights. The block classification shown in Figure 23A is similar to the block classification discussed with respect to Figures 15 and 21A. As discussed with respect to FIG. 21A , the peripheral channel block 1508 for cycles 1 through N1 is associated with a corresponding weight set (p1,1-N1). As shown in FIG. 23A, the weight set (pl, 1-N1) includes a corresponding spatial weight (s-pl, (1-N1)) and a corresponding temporal weight (t-pl, (1-N1)). Spatial weights (s-pl,(1-N1)) are used to configure the spatial layer of the neural network model when the neural network model is used to process cluster sensor data for the perimeter channel block 1508 for cycles 1 to N1 . When the neural network model is used to process cluster sensor data for the perimeter channel block 1508 for cycle 1 to N1, the temporal weights (t-pl,(1-N1)) are used to configure the temporal layer of the neural network model .

Similarly, as also discussed with respect to FIG. 21A , perimeter channel blocks 1508 for cycles N1 through N2 are associated with corresponding weight sets (pl, N1-N2). As shown in FIG. 23A , the set of weights (pl, N1-N2) includes corresponding spatial weights (s-pl, (N1-N2)) and corresponding temporal weights (t-pl, (N1-N2)). The various other weight sets of FIG. 23A similarly have corresponding spatial and temporal weights.

Figure 23B shows various sets of weights for various classes of tiles and for various cycles, where the different sets of weights for a particular class of tiles include common spatial weights and different temporal weights. The block classification shown in Figure 23A is similar to the block classification discussed with respect to Figures 15, 21A, and 23A. However, unlike FIG. 23A , in FIG. 23B the weight sets (pl,(1-N1)), (pl,(N1-N2)) and (pl,(N2-N2)) for the perimeter channel block 1508 )) have common spatial weights (s-pl). Thus, the same or common spatial weight (s-pl) is used for the perimeter channel block 1508 and for each cycle in the cycle subseries 1 to N1, (N+1) to N2, and (N2+1) to N sub-series.

Weight sets (pl, (1-N1)), (pl, (N1-N2)) and (pl, (N2-N)) have different time weights, such as time weights (t-pl, (1-N2) respectively N1)), time weight (t-pl, (N1-N2)) and time weight (t-pl, (N2-N)).

Similarly, the weight sets (cl,(1-N1)), (cl,(N1-N2)) and (cl,(N2-N)) for the central channel block 1510 have common spatial weights (s- cl). Thus, the same or common spatial weight (s-cl) is used for the central channel block 1510 and for each cycle in the cycle subseries 1 to N1, (N+1) to N2, and (N2+1) to N sub-series.

Weight sets (cl,(1-N1)), (cl,(N1-N2)) and (cl,(N2-N)) have different time weights, such as time weights (t-cl,(1-N2) respectively N1)), time weight (t-cl, (N1-N2)) and time weight (t-cl, (N2-N)).

In one embodiment and as discussed with respect to FIGS. 17A and 17B , fading, phasing, and/or prephasing cause degradation of sensor data as the sequencing cycle progresses. Such degradations are addressed by temporal layers of the neural network model, such as layers within block 720 of FIG. 7 or layers 924, 925 of FIG. 9 . Thus, in Figure 23B, the time weights for the various subseries of sequencing cycles are trained differently. For example, the time weight for a given block class for cycles 1 to N1 is different from the time weight for the same block class for cycles N1 to N2. In contrast, because spatial layers (such as the layers within blocks 701, . . . , 705 of FIG. Common space weights for block categories, as shown in Figure 23B.

Thus, when processing sensor data for a particular block class (say, for the peripheral channel block 1508), the set of weights (pl,(1-N1 )), and configure the neural network based base caller 2308 with these spatial and temporal weights. For example, the spatial layer of the neural network based base caller 2308 is configured with common spatial weights (s-pl) and the neural network based base calling is configured with temporal weights (t-pl,(1-N1)) The time layer of the device 2308. The configured neural network based base caller 2308 applies the configured spatial and temporal layers to the sensor data for the peripheral channel block 1508 for cycles 1 to N1 to generate the peripheral channel block 1508 for Base calling classification data for cycles 1 to N1.

Subsequently, the temporal weights (t-pl,(N1-N2)) of the weight set (pl,(N1-N2)) are loaded without loading the weight set before processing the sensor data for the loop (N1+1) Any corresponding spatial weights for . The temporal layers of the neural network based base caller 2308 are configured with temporal weights (t-pl,(N1-N2)). The neural network based base caller 2308 then applies a previously configured spatial layer (e.g., which was previously assigned with the common spatial weights (s -pl) configuration) and a reconfigured temporal layer (e.g., it is reconfigured with temporal weights (t-pl,(N1-N2))) to generate peripheral channel block 1508 for cycle (N1+1) Base call classification data to N2.

Subsequently, the temporal weights (t-pl,(N2-N)) of the weight set (pl,(N2-N)) are loaded without loading the weight set before processing the sensor data for the loop (N2+1) Any corresponding spatial weights for . The temporal layers of the neural network based base caller 2308 are reconfigured with temporal weights (t-pl,(N2-N)). The neural network-based base caller 2308 then applies a previously configured spatial layer (e.g., which was previously assigned with the common spatial weights (s− pl) configuration) and a reconfigured temporal layer (e.g., it is reconfigured with temporal weights (t-pl, (N2-N))) to generate surrounding channel blocks for looping (N2+1) to N base call classification data for .

Base call classification data for other block classes, such as central lane block 1510, is generated in a correspondingly similar manner, which will be understood by those skilled in the art based on the above discussion and the illustration of Figure 23B.

FIG. 23C illustrates a system 2300 for selecting a set of weights based on one or more sequencing run parameters 2382 . For example, weight set selection logic 2386 that may be executed on configurable processor 450 and/or host processor 2304 is shown. Weight set selection logic 2386 receives one or more sequencing run parameters 2382 and one or more other weight set selection criteria discussed with respect to FIGS. 14-23B . Weight set selection logic 2386 selects from a plurality of candidate weight sets 2384a, ..., 2384N based on one or more sequencing run parameters 2382 and/or one or more other weight set selection criteria discussed with respect to FIGS. set of weights. In the example of FIG. 23B, weight set selection logic 2386 selects weight set 2384b. The selected set of weights is then loaded in the configurable processor 450 and used to configure the neural network topology for base calling, as discussed herein.

One or more sequencing run parameters 2382 may include one or more appropriate parameters associated with the current sequencing run. For example, reaction components used in a sequencing run (such as reagents, enzymes, samples, other biomolecules, and buffer solutions) can affect sensor data, and weight sets can be selected based on the type, parameters, or batch of reaction components used . For example, phasing characteristics (see FIG. 17B ) can be based on the reagent pack used for the sequencing run, and can vary based on the type, age, and/or lot of the reagent pack. Accordingly, various candidate weight sets can be generated for various types of batches of reaction components, and the weight set selection logic 2386 can select a weight set based on the reaction components used for the current sequencing cycle.

In another example, the weight set selection logic 2386 can estimate the phasing characteristics and select a weight set based on the phasing characteristics. For example, different sets of weights can be generated for different phasing features. Then early in the sequencing run, phasing parameters can be estimated and used to select weight sets. In yet another example, multiple candidate weight sets can be tried, and the weight set with the lowest error rate (or highest signal-to-noise ratio) can be selected for the entire sequencing run.

Figure 24 is a block diagram of a base calling system 2400 according to one implementation. The base calling system 2400 is operable to obtain any information or data related to at least one of a biological substance or a chemical substance. In some implementations, base calling system 2400 is a workstation that can resemble a desktop device or a desktop computer. For example, most (or all) of the systems and components used to perform the desired reactions may be located within a common housing 2416 .

In certain implementations, base calling system 2400 is a nucleic acid sequencing system (or sequencer) configured for various applications including, but not limited to, de novo sequencing, resequencing of whole genomes or targeted genomic regions, and Metagenomics. Sequencers can also be used for DNA or RNA analysis. In some implementations, base calling system 2400 can also be configured to generate reactive sites in a biosensor. For example, base calling system 2400 can be configured to receive a sample and generate surface-attached clusters of clonally amplified nucleic acid derived from the sample. Each cluster can constitute or be part of a reactive site in a biosensor.

Exemplary base calling system 2400 can include a system socket or interface 2412 configured to interact with biosensor 2402 to perform desired reactions within biosensor 2402 . In the description below with respect to FIG. 24 , biosensor 2402 is loaded into system socket 2412 . It should be understood, however, that a cartridge including biosensor 2402 may be inserted into system receptacle 2412 and, in some cases, removed temporarily or permanently. As noted above, the cartridge may include, among other things, a fluid control component and a fluid storage component.

In a particular implementation, base calling system 2400 is configured to perform a large number of parallel reactions within biosensor 2402 . Biosensor 2402 includes one or more reaction sites where a desired reaction can occur. Reaction sites may, for example, be immobilized to the solid surface of the biosensor or to beads (or other movable substrates) located within corresponding reaction chambers of the biosensor. A reaction site can include, for example, a cluster of clonally amplified nucleic acids. Biosensor 2402 may include a solid-state imaging device (eg, a CCD or CMOS imaging device) and a flow cell mounted thereto. A flow cell may include one or more flow channels that receive a solution from the base calling system 2400 and direct the solution to a reaction site. Optionally, biosensor 2402 may be configured to engage a thermal element for transferring thermal energy into or out of the flow channel.

Base calling system 2400 may include various components, assemblies and systems (or subsystems) that interact with each other to perform a predetermined method or assay protocol for biological or chemical analysis. For example, base calling system 2400 includes a system controller 2404 that can communicate with various components, components, and subsystems of base calling system 2400 and biosensor 2402 . For example, in addition to system socket 2412, base calling system 2400 may also include: fluid control system 2406 for controlling the flow of fluids throughout the fluidic network of base calling system 2400 and biosensors 2402 a fluid storage system 2408 configured to hold all fluids (eg, gases or liquids) that can be used by the bioassay system; a temperature control system 2410 that can regulate the fluid network, fluid storage system 2408, and and/or the temperature of the fluid in the biosensor 2402; and an illumination system 2409 configured to illuminate the biosensor 2402. As noted above, if a cartridge with biosensor 2402 is loaded into system receptacle 2412, the cartridge may also include fluid control components and fluid storage components.

As also shown, the base calling system 2400 can include a user interface 2414 for interacting with a user. For example, user interface 2414 may include a display 2413 for displaying or requesting information from a user and a user input device 2415 for receiving user input. In some implementations, the display 2413 and the user input device 2415 are the same device. For example, user interface 2414 may include a touch-sensitive display configured to detect the presence of an individual touch and also identify the location of the touch on the display. However, other user input devices 2415 may be used, such as mice, touch pads, keyboards, keypads, handheld scanners, voice recognition systems, motion recognition systems, and the like. As will be discussed in more detail below, base calling system 2400 can communicate with various components including biosensor 2402 (eg, in the form of a cartridge) to perform desired reactions. The base calling system 2400 can also be configured to analyze data obtained from biosensors to provide desired information to the user.

System controller 2404 may include any processor-based or microprocessor-based system, including those using microcontrollers, reduced instruction set computers (RISCs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), logic circuits and any other circuit or processor capable of performing the functions described herein. The above examples are exemplary only, and thus are not intended to limit in any way the definition and/or meaning of the term system controller. In an exemplary implementation, the system controller 2404 executes a set of instructions stored in one or more storage elements, memories, or modules to at least one of obtain detection data and analyze detection data. The detection data may include multiple pixel signal sequences such that a pixel signal sequence from each of the millions of sensors (or pixels) may be detected over many base calling cycles. The storage element may be in the form of an information source or a physical memory element within the base calling system 2400 .

The set of instructions may include various commands that instruct base calling system 2400 or biosensor 2402 to perform specific operations, such as the various implemented methods and processes described herein. A set of instructions may be in the form of a software program that may form part of one or more tangible non-transitory computer readable media. As used herein, the terms "software" and "firmware" are interchangeable and include any computer program stored in memory for execution by a computer, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The memory types described above are exemplary only, and thus do not limit the types of memory that can be used to store computer programs.

Software can be in various forms, such as system software or application software. Furthermore, software may be in the form of a collection of stand-alone programs, or a program module within a larger program, or a portion of a program module. The software may also include modular programming in the form of object-oriented programming. After the assay data is obtained, the assay data can be processed by the base calling system 2400 automatically, in response to user input, or in response to a request from another processing machine (eg, remotely via a communication link). In the illustrated implementation, the system controller 2404 includes an analysis module 2538 (shown in FIG. 25 ). In other implementations, the system controller 2404 does not include the analysis module 2538, but can access the analysis module 2538 (eg, the analysis module 2538 can be hosted separately on the cloud).

System controller 2404 may be connected to biosensor 2402 and other components of base calling system 2400 via communication links. System controller 2404 may also be communicatively coupled to off-site systems or servers. Communication links can be hardwired, wired or wireless. System controller 2404 may receive user input or commands from user interface 2414 and user input device 2415 .

Fluid control system 2406 includes a fluid network and is configured to direct and regulate the flow of one or more fluids through the fluid network. The fluid network may be in fluid communication with biosensor 2402 and fluid storage system 2408 . For example, selected fluids may be drawn from fluid storage system 2408 and directed to biosensor 2402 in a controlled manner, or fluid may be drawn from biosensor 2402 and directed toward a waste reservoir in fluid storage system 2408, for example. Although not shown, fluid control system 2406 may include flow sensors that detect the flow rate or pressure of fluid within the fluid network. The sensors may communicate with the system controller 2404.

Temperature control system 2410 is configured to regulate the temperature of fluid at various regions of fluid network, fluid storage system 2408 , and/or biosensor 2402 . For example, temperature control system 2410 may include a thermal cycler that interfaces with biosensor 2402 and controls the temperature of fluid flowing along reaction sites in biosensor 2402 . Temperature control system 2410 may also regulate the temperature of solid elements or components of base calling system 2400 or biosensor 2402 . Although not shown, temperature control system 2410 may include sensors for detecting the temperature of fluid or other components. The sensors may communicate with the system controller 2404.

Fluid storage system 2408 is in fluid communication with biosensor 2402 and can store various reaction components or reactants for performing desired reactions therein. Fluid storage system 2408 may also store fluids for washing or cleaning the fluid network and biosensors 2402 and for diluting reactants. For example, fluid storage system 2408 may include various reservoirs to store samples, reagents, enzymes, other biomolecules, buffered solutions, aqueous and non-polar solutions, and the like. Additionally, fluid storage system 2408 may also include a waste reservoir for receiving waste from biosensor 2402 . In implementations that include a cartridge, the cartridge may include one or more of a fluid storage system, a fluid control system, or a temperature control system. Accordingly, one or more components described herein in relation to those systems may be housed within the cartridge housing. For example, a cartridge may have various reservoirs to store samples, reagents, enzymes, other biomolecules, buffer solutions, aqueous and non-polar solutions, waste, and the like. Accordingly, one or more of the fluid storage system, fluid control system, or temperature control system may be removably engaged with the bioassay system via a cartridge or other biosensor.

The illumination system 2409 may include a light source (eg, one or more LEDs) and a plurality of optical components for illuminating the biosensor. Examples of light sources may include lasers, arc lamps, LEDs, or laser diodes. Optical components may be, for example, reflectors, dichroic mirrors, beam splitters, collimators, lenses, filters, wedges, prisms, mirrors, detectors, and the like. In implementations using an illumination system, the illumination system 2409 can be configured to direct excitation light to the reaction sites. As an example, a fluorophore may be excited by light at a green wavelength, so the wavelength of the excitation light may be approximately 532 nm. In one implementation, the illumination system 2409 is configured to produce illumination parallel to the surface normal of the surface of the biosensor 2402 . In another implementation, the illumination system 2409 is configured to produce illumination at an off-angle relative to the surface normal of the surface of the biosensor 2402 . In yet another implementation, the lighting system 2409 is configured to generate lighting with multiple angles, including some parallel lighting and some off-angle lighting.

System socket or interface 2412 is configured to at least one of mechanically, electrically, and fluidly engage biosensor 2402 . System receptacle 2412 can hold biosensor 2402 in a desired orientation to facilitate fluid flow through biosensor 2402 . System socket 2412 may also include electrical contacts configured to engage biosensor 2402 such that base calling system 2400 may communicate with and/or provide power to biosensor 2402 . Additionally, system receptacle 2412 may include a fluid port (eg, a nozzle) configured to engage biosensor 2402 . In some implementations, biosensor 2402 is removably coupled to system socket 2412 mechanically, electrically, and fluidically.

Additionally, the base calling system 2400 can communicate remotely with other systems or networks or with other bioassay systems 2400 . Detection data obtained by the biometric system 2400 may be stored in a remote database.

FIG. 25 is a block diagram of a system controller 2404 that may be used in the system of FIG. 24 . In one implementation, the system controller 2404 includes one or more processors or modules that can communicate with each other. Each of the processors or modules may include an algorithm (eg, instructions stored on a tangible and/or non-transitory computer-readable storage medium) or sub-algorithm for performing a particular process. The system controller 2404 is shown conceptually as a collection of modules, but may be implemented with any combination of dedicated hardware boards, DSPs, processors, and the like. Alternatively, system controller 2404 may be implemented using an off-the-shelf PC with a single processor or multiple processors, with functional operations distributed among the processors. As a further option, the modules described below may be implemented using a hybrid configuration, where some of the modular functions are performed using dedicated hardware, while the remaining modular functions are performed using off-the-shelf PCs or the like. A module may also be implemented as a software module within a processing unit.

During operation, communication port 2520 may transmit information (eg, commands) or receive information (eg, data) from biosensor 2402 ( FIG. 24 ) and/or subsystems 2406 , 2408 , 2410 ( FIG. 24 ). In a specific implementation, the communication port 2520 may output a plurality of pixel signal sequences. Communication port 2520 may receive user input from user interface 2414 ( FIG. 24 ) and transmit data or information to user interface 2414 . Data from biosensors 2402 or subsystems 2406, 2408, 2410 may be processed in real-time by system controller 2404 during a biometric session. Additionally or alternatively, data may be temporarily stored in system memory during a biometric session and processed at a slower rate than real-time or offline operation.

As shown in FIG. 25 , system controller 2404 may include a number of modules 2531 - 2539 in communication with master control module 2530 . The main control module 2530 can communicate with the user interface 2414 (FIG. 24). Although modules 2531 - 2539 are shown in direct communication with main control module 2530 , modules 2531 - 2539 may communicate directly with each other, with user interface 2414 and biosensor 2402 . In addition, the modules 2531 to 2539 may communicate with the main control module 2530 through other modules.

The plurality of modules 2531-2539 includes system modules 2531-2533, 2539 in communication with subsystems 2406, 2408, 2410, and 2409, respectively. Fluid control module 2531 may communicate with fluid control system 2406 to control valves and flow sensors of the fluid network to control the flow of one or more fluids through the fluid network. The fluid storage module 2532 can notify the user when the fluid level is low or when the waste reservoir is at or near capacity. The fluid storage module 2532 can also communicate with the temperature control module 2533 so that the fluid can be stored at a desired temperature. The illumination module 2539 can communicate with the illumination system 2409 to illuminate the reaction site at specified times during the protocol, such as after a desired reaction (eg, a binding event) has occurred. In some implementations, the illumination module 2539 can communicate with the illumination system 2409 to illuminate the reaction site at a specified angle.

Plurality of modules 2531 to 2539 may also include a device module 2534 that communicates with biosensor 2402 and an identification module 2535 that determines identification information associated with biosensor 2402 . Device module 2534 may communicate with system socket 2412 to confirm that the biosensor has established electrical and fluid connection with base calling system 2400, for example. Identification module 2535 may receive a signal identifying biosensor 2402 . Identification module 2535 may use the identity of biosensor 2402 to provide other information to the user. For example, identification module 2535 may determine and then display a lot number, date of manufacture, or a protocol recommended to run with biosensor 2402 .

The number of modules 2531 to 2539 also includes an analysis module 2538 (also referred to as a signal processing module or signal processor) that receives and analyzes signal data (eg, image data) from the biosensor 2402 . Analysis module 2538 includes memory (eg, RAM or flash memory) for storing detection data. The detection data may include multiple pixel signal sequences such that a pixel signal sequence from each of the millions of sensors (or pixels) may be detected over many base calling cycles. Signal data may be stored for subsequent analysis, or may be transmitted to user interface 2414 to display desired information to the user. In some implementations, the signal data can be processed by a solid-state imaging device (eg, a CMOS image sensor) before the signal data is received by the analysis module 2538 .

The analysis module 2538 is configured to obtain image data from the light detectors at each of the plurality of sequencing cycles. The image data is derived from the emission signal detected by the photodetector and passed through a neural network (e.g., neural network-based template generator 2548, neural network-based base caller 2558 (see, e.g., FIGS. 7, 9 and 10 ) and/or neural network-based quality scorer 2568) process the image data for each of the plurality of sequencing cycles and generate bases for at least some of the analytes at each of the plurality of sequencing cycles base checkout.

Protocol module 2536 and protocol module 2537 communicate with main control module 2530 to control the operation of subsystems 2406, 2408, and 2410 when performing predetermined assay protocols. The protocol modules 2536 and 2537 may include instruction sets for instructing the base calling system 2400 to perform specific operations according to a predetermined protocol. As shown, the protocol module may be a sequencing by synthesis (SBS) module 2536 configured to issue various commands for performing the sequencing by synthesis process. In SBS, the extension of a nucleic acid primer along a nucleic acid template is monitored to determine the sequence of nucleotides in the template. The underlying chemical process can be polymerization (eg, catalyzed by a polymerase) or ligation (eg, catalyzed by a ligase). In certain polymerase-based SBS implementations, fluorescently labeled nucleotides are added to primers (thus allowing primer extension) in a template-dependent manner, allowing detection of the order and type of nucleotides added to the primers. Can be used to determine the sequence of the template. For example, to initiate a first SBS cycle, a command can be issued to deliver one or more labeled nucleotides, DNA polymerase, etc. to/through a flow cell containing an array of nucleic acid templates. Nucleic acid templates can be located at corresponding reaction sites. Those reactive sites where primer extension results in incorporation of labeled nucleotides can be detected by imaging events. During an imaging event, illumination system 2409 can provide excitation light to the reaction sites. Optionally, the nucleotide may also include a reversible termination property that terminates further primer extension once the nucleotide has been added to the primer. For example, a nucleotide analog having a reversible terminator moiety can be added to the primer such that subsequent extension does not occur until delivery of an unblocking agent to remove the moiety. Thus, for implementations using reversible termination, a command can be issued to deliver the deblocking agent to the flow cell (either before or after detection occurs). One or more commands may be issued to effectuate washing between various delivery steps. This cycle can then be repeated n times to extend the primer by n nucleotides to detect a sequence of length n. Exemplary sequencing techniques are described in, for example, Bentley et al., Nature 456:53-59 (2008), WO 04/018497, US 7,057,026, WO 91/06678, WO 07/123744, US 7,329,492, US 7,211,414, US 7,315,019 and US 7,405,281 , each of these documents is incorporated herein by reference.

For the nucleotide delivery step of the SBS cycle, a single type of nucleotide can be delivered at a time, or multiple different nucleotide types (eg, A, C, T, and G together) can be delivered. For nucleotide delivery configurations where only a single type of nucleotide is present at a time, different nucleotides need not have different labels, as they can be distinguished based on the time interval inherent in individualized delivery. Accordingly, a sequencing method or device may use single-color detection. For example, an excitation source need only provide excitation at a single wavelength or range of wavelengths. For nucleotide delivery configurations where delivery results in the simultaneous presence of multiple different nucleotide types in the flow cell, the sites where different nucleotide types are incorporated can be distinguished based on the different fluorescent labels attached to the corresponding nucleotide types in the mixture. point. For example, four different nucleotides, each with one of four different fluorophores, can be used. In one implementation, four different fluorophores can be distinguished using excitation in four different regions of the spectrum. For example, four different excitation radiation sources can be used. Alternatively, fewer than four different excitation sources may be used, but optical filtering of the excitation radiation from a single source may be used to generate different ranges of excitation radiation at the flow cell.

In some implementations, fewer than four different colors can be detected in a mixture with four different nucleotides. For example, pairs of nucleotides can be detected at the same wavelength, but based on differences in the intensity of one member of the pair relative to the other, or based on one member of the pair resulting in a signal that differs from the detected signal of the other member of the pair. Distinguished by changes (eg, by chemical modification, photochemical modification, or physical modification) compared to the appearance or disappearance of a distinct signal. Exemplary devices and methods for differentiating four different nucleotides using detection using fewer than four colors are described, for example, in US Patent Application Serial Nos. 61/538,294 and 61/619,878, which are incorporated herein by reference in their entirety. US Application 13/624,200, filed September 21, 2012, is also incorporated by reference in its entirety.

Number of protocol modules may also include a sample preparation (or generation) module 2537 configured to issue commands to fluid control system 2406 and temperature control system 2410 to amplify products within biosensor 2402 . For example, biosensor 2402 can be coupled to base calling system 2400 . Amplification module 2537 may issue instructions to fluid control system 2406 to deliver the necessary amplification components to reaction chambers within biosensor 2402 . In other implementations, the reaction site may already contain some components for amplification, such as template DNA and/or primers. After delivering the amplification components to the reaction chamber, the amplification module 2537 can instruct the temperature control system 2410 to cycle through different temperature stages according to known amplification protocols. In some implementations, amplification and/or nucleotide incorporation is performed isothermally.

The SBS module 2536 can issue commands to perform bridge PCR, in which clusters of clonal amplicons form on localized regions within the channels of the flow cell. After amplicons are generated by bridge PCR, the amplicons can be "linearized" to make single-stranded template DNA or sstDNA, and sequencing primers can be hybridized to universal sequences flanking the region of interest. For example, a reversible terminator-based sequencing-by-synthesis approach can be used as described above or as follows.

Each base calling or sequencing cycle can extend sstDNA by a single base, which can be done, for example, by using a modified DNA polymerase and a mixture of the four types of nucleotides. Different types of nucleotides can have unique fluorescent labels, and each nucleotide can also have a reversible terminator that allows only a single base incorporation to occur each cycle. Following the addition of a single base to the sstDNA, excitation light can be incident on the reaction site and fluorescence emission can be detected. After detection, the fluorescent label and terminator can be chemically cleaved from the sstDNA. This may be followed by another similar base calling or sequencing cycle. In such sequencing protocols, the SBS module 2536 can instruct the fluid control system 2406 to direct the flow of reagent and enzyme solutions through the biosensor 2402 . Exemplary reversible terminator-based SBS methods that can be used with the devices and methods described herein are described in U.S. Patent Application Publication 2007/0166705 Al, U.S. Patent Application Publication 2006/0188901 Al, U.S. Patent 7,057,026, U.S. Patent Application Publication 2006 /0240439 Al, US Patent Application Publication 2006/02814714709 Al, PCT Publication WO 05/065814, PCT Publication WO 06/064199, each of which is incorporated herein by reference in its entirety. Exemplary reagents for reversible terminator-based SBS are described in US 7,541,444, US 7,057,026, US 7,427,673, US 7,566,537, and US 7,592,435, each of which is incorporated herein by reference in its entirety.

In some implementations, the amplification module and the SBS module can be operated in a single assay protocol, where, for example, a template nucleic acid is amplified and then sequenced within the same cassette.

The base calling system 2400 may also allow the user to reconfigure the assay protocol. For example, base calling system 2400 can provide a user through user interface 2414 with options for modifying the determined protocol. For example, base calling system 2400 may request a temperature for an annealing cycle if it is determined that biosensor 2402 will be used for amplification. Additionally, the base calling system 2400 can issue a warning to the user if the user has provided user input that is generally not acceptable for the selected assay protocol.

In a specific implementation, the biosensor 2402 includes millions of sensors (or pixels), and each sensor (or pixel) generates a plurality of pixel signal sequences in subsequent base calling cycles. The analysis module 2538 detects and assigns a plurality of pixel signal sequences to corresponding sensors (or pixels) based on the row-by-row and/or column-by-column positions of the sensors on the sensor array.

Each sensor in the sensor array can generate sensor data for a block of the flow cell in a region on the flow cell where the cluster of genetic material was disposed during a base calling operation. The sensor data may include image data in the pixel array. For a given cycle, the sensor data may include more than one image, resulting in multiple features per pixel as block data.

FIG. 26 is a simplified block diagram of a computer system 2600 that can be used to implement the disclosed techniques. Computer system 2600 includes at least one central processing unit (CPU) 2672 in communication with a number of peripheral devices via bus subsystem 2655 . These peripherals may include storage subsystem 2610 including, for example, memory device and file storage subsystem 2636 , user interface input device 2638 , user interface output device 2676 , and network interface subsystem 2674 . Input and output devices allow a user to interact with computer system 2600 . Network interface subsystem 2674 provides interfaces to external networks, including interfaces to corresponding interface devices in other computer systems.

User interface input devices 2638 may include: a keyboard; pointing devices such as a mouse, trackball, touch pad, or tablet; a scanner; a touch screen incorporated into a display; audio input devices such as a voice recognition system and a microphone; input device. In general, use of the term "input device" is intended to include all possible types of devices and manners of entering information into computer system 2600 .

User interface output devices 2676 may include a display subsystem, a printer, a facsimile machine, or a non-visual display such as an audio output device. The display subsystem may include an LED display, a cathode ray tube (CRT), a flat panel device such as a liquid crystal display (LCD), a projection device, or some other mechanism for producing a visible image. The display subsystem may also provide non-visual displays, such as audio output devices. In general, use of the term "output device" is intended to include all possible types of devices and means of outputting information from computer system 2600 to a user or to another machine or computer system.

Storage subsystem 2610 stores programming structures and data structures that provide the functionality of some or all of the modules and methods described herein. These software modules are typically executed by the deep learning processor 2678.

In one specific implementation, the neural network is implemented using deep learning processors 2678, which may be configurable and reconfigurable processors, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), and/or Or coarse-grained reconfigurable architecture (CGRA) and graphics processing unit (GPU) or other configured devices. The deep learning processor 2678 can be hosted by a deep learning cloud platform such as Google Cloud Platform ^™ , Xilinx ^™ , and Cirrascale ^™ . Examples of Deep Learning Processor 14978 include Google's Tensor Processing Unit (TPU) ^TM , Rackmount Solutions (such as GX4 Rackmount Series ^TM , GX149 Rackmount Series ^TM ), NVIDIA DGX-1 ^TM , Microsoft's Stratix V FPGA ^TM , Graphcore's Intelligent Processor Unit (IPU) ^TM , Qualcomm's Zeroth Platform ^TM with Snapdragon processors ^TM , NVIDIA's Volta ^TM , NVIDIA's DRIVE PX ^TM , NVIDIA's JETSON TX1/TX2 MODULE ^TM , Intel's Nirvana ^TM , Movidius VPU ^TM , Fujitsu DPI ^TM , DynamicIQ ^TM of ARM, IBM TrueNorth ^TM , etc.

Memory subsystem 2622 used in storage subsystem 2610 may include a number of memories including main random access memory (RAM) 2634 for storing instructions and data during program execution and read only memory (ROM) in which fixed instructions are stored. )2632. File storage subsystem 2636 may provide persistent storage for program files and data files and may include hard drives, floppy disk drives and associated removable media, CD-ROM drives, optical drives, or removable media cartridges. Modules that implement certain implemented functions may be stored in the storage subsystem 2610 by the file storage subsystem 2636, or in other machines accessible by the processor.

Bus subsystem 2655 provides a mechanism for the various components and subsystems of computer system 2600 to communicate with each other as intended. Although the bus subsystem 2655 is shown schematically as a single bus, alternative implementations of the bus subsystem may use multiple buses.

Computer system 2600 itself may be of various types, including a personal computer, portable computer, workstation, computer terminal, network computer, television, mainframe, server farm, a widely distributed group of loosely networked computers, or any other data processing system or user equipment. Due to the ever-changing nature of computers and networks, the description of computer system 2600 depicted in FIG. 26 is intended only as a specific example for illustrating a preferred implementation of the invention. Many other configurations of computer system 2600 are possible, having more or fewer components than the computer system depicted in FIG. 26 .

Claims (29)

1. A system comprising: host processor; memory, the memory being accessible by the host processor, the memory storing: The topology of the neural network, A plurality of weight sets for configuring the topology to perform base calling operations, the weight sets in the plurality of weight sets are trained on corresponding training data sets in the plurality of training data sets, the training data sets correspond to A corresponding sequencing event in a plurality of sequencing events of the base calling operation spanning the time progression of the base calling operation through a subseries of sensing cycles in a series of sensing cycles and the the spatial progression of base calling operations through positions on the biosensor, and sensor data for a sensing cycle in the series of sensing cycles; and a configurable processor capable of accessing the memory and configured with dataflow logic to: loading said topology on a processing element of said configurable processor, selecting a weight set from the plurality of weight sets based at least in part on a subject subseries of sensing cycles and/or a subject position on the biosensor, loading on the processing element subject sensor data for the subject sub-series of the sensing cycle and the subject's location on the biosensor, and loading weights in the selected set of weights on the processing element to configure the topology with the weights, and causing the neural network to apply the weights in the selected set of weights to the subject sensor data to Generate base call classification data. 2. The system of claim 1 , wherein the sensing cycle sub-series comprises an initial sensing cycle sub-series, an intermediate sensing cycle sub-series and a final sensing cycle sub-series, and wherein the training data set and the The weight sets respectively correspond to the initial sensing cycle sub-series, the intermediate sensing cycle sub-series and the final sensing cycle sub-series. 3. The system according to claim 1 or 2, wherein the locations on the biosensor include edge locations and non-edge locations, and wherein the training data set and the weight set respectively correspond to the edge locations and the non-edge locations. 4. The system of any one of claims 1 to 3, wherein the locations on the biosensor include a first quadrant location, a second quadrant location, a third quadrant location, and a fourth quadrant location, and wherein The training data set and the weight set respectively correspond to the first quadrant position, the second quadrant position, the third quadrant position and the fourth quadrant position. 5. The system according to any one of claims 1 to 4, wherein the biosensor is partitioned into a plurality of blocks, and the edge location, the non-edge location, the first quadrant location, the Each of the second quadrant location, the third quadrant location, and the fourth quadrant location includes a corresponding one or more of the plurality of blocks. 6. The system of any one of claims 1 to 5, wherein the sequencing events span the time progression of the base calling operation through base called paired-end reads, and wherein the training dataset and the weight sets respectively correspond to reads in the paired-end reads. 7. The system according to any one of claims 1 to 6, wherein: The sensing cycle sub-series includes an initial sensing cycle sub-series, an intermediate sensing cycle sub-series and a final sensing cycle sub-series; said locations on said biosensor include edge locations and non-edge locations; and The training data set and thus the weight set correspond to (i) the initial sensing cycle sub-series and the edge positions, (ii) the intermediate sensing cycle sub-series and the edge positions, respectively, (iii) ) said final sensing cycle sub-series and said edge positions, (iv) said initial sensing cycle sub-series and said non-edge positions, (v) said intermediate sensing cycle sub-series and said non-edge positions , and (vi) said final sensing cycle subseries and said non-edge locations. 8. The system according to any one of claims 1 to 7, wherein: The sensing cycle sub-series includes an initial sensing cycle sub-series, an intermediate sensing cycle sub-series and a final sensing cycle sub-series; said locations on said biosensor include locations of a first category and locations of a second category; and The training data set and thus the weight set correspond to (i) the initial sensing cycle sub-series and the position of the first category, (ii) the intermediate sensing cycle sub-series and the first class respectively. The location of the category, (iii) the location of the final sensing cycle subseries and the first category, (iv) the initial sensing cycle subseries and the non-edge location of the second category, (v) the (vi) the position of the intermediate subseries of sensing cycles and the second category, and (vi) the position of the final subseries of sensing cycles and the second category. 9. The system of any one of claims 1 to 8, wherein the configurable processor is further: determining one or more parameters of the current sequencing run; and The set of weights is selected from the plurality of sets of weights further based on the determined one or more parameters of the current sequencing run. 10. The system of claim 9, wherein the determined one or more parameters of the current sequencing run include one or more of: characteristics of reaction components used in the biosensor Or phased features associated with said sensor data. 11. A system comprising: host processor; memory, the memory being accessible by the host processor, the memory storing: The topology of the neural network, a first set of weights, a second set of weights and a third set of weights for configuring the topology to perform a base calling operation, the first set of weights, the second set of weights and the third set of weights respectively corresponding to a first sensing cycle sub-series, a second sensing cycle sub-series and a third sensing cycle sub-series in a series of sensing cycles, and first sensor data, second sensor data, and third sensor data corresponding to the first, second, and third sensing cycle sub-series, respectively; and a configurable processor capable of accessing the memory and configured with dataflow logic to: loading said topology on a processing element of said configurable processor, loading the first sensor data on the processing element, loading the first set of weights on the processing element to configure the topology with weights in the first set of weights, and causing the neural network to applying the weights in the first set of weights to the first sensor data to generate first base call classification data for a sensing cycle in the first sub-series of sensing cycles, loading the second sensor data on the processing element, loading the second set of weights on the processing element to configure the topology with weights in the second set of weights, and causing the neural network to applying the weights in the second set of weights to the second sensor data to generate second base call classification data for sensing cycles in the second sub-series of sensing cycles, and loading the third sensor data on the processing element, loading the third set of weights on the processing element to configure the topology with weights in the third set of weights, and causing the neural network to The third sensor data applies the weights of the third set of weights to generate third base call classification data for a sensing cycle in the third sub-series of sensing cycles. 12. The system of claim 11, wherein the memory further stores: A fourth weight set, a fifth weight set, and a subsequent weight set for configuring the topology to perform a base call operation, the fourth weight set, the fifth weight set, and the subsequent weight set respectively correspond to a fourth sub-series of sensing cycles, a fifth sub-series of sensing cycles, and a subsequent sub-series of sensing cycles in the series of sensing cycles; and Fourth sensor data, fifth sensor data and subsequent sensor data for said fourth sub-series of sensing cycles, said fifth sub-series of sensing cycles and said subsequent sub-series of sensing cycles. 13. The system of claim 11 or 12, wherein the configurable processor is configured with data flow logic to: loading the fourth sensor data on the processing element, loading the fourth set of weights on the processing element to configure the topology with weights in the fourth set of weights, and causing the neural network to applying the weights in the fourth set of weights to the fourth sensor data to generate fourth base call classification data for sensing cycles in the fourth sub-series of sensing cycles; loading the fifth sensor data on the processing element, loading the fifth set of weights on the processing element to configure the topology with weights in the fifth set of weights, and causing the neural network to applying the weights in the fifth set of weights to the fifth sensor data to generate fifth base call classification data for a sensing cycle in the fifth subseries of sensing cycles; and loading the subsequent sensor data and the subsequent set of weights on the processing element to configure the topology with weights in the subsequent set of weights, and causing the neural network to apply the subsequent weights to the subsequent sensor data The weights are aggregated to generate subsequent base calling classification data for sensing cycles in the subseries of subsequent sensing cycles. 14. The system according to any one of claims 11 to 13, wherein the topology employs sensor data from successive sensing cycles as input, and the topology comprises a spatial layer and a temporal layer, the spatial layer Instead of combining the sensor data and the resulting feature maps between the successive sensing cycles, the temporal layers combine the resulting feature maps between the successive sensing cycles. 15. The system according to any one of claims 11 to 14, wherein the first set of weights comprises first spatial weights for the spatial layer and first temporal weights for the temporal layer, the first The second set of weights includes a second spatial weight for the spatial layer and a second temporal weight for the temporal layer, and the third set of weights includes a third spatial weight for the spatial layer and a second temporal weight for the temporal layer The third time weight of . 16. The system according to any one of claims 11 to 15, wherein said first set of weights comprises spatial weights for said spatial layer and first temporal weights for said temporal layer, said second weight set includes a second temporal weight for the temporal stratum, and the third set of weights includes a third temporal weight for the temporal stratum, and wherein the configurable processor is configured with dataflow logic to: loading the first sensor data on the processing element, loading the spatial weights and the first temporal weights on the processing element to configure the spatial layer with the spatial weights and with the first temporal weights configuring the temporal layer, and causing the neural network to apply the configured spatial layer and the configured temporal layer to the first sensor data to generate a metric for a sensing cycle in the first sensing cycle sub-series First base call classification data; loading the second sensor data on the processing element, loading the second temporal weights on the processing element to reconfigure the temporal layer with weights in the second temporal weights without reconfiguring the a spatial layer, and causing the neural network to apply a reconfigured temporal layer and a previously configured spatial layer to the second sensor data to generate a second base for a sensing cycle in the second subseries of sensing cycles base call classification data; and The third sensor data is loaded on the processing element, the third temporal weights are loaded on the processing element to reconfigure the temporal layer with weights in the third temporal weights without reconfiguring the a spatial layer, and causing the neural network to apply a reconfigured temporal layer and a previously configured spatial layer to the third sensor data to produce a third base for a sensing cycle in the third sensing cycle subseries Base checkout categorical data. 17. The system of any one of claims 11 to 16, wherein weights in the first set of weights, the second set of weights and the third set of weights are quantized using different scaling factors. 18. The system of any one of claims 11 to 17, wherein weights in the first set of weights, the second set of weights, and the third set of weights correspond to a first sequencing chemistry, a second sequencing chemistry, respectively. Sequencing chemistry and third sequencing chemistry. 19. The system of any one of claims 11 to 18, wherein weights in the first set of weights, the second set of weights, and the third set of weights correspond to the first sequencing assay, the second set of weights, respectively. Sequencing Assay and Third Sequencing Assay. 20. The system of any one of claims 11 to 19, wherein weights in the first set of weights, the second set of weights, and the third set of weights correspond to the first sequencing configuration, the second A sequencing configuration and a third sequencing configuration. 21. A computer-implemented method for generating base calling classification data, the method comprising: loading the topology of the neural network on a processing element of a processor for performing a base calling operation; Storing (i) first sensor data from clusters within a first one or more blocks of a flow cell, (ii) second sensor data from clusters within a second one or more blocks of said flow cell. sensor data, (iii) a first set of weights comprising a first one or more weights, and (iv) a second set of weights comprising a second one or more weights, wherein said first sensor data and said second sensor data is generated during a subset of sensing cycles in the series of sensing cycles; configuring the topology of the neural network with the first set of weights, and processing the first sensor data with the neural network configured using the first set of weights and generating a a plurality of blocks and first base call classification data for the subset of sensing cycles; and configuring the topology of the neural network with the second set of weights, and processing the second sensor data with the neural network configured using the second set of weights and generating a A plurality of blocks and second base calling classification data for the subset of sensing cycles. 22. The method of claim 21 , wherein the sensing cycle subset is a first sensing cycle subset, and wherein the method further comprises: storing (i) third sensor data from clusters within said first one or more blocks, (ii) fourth sensor data from clusters within said second one or more blocks, ( iii) a third set of weights, and (iv) a fourth set of weights, wherein said third sensor data and said fourth sensor data are generated during a second subset of sensing cycles in said series of sensing cycles, the second subset of sensing cycles follows the first subset of sensing cycles in the series of sensing cycles; configuring the topology of the neural network with the third set of weights, and processing the third sensor data with the neural network configured using the third set of weights and generating a a plurality of blocks and third base call classification data for the second subset of sensing cycles; and configuring the topology of the neural network with the fourth set of weights, and processing the fourth sensor data with the neural network configured using the fourth set of weights and generating a A plurality of blocks and fourth base call classification data for the second subset of sensing cycles. 23. The method of claim 21 or 22, wherein: said first one or more zones are within a first region of said flow cell; and The second one or more zones are within a second region of the flow cell. 24. A method according to any one of claims 21 to 23, wherein: said first one or more sections are edge sections of said flow cell; and The second one or more segments are non-edge segments of the flow cell. 25. The method of any one of claims 21 to 24, further comprising: generating the first set of weights by training the neural network on sensor data generated only from edge blocks; and The second set of weights is generated by training the neural network on sensor data generated only from non-edge blocks. 26. A system comprising: host processor; a memory accessible by the host processor, the memory storing (i) a topology of the neural network, and (ii) a plurality of weights for configuring the topology to perform a base calling operation, wherein the plurality of weights is based on tile location, a series of sensing cycles, and/or sensor data; and a configurable processor capable of accessing the memory and configured with dataflow logic to: loading said topology on a processing element of said configurable processor, The plurality of weights is loaded on the processing element to configure the topology with the plurality of weights to cause the neural network to generate base call classification data. 27. The system of claim 26, wherein the plurality of weights is a first plurality of weights, the tile location is a first tile location, and the series of sensing cycles is a first series of sensing cycles loop, and the sensor data is the first sensor data, and where: The memory is configured to further store a second plurality of weights for configuring the topology to perform a base calling operation, wherein the second plurality of weights is based on a second tile position, a second series of sensing cycles and/or second sensor data; and The configurable processor is configured with dataflow logic to: The second plurality of weights is loaded on the processing element to configure the topology with the second plurality of weights to cause the neural network to generate additional base calling classification data. 28. The system of claim 27, wherein: the first block location is on a first region within the flow cell; and The second block location is on a second region within the flow cell. 29. The system of claim 27 or 28, wherein: The second series of sensing cycles occurs after the first series of sensing cycles.

Images