CN107818329A - A kind of MASS SPECTRAL DATA ANALYSIS method - Google Patents

Abstract

The invention provides a mass spectrum data analysis method, which comprises a sample data collection step, a sample data preprocessing step, a data model construction and cross-validation step, a data model optimization step, and a sample group judgment step.

Description

A method for mass spectrometry data analysis

technical field

The invention relates to the field of machine learning applications, in particular to a mass spectrometry data analysis method.

Background technique

Machine learning (Machine Learning, ML) is a multi-field interdisciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines, specializing in the study of how computers simulate or realize the learning behavior of human groups , to acquire new knowledge or skills, and reorganize the existing knowledge structure to continuously improve its own performance. Machine learning is the core of artificial intelligence and can be applied to many fields such as data mining, computer vision, natural language processing, biometric identification, search engines, medical diagnosis, detection of credit card fraud, securities market analysis, and DNA sequence sequencing. Machine learning algorithms are a set of algorithms that automatically analyze and obtain laws from known data and use the laws to predict unknown data.

Mass spectrometry data is to use a special instrument to ionize the sample to generate charged ions with different charge-to-mass ratios, and then use an external electric field to separate the ions with different charge-to-mass ratios in space or time to obtain mass spectrometry data. Ions with different mass-to-charge ratios are separated by a mass analyzer, detected and recorded, and processed by a computer to generate a mass spectrum.

In the fields of biology, chemistry, and medicine, it often involves the classification of body fluid samples according to their components. Generally speaking, most technicians use separate analysis and comparison methods. The advantage of this method is that the sample components are clear and easy to classify. Accurate; its disadvantage is that when there are many types of body fluid samples to be classified, it takes a lot of time and resources, and the labor cost is high. How to infer the type of a new body fluid sample based on known types of body fluid samples has always been an important research topic for researchers.

Taking the medical field as an example, some of the same special components often exist in the body fluids of patients with certain diseases known so far. These components may be the cause of the patient suffering from the same disease, or may be the manifestation of a certain type of disease. Clinically, if a certain type of component is detected in a patient's body fluid, the patient can be associated with a certain disease or a certain type of disease, providing data support for clinical diagnosis. Since the human body is a very complex organism, the diagnosis of diseases and the selection of treatment options require professional medical personnel to make judgments based on the massive data of each individual, resulting in low diagnostic efficiency and high labor costs. When there are a large number of patients who need to be examined, patients need to queue for a long time, and doctors will work hard continuously. The time for diagnosis and treatment of each patient is short, and misdiagnosis is prone to occur. Therefore, in clinical medicine, there is a need for a medical device that can analyze the composition of a large number of body fluid samples at the same time, and can detect and analyze in a short time whether or not Contains certain specific ingredients, so as to assist medical personnel to make a diagnosis more conveniently and accurately.

Contents of the invention

The purpose of the present invention is to provide a mass spectrometry data analysis method to solve the technical problems in the prior art that when there are a large number of body fluid samples to be classified, a large amount of time and resources are consumed and labor costs are high.

In order to solve the above technical problems, the present invention provides a mass spectrometry data analysis method, comprising the following steps: a sample data collection step for collecting mass spectrometry data of two or more body fluid samples and generating a mass spectrogram according to the mass spectrometry data; the body fluid sample Including two or more training samples and at least one test sample; the training samples are divided into two or more groups, and the training samples of the same group are marked with the same group label; the sample data preprocessing step is used to at least one group Preprocessing the mass spectrum data, performing coordinate transformation processing on the mass spectrum to obtain standardized mass spectrum data of the training sample and the test sample; data model construction and cross-validation steps for utilizing the standardized mass spectrum data of the training sample and the group labels of the training samples to construct a primary data model, and carry out at least one cross-validation process on the primary data model according to the standardized mass spectrum data of the training samples; the data model optimization step is used to construct according to the results of the cross-validation Optimizing the data model; and a sample group judgment step for obtaining the group label of the test sample by using the standardized mass spectrum data of the test sample and the optimized data model.

Further, the sample data collection step specifically includes the following steps: acquiring more than two body fluid samples; arranging all the body fluid samples in a matrix on a flat plate; and using mass spectrometry to collect mass spectrum data of the body fluid samples and generating Mass spectrum; at least one set of mass spectrum data is collected for each body fluid sample.

Further, the test sample is located in the middle of the flat plate, and the training sample surrounds the test sample; the flat plate includes but is not limited to a matrix metal plate; the group labels of any two adjacent training samples are different; The distance between any two adjacent body fluid samples is greater than or equal to 2mm and less than or equal to 5mm.

Further, each set of mass spectrometry data includes the mass-to-charge ratio of an ion in the body fluid sample and the measured signal intensity value corresponding to the ion; each set of mass spectrometry data corresponds to a sampling point in the mass spectrogram; the abscissa of each sampling point Indicates the mass-to-charge ratio of an ion, and its ordinate indicates the measured intensity value of the signal corresponding to the ion.

Further, the sample data preprocessing step specifically includes the following steps: a baseline correction step, for performing baseline correction processing on the mass spectrum data in the mass spectrogram; a resampling step, for using a resampling algorithm to correct the baseline The mass-to-charge ratio of the ions in the mass spectrum data is resampled, the abscissa of the mass spectrum is transformed, and the mass-to-charge ratios of all mass spectrum data are unified to obtain resampled mass spectrum data; the standardization step is used to resample the mass spectrum data The signal intensity value of the neutral ion is standardized, and the ordinate transformation is performed on the mass spectrum to obtain standardized mass spectrum data.

Further, the baseline correction step specifically includes the following steps: a signal calculation step for calculating the baseline signal intensity corresponding to at least one mass-to-charge ratio value in a set of mass spectrum data by using a window function; a signal correction step for calculating the baseline signal intensity according to the baseline signal Intensity correction corresponds to the measured signal intensity of the mass-to-charge ratio; the signal calculation step and the signal correction step are repeated to sequentially complete the correction of each group of mass spectrum data of each body fluid sample.

Further, the resampling step specifically includes the following steps: an effective mass-to-charge ratio selection step, used to select the effective mass-to-charge ratio interval and the number of effective mass-to-charge ratios; an effective mass-to-charge ratio calculation step, to use the resampling algorithm to calculate the The mass-to-charge ratio of the sampled mass spectrum data; the interpolation processing step is used to interpolate the mass spectrum after baseline correction by using the mass-to-charge ratio after resampling and the mass-to-charge ratio number, and the abscissa of the mass spectrum after baseline correction is changed from charge to mass Ratio values are transformed into mass-to-charge ratio numbers.

Further, the resampling algorithm refers to: set the mass-to-charge ratio interval of the effective mass spectrum data after resampling to [y ₁ , y ₂ ], and the number of mass-to-charge ratio coordinates after resampling is N; use the following formula to calculate the resampling Mass-to-charge ratio coordinates after

Wherein, N is greater than 10 ⁴ and less than 10 ⁵ .

Further, the standardization step specifically includes the following steps: a step of calculating the absolute sum of signal intensities to calculate the absolute sum S of ion signal intensities in all resampled mass spectrum data; a step of setting the sum of standardized signal intensities to Set the absolute value sum of ion signal intensity values in all resampled mass spectrometry data after normalization to be a constant T; the signal intensity value change multiple calculation step is used to calculate the change multiple T/S of each signal intensity value; the signal intensity value change step, for performing synchronous amplification or synchronous reduction processing on each ion signal intensity value in the resampled mass spectrum data.

Further, the data model construction and cross-validation steps specifically include the following steps: choose a training sample as a standard training sample, and its group label is known; take the position of the standard training sample as the center of the circle, and use a specific length r For the radius, a circular area is set on the flat plate; a matrix D is constructed according to the standardized mass spectrum data of other training samples in the circular area except the standard training sample, and each column of data in the matrix D is respectively A group of standardized mass spectrum data corresponding to a training sample; according to the group labels of other training samples in the circular area except the standard training sample, the vector is obtained The group label of each training sample is recorded in the vector Medium; use sparse learning optimization algorithm to build primary data model Multiplying more than two groups of mass spectrum data of the standard training sample by the data model, arranging the products into a sequence according to the numerical value, rounding the median value, and obtaining the estimated group label of the standard training sample ; compare the inferred group label of the standard training sample with its group label, if the two are the same, then determine that the group label of the standard training sample is guessed correctly, and add one to the correctness counter; each training sample is used as a standard in turn For the training sample, repeat the above steps, carry out cross-validation processing on all the training samples, and calculate the group label judgment accuracy rate of the training sample under the condition that the radius is r, and the group label judgment accuracy rate is correctness The ratio of the numerical value of the counter to the total number of training samples; adjust the size of the radius r, repeat the above steps, and calculate the group label judgment accuracy rate when the radius r is a different value; judge from more than two group labels Select a maximum accuracy rate in the accuracy rate, and obtain the optimal value R of the radius r corresponding to the maximum accuracy rate. Further, described data model optimization step, specifically comprises the steps: take the position of a test sample as the center of circle, take the length of radius optimal value R as radius, set a circular area on the flat plate; Construct a matrix D _W according to the standardized mass spectrum data of all training samples in the circular area, and each column of data in the matrix D _W corresponds to a group of standardized mass spectral data of a training sample; according to all training samples in the circular area The group labels get the vector The group label of each training sample is recorded in the vector corresponding to the training sample in the form of natural numbers Medium; use sparse learning optimization algorithm to build an optimized data model The data model optimization step specifically includes the following steps: taking the position of a test sample as the center and the length of the optimal value R of the radius as the radius, setting a circular area on the flat plate; according to the circular area The standardized mass spectrometry data of all training samples in the matrix D W constructs a matrix D _W , and each column of data in the matrix D _W corresponds to a group of standardized mass spectrometry data of a training sample; according to the group labels of all training samples in the circular area, the vector is obtained The group label of each training sample is recorded in the vector corresponding to the training sample in the form of natural numbers Medium; use sparse learning optimization algorithm to build an optimized data model

Further, the sample group judgment step specifically includes the following steps: multiplying a set of mass spectrum data of a test sample by the data model, rounding the product, and obtaining the group label of the test sample; or Multiplying more than two groups of mass spectrometry data of a test sample with the data model, arranging the products into a series according to the numerical value, rounding the median value to obtain the group label of the test sample.

The advantage of the present invention is that it provides a mass spectrometry data analysis method, which can construct a grouper model according to the groups of known body fluid samples, and obtain the data model with the highest correct rate through multiple cross-validation of multiple training samples, which can be processed simultaneously Mass spectrometry data of a large number of body fluid samples and grouping them according to body fluid sample components.

Description of drawings

Fig. 1 is the flow chart of the mass spectrometry data analysis method described in the embodiment of the present invention;

Fig. 2 is a method flowchart of the sample data collection step described in the embodiment of the present invention;

Fig. 3 is the mass spectrogram generated before the preprocessing of the mass spectral data described in the embodiment of the present invention;

Fig. 4 is a method flowchart of the sample data preprocessing step described in the embodiment of the present invention;

Fig. 5 is a method flow chart of the sample data baseline correction step described in the embodiment of the present invention;

Fig. 6 is a mass spectrogram generated after baseline correction of the mass spectral data described in the embodiment of the present invention;

Fig. 7 is a method flow chart of the mass spectrum data resampling processing steps described in the embodiment of the present invention;

Fig. 8 is a schematic diagram of the effective mass-to-charge ratio in the resampled mass spectrum data according to the embodiment of the present invention;

FIG. 9 is a mass spectrum generated by resampling mass spectrum data according to an embodiment of the present invention;

Fig. 10 is a method flowchart of the standardization processing steps of mass spectrum data according to the embodiment of the present invention;

Figure 11 shows the mass spectrogram generated by the standardized mass spectrometry data of the present embodiment;

12 is a method flow chart of the data model construction and cross-validation steps described in the embodiment of the present invention;

Fig. 13 is a method flowchart of the data model optimization step according to the embodiment of the present invention.

Detailed ways

An embodiment of the present invention is provided below, with reference to the accompanying drawings, to demonstrate that the present invention can be implemented.

As shown in FIG. 1 , this embodiment provides a mass spectrometry data analysis method, including the following steps S1) to S5).

Step S1) The sample data collection step is used to collect at least one set of mass spectrum data of two or more body fluid samples and generate a mass spectrum according to the mass spectrum data. The body fluid samples include more than two training samples and at least one test sample; the training samples are divided into more than two groups (also referred to as categories), and the training samples of the same group are marked with the same group label. The body fluid sample can be a certain kind of body fluid from the human body or other organisms. In this embodiment, human blood samples are preferred, and the group labels are 0 and 1 respectively. The samples of group 0 come from patients with certain diseases (such as diabetic patients, Hemophiliacs, etc.), the samples of group 1 come from healthy people without the disease, the training samples are blood samples with known group labels, and each blood sample is marked with 0 or 1. In other embodiments, the group labels may also be identified as other natural numbers.

As shown in Fig. 2, step S1) specifically includes the following steps: Step S101) obtains more than two body fluid samples; generally dozens or hundreds of samples can be selected. Step S102) Arranging all the body fluid samples in the form of droplets on a flat plate (preferably a matrix metal plate) to form a matrix, the test sample is located in the middle of the flat plate, and the training samples surround the test sample; any two The group labels of two adjacent training samples are all different; the distance between any two adjacent body fluid samples is greater than or equal to 2mm and less than 5mm; the flat plate includes but not limited to a matrix metal plate. Step S103) Using mass spectrometry to collect mass spectrometry data of the body fluid sample and generate a mass spectrogram, as shown in Figure 3; collecting at least one set of mass spectrometry data for each body fluid sample, preferably more than three sets, to reduce the negative impact of mass spectrometry data errors , improve the accuracy rate, and realize pattern classification on the basis of multiple sets of data in the same sample, which can effectively reduce the interference caused by the error of a single set of data. Each group of mass spectrometry data includes the mass-to-charge ratio of an ion in the body fluid sample and the measured signal intensity value corresponding to the ion; for each sampling point in the mass spectrogram, its abscissa represents the mass-to-charge ratio of an ion, and its ordinate represents the mass-to-charge ratio of an ion. For the measured signal intensity values of the ions, see Figure 3 for details.

Step S2) The sample data preprocessing step is used to preprocess at least one set of mass spectrum data, perform coordinate transformation processing on the mass spectrum, and obtain standardized mass spectrum data of the training sample. Due to factors such as sample processing, instrument performance, external contamination, etc., the mass spectral data directly obtained by the mass spectrometer needs to be properly preprocessed to improve the grouping accuracy.

As shown in Figure 4, step S2) specifically includes steps S201) to step S203). For the mass spectrum data on the mass spectrum, through three processing steps of baseline correction, resampling and standardization, it is possible to avoid too many external factors affecting the Grouping precision for mass spectrometry data.

Step S201) The baseline correction step is used to perform baseline correction processing on the mass spectrum data on the mass spectrum. The baseline is the basic intensity value in the mass spectrum data. The function of the baseline correction step is to identify and remove a large deviation from the baseline in the mass spectrum , to remove data with large deviations in the mass spectrometry data. As shown in Figure 5, the step S201) the baseline correction step specifically includes the following steps: step S2011) the signal calculation step, in order to utilize the window function to calculate the baseline signal intensity of at least one mass-to-charge ratio in a set of mass spectrum data; step S2012) signal correction a step of correcting the measured signal strength corresponding to the mass-to-charge ratio according to the baseline signal strength, screening and removing invalid data with large deviations; repeating steps S2011) to S2012), and sequentially completing each body fluid sample Calibration of group mass spectrometry data. When the computer is used to realize engineering test signal processing, it is not possible to measure and calculate the infinitely long signal, but to take its limited time segment for analysis, intercept a time segment from the signal, and then use the intercepted signal time segment for period delay. By extension processing, a virtual infinitely long signal can be obtained, and mathematical processing such as Fourier transform and correlation analysis can be performed on the signal. In specific applications, different interception functions can be used to truncate the signal, and the interception function is called a window function. In this embodiment, the window function STEP is set to 50, and WINDOW is set to 50. After the baseline correction is completed, the mass spectrum after the baseline correction is obtained, as shown in Figure 6 for details, the abscissa represents the mass-to-charge ratio of an ion, and the ordinate represents the measured intensity value of the signal corresponding to the ion.

Step S202) a resampling step, for resampling the ion mass-to-charge ratios in the baseline-corrected mass spectrometry data using a resampling algorithm, performing abscissa transformation on the mass spectrograms, and unifying the mass-to-charge ratios of all mass spectrometry data, Remove mass spectral data with large deviations to obtain resampled mass spectral data.

As shown in FIG. 7 , the resampling step S202) specifically includes the following steps S2021) to S2023). S2021) The effective mass-to-charge ratio selection step is used to select the effective mass-to-charge ratio interval and the effective mass-to-charge ratio quantity; construct the effective mass-to-charge ratio schematic diagram in the resampled data, and its abscissa represents the effective mass-to-charge ratio number retained after resampling, Its ordinate indicates the value of the mass-to-charge ratio corresponding to the mass-to-charge ratio number. S2022) The effective mass-to-charge ratio calculation step is used to calculate the mass-to-charge ratio of the resampled mass spectrum data using the resampling algorithm; the resampling algorithm refers to: set the mass-to-charge ratio interval of the effective mass spectrum data after resampling to [y ₁ , y ₂ ], the number of mass-to-charge ratio coordinates after resampling is N; use the following formula to calculate the mass-to-charge ratio coordinates after resampling

Among them, N is greater than 10 ⁴ and less than 10 ⁵ , which has achieved a balance between algorithm accuracy and calculation speed. S2023) Interpolation processing step, for using the mass-to-charge ratio after resampling and the mass-to-charge ratio number to perform interpolation processing on the mass spectrogram after baseline correction, and transform the abscissa of the mass spectrogram after baseline correction from charge-to-mass ratio value to mass-to-charge Than no. In this embodiment, the mass-to-charge ratios of the resampled mass spectrum data are distributed within the mass-to-charge ratio range of 98.9 to 1003.1, and 10,000 sets of valid mass spectrum data are retained, and the mass-to-charge ratio coordinates after resampling are calculated using the following formula Corresponding to the effective mass spectrum data, there are a total of 10,000 mass-to-charge ratios, as shown in Figure 8, which is a schematic diagram of the effective mass-to-charge ratios in the resampled data of this embodiment, and its abscissa represents the number of effective mass-to-charge ratios retained after resampling. The ordinate indicates the value of the mass-to-charge ratio corresponding to the mass-to-charge ratio number.

In the process of interpolating the mass spectrum, the redundant mass spectrum data in the mass spectrum after baseline correction (as shown in Figure 6) is removed, and only the effective mass spectrum data for resampling is retained; the abscissa of the mass spectrum after baseline correction is given by The charge-to-mass ratio value is transformed into a mass-to-charge ratio number, and the ordinate remains unchanged, so that the resampling of each group of original mass spectrum data can be completed. As shown in FIG. The effective mass-to-charge ratio number after resampling, and its ordinate indicates the measured intensity value of the ion signal corresponding to the mass-to-charge ratio number. After the resampling step, on the mass spectrogram, the intervals with relatively small mass-to-charge ratios contain more sampled values, the intervals with relatively large mass-to-charge ratios contain fewer sampled values, and the smaller intervals with mass-to-charge ratios contain more information Corresponding to the assumption that the mass-charge ratio is relatively large.

Step S203) A standardization step, for standardizing ion signal intensity values in the resampled mass spectrum data, and performing ordinate transformation on the mass spectrum to obtain standardized mass spectrum data. As shown in FIG. 10 , the standardization step in step S203) specifically includes the following steps (step S2031) to step S2034). Step S2031) a signal intensity absolute value sum calculation step, used to calculate the sum S of the absolute value of ion signal intensity values in all resampled mass spectrum data; step S2032) a normalized signal intensity value sum setting step, used to set the normalized processing The sum of the absolute values of the ion signal strength values in all resampled mass spectrum data is a constant T, and in this embodiment, the constant is set to 10000; step S2033) signal strength value change multiple calculation step, in order to calculate the change of each signal strength value Multiple T/S; step S2034) signal strength value change step, for synchronously amplifying or synchronously reducing the signal strength value of each ion in the resampled mass spectrum data, performing ordinate transformation on the mass spectrogram, and signal strength The change factor of the value is T/S in step S2033). As shown in Figure 11 is the mass spectrogram of the standardized mass spectrometry data in this embodiment, the abscissa indicates the effective mass-to-charge ratio number after resampling, and the ordinate indicates the normalized intensity value of the ion signal corresponding to the mass-to-charge ratio number. The technical effect of the standardization step is that mapping the intensity of the mass spectrum data to a unified range can ensure that the distribution range of the intensity of each group of mass spectrum data is basically consistent, thereby enhancing the comparability of different mass spectrum data.

Step S3) Data model construction and cross-validation steps, for constructing a primary data model by using the mass spectrum data of the training samples and the group labels of the training samples, and performing an operation on the primary data model according to the mass spectrum data of the training samples n times (n is the number of training samples) of cross-validation processing, using the mass spectrum data and group labels of known training samples to perform machine learning and build a model. As shown in Figure 12, step S3) specifically includes the following steps: step S301) select a training sample as a standard training sample, and its group label is known; step S302) take the position of the standard training sample as the center of the circle, and use The length r is a radius, and a circular area is set on the flat plate; step S303) a matrix D is constructed according to the standardized mass spectrum data of other training samples in the circular area except the standard training sample, and the matrix D Each column of data corresponds to a group of standardized mass spectrometry data of a training sample; step S304) Acquire vectors according to the group labels of other training samples in the circular area except the standard training samples The group label of each training sample is recorded in the vector Middle; Step S305) Utilize the sparse learning optimization algorithm to establish the primary data model Step S306) Multiplying two or more sets of mass spectrum data of the standard training sample by the data model, arranging the products into a series according to the numerical value, rounding the median value, and obtaining the guess of the standard training sample Group label; since the group label in the present invention is all integers (0 or 1), it is necessary to round the digits after the decimal point to obtain an integer, which is the rounding process; step S307) compares the standard training sample Infer that the group label is the same as the known group label of the standard training sample, if the two are the same, it is determined that the group label of the standard training sample is guessed correctly, and the correctness counter is increased by one; step S308) sequentially counts each The training sample is used as a standard training sample, repeat steps S301) to step S307), carry out cross-validation processing on all training samples, and calculate the group label judgment accuracy rate of the training sample when the radius is r, the group The accuracy rate of different label judgment is the ratio of the value of the accuracy counter to the total number of training samples; step S309) adjusts the size of the radius r, repeats steps S301) to step S308), and calculates how many A group label judgment accuracy rate; step S310) select a maximum accuracy rate from the plurality of group label judgment accuracy rates, and obtain the value of the radius r corresponding to the maximum accuracy rate, which is the optimal value of the radius R.

Machine learning algorithm is a kind of algorithm that automatically analyzes and obtains laws from data, and uses the laws to predict unknown data. The invention adopts the Lasso regression algorithm in machine learning to analyze the mass spectrum data, and learns to build a model, including two processes of training and testing. The basic idea of the Lasso algorithm is to minimize the sum of squared residuals under the constraint that the sum of the absolute values of the regression coefficients is less than a constant, so as to generate some regression coefficients that are strictly equal to 0 and obtain an interpretable model.

In this embodiment, the Lasso algorithm is used to perform n times of cross-validation (n is the number of training samples), and the model obtained by each cross-validation is matched with 11 intensity thresholds 0, 0.1, ..., 1, and the corresponding 11 group labels are accurately judged rate; repeated n times, a total of n*11 data models (groupers) were obtained, and each data model corresponds to a group label judgment accuracy rate. Adjust the specific radius r=2.0mm, 2.2mm, 2.4mm..., 4.8mm, 5mm to obtain n*11*16 group label judgment accuracy rates, arrange the values of all group label judgment accuracy rates according to size, and find out The maximum accuracy rate value, and then find the radius corresponding to the maximum accuracy rate value, which is the optimal value of the radius R.

Step S4) a data model optimization step for constructing an optimized data model according to the cross-validation result. As shown in Figure 13, step S4) specifically includes the following steps: step S401) takes the position of a test sample as the center of the circle, and takes the length of the optimal value R of the radius in step S310) as the radius, and sets a Circular area; Step S402) Construct matrix D _W according to the standardized mass spectrum data of all training samples in the circular area, each column of data in the matrix D _W corresponds to a set of standardized mass spectrum data of a training sample; Step S403) Acquire vectors according to the group labels of all training samples in the circular area The group label of each training sample is recorded as an integer in the vector corresponding to the training sample Middle; step S404) establishes the optimized data model Build an optimized data model The sparse learning optimization algorithm is used in the process.

Step S5) a sample group judgment step, for obtaining the group label of the test sample by using the mass spectrum data of the test sample and the optimized data model. In said step S5), a group of mass spectrum data of a test sample is multiplied by the data model, and the product is rounded to obtain the group label of the test sample; or two groups of a test sample are The above mass spectrum data is multiplied by the data model, and the products are arranged into a series according to the numerical value, and the median value is rounded to obtain the group label of the test sample. In this embodiment, if the rounded result is 0, it can be considered that the person corresponding to the test sample has a mass spectrum data pattern associated with a certain disease, thereby assisting the doctor to make a diagnosis; if the rounded result is 1, it can be considered The person corresponding to the test sample does not have the mass spectrometry data pattern associated with the disease, thereby assisting the physician in making a diagnosis.

The invention provides a mass spectrometry data analysis method, which can construct a grouper model according to the groups of known body fluid samples, obtain the data model with the highest correct rate through multiple cross-validation of multiple training samples, and can process a large number of body fluid samples at the same time mass spectrometry data and group them according to body fluid sample composition. In clinical medicine, the technical solution of the present invention can be applied to assist in the intelligent diagnosis of diseases. By using computer technology to simultaneously detect multiple groups of blood samples of multiple people to be tested, it can be judged in a short time whether multiple people to be tested have a certain Disease-associated mass spectrometry data patterns assist physicians in rapid diagnosis.

The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications should also be considered Be the protection scope of the present invention.

Claims (12)

1. a mass spectrometry data analysis method, is characterized in that, comprises the steps: The sample data collection step is used to collect mass spectrum data of more than two body fluid samples and generate a mass spectrum according to the mass spectrum data; the body fluid samples include more than two training samples and at least one test sample; The training samples are divided into two or more groups, and the training samples of the same group have the same group label; The sample data preprocessing step is used to preprocess at least one set of mass spectrum data, perform coordinate transformation processing on the mass spectrum, and obtain standardized mass spectrum data of the training sample and the test sample; The data model construction and cross-validation steps are used to construct a primary data model using the standardized mass spectrum data of the training samples and the group labels of the training samples, and perform at least One-time cross-validation processing; A data model optimization step, for constructing an optimized data model according to the results of the cross-validation; and The sample group judgment step is used to obtain the group label of the test sample by using the standardized mass spectrum data of the test sample and the optimized data model. 2. mass spectrum data analysis method as claimed in claim 1, is characterized in that, The sample data collection step specifically includes the following steps: Obtain two or more body fluid samples; arranging all of said bodily fluid samples in a matrix on a plate; and The mass spectrometry data of the body fluid sample is collected by mass spectrometry and a mass spectrogram is generated; at least one set of mass spectrum data is collected for each body fluid sample. 3. mass spectrum data analysis method as claimed in claim 2, is characterized in that, The test sample is located in the middle of the flat panel, and the training sample surrounds the test sample; The flat plate includes, but is not limited to, a matrix metal plate; The group labels of any two adjacent training samples are different; The distance between any two adjacent body fluid samples is greater than or equal to 2mm and less than or equal to 5mm. 4. mass spectrum data analysis method as claimed in claim 1 or 2, is characterized in that, Each set of mass spectrometry data includes the mass-to-charge ratio value of an ion in the sample and the measured signal intensity value corresponding to the ion; Each set of mass spectrum data corresponds to a sampling point in the mass spectrum; The abscissa of each sampling point represents the mass-to-charge ratio of an ion, and the ordinate represents the measured intensity value of the signal corresponding to the ion. 5. mass spectrum data analysis method as claimed in claim 1, is characterized in that, The sample data preprocessing step specifically includes the following steps: a baseline correction step, for performing baseline correction processing on the mass spectrum data in the mass spectrogram; The resampling step is used to resample the mass-to-charge ratio of ions in the mass spectrum data after baseline correction by using a resampling algorithm, perform abscissa transformation on the mass spectrum, and unify the mass-to-charge ratios of all mass spectrum data to obtain resampling mass spectrometry data; and The standardization step is used to standardize ion signal intensity values in the resampled mass spectrum data, and perform ordinate transformation on the mass spectrum to obtain standardized mass spectrum data. 6. mass spectrum data analysis method as claimed in claim 5, is characterized in that, The baseline correction step specifically includes the following steps: The signal calculation step is used to calculate the baseline signal intensity corresponding to at least one mass-to-charge ratio value in a set of mass spectrum data by using a window function; a signal correction step for correcting the measured signal intensity corresponding to the mass-to-charge ratio based on the baseline signal intensity; and The signal calculation step and the signal correction step are repeated to sequentially complete the correction of each set of mass spectrum data of each body fluid sample. 7. mass spectrum data analysis method as claimed in claim 5, is characterized in that, The resampling step specifically includes the following steps: The effective mass-to-charge ratio selection step is used to select the effective mass-to-charge ratio range and the effective mass-to-charge ratio quantity; The effective mass-to-charge ratio calculation step is used to calculate the mass-to-charge ratio of the resampled mass spectrum data using a resampling algorithm; The interpolation processing step is used to interpolate the mass spectrum after baseline correction by using the mass-to-charge ratio after resampling and the mass-to-charge ratio number, and convert the abscissa of the mass spectrum after baseline correction from the charge-to-mass ratio value to the mass-to-charge ratio number . 8. mass spectrum data analysis method as claimed in claim 7, is characterized in that, The resampling algorithm refers to: Set the mass-to-charge ratio interval of effective mass spectrum data after resampling to [y ₁ , y ₂ ], and the number of mass-to-charge ratio coordinates after resampling to be N; Use the following formula to calculate the mass-to-charge ratio coordinates after resampling Wherein, N is greater than 10 ⁴ and less than 10 ⁵ . 9. mass spectrum data analysis method as claimed in claim 5, is characterized in that, Described standardization step specifically comprises the following steps: A signal intensity absolute value sum calculation step, used to calculate the absolute value sum S of ion signal intensity values in all resampled mass spectrum data; The normalized signal intensity value sum setting step is used to set the absolute value sum of the ion signal intensity values in all resampled mass spectrum data after the normalization process as a constant T; The signal strength value change multiple calculation step is used to calculate the change multiple T/S of each signal strength value; The signal intensity value changing step is used for synchronously amplifying or synchronously reducing the signal intensity value of each ion in the resampled mass spectrum data. 10. mass spectrum data analysis method as claimed in claim 1, is characterized in that, The data model construction and cross-validation steps specifically include the following steps: Choose a training sample as a standard training sample, whose group label is known; Set a circular area on the plate with the position of the standard training sample as the center and a specific length r as the radius; According to the standardized mass spectrum data of other training samples in the circular area except the standard training samples, a matrix D is constructed, and each column of data in the matrix D corresponds to a set of standardized mass spectrum data of a training sample; Acquire vectors according to the group labels of other training samples in the circular area except the standard training samples The group label of each training sample is recorded in the vector middle; Create a primary data model Multiplying more than two groups of standardized mass spectrum data of the standard training sample by the data model, arranging the products into a sequence according to the numerical value, rounding the median value, and obtaining the estimated group of the standard training sample Label; Comparing the estimated group label of the standard training sample with its group label, if the two are the same, it is determined that the group label of the standard training sample is guessed correctly, and the correctness counter is increased by one; Taking each training sample as a standard training sample in turn, repeating the above steps, performing cross-validation processing on all training samples, and calculating the group label judgment accuracy rate of the training samples when the radius is r, the group The accuracy rate of different label judgment is the ratio of the numerical value of the accuracy counter and the total number of training samples; Adjust the size of the radius r, repeat the above steps, and calculate the group label judgment accuracy rate when the radius r is a different value; and Select a maximum accuracy rate from two or more group label judgment accuracy rates, and obtain the optimal value R of the radius r corresponding to the maximum accuracy rate. 11. mass spectrum data analysis method as claimed in claim 1, is characterized in that, The data model optimization step specifically includes the following steps: Taking the position of a test sample as the center of the circle, and taking the length of the optimal radius R as the radius, set a circular area on the flat plate; Construct a matrix D _W according to the standardized mass spectrum data of all training samples in the circular area, and each column of data in the matrix D _W corresponds to a set of standardized mass spectrum data of a training sample; Acquire vectors according to the group labels of all training samples in the circular area The group label of each training sample is recorded in the vector corresponding to the training sample in the form of natural numbers in; and Building Optimal Data Models Using Sparse Learning Optimization Algorithms . 12. mass spectrum data analysis method as claimed in claim 1, is characterized in that, The sample group judgment step specifically includes the following steps: multiplying a set of mass spectrum data of a test sample by the data model, and rounding the product to obtain the group label of the test sample; or Multiplying more than two groups of mass spectrometry data of a test sample with the data model, arranging the products into a series according to the numerical value, rounding the median value to obtain the group label of the test sample.