AT527861A1 - DETERMINING A CANCER RISK SCORE - Google Patents

Abstract

A computer-implemented method for determining a cancer risk score for an individual comprises: - obtaining 1H NMR spectroscopy data (300a-c) for a biofluid sample of an individual at a computing device (100), wherein the 1H NMR spectroscopy data (300a-c) is indicative of an NMR signal intensity as a function of chemical shift; - evaluating the 1H NMR spectroscopy data (300a-c) with at least one trained machine learning model (114) of the computing device (100) with respect to a molecular profile (310a-c) of the biofluid sample, wherein the molecular profile includes all hydrogen peaks in the 1H NMR spectroscopy data that can be assigned to one or more of a metabolite, a protein, an amino acid, a micromolecule, and a macromolecule contained in the biofluid sample; - Classifying the molecular profile (310a-c) based on the evaluation with the at least one trained machine learning model (114) into at least a first class and a second class of molecular profiles, wherein the first class is representative of molecular profiles (310a) associated with healthy individuals and the second class is representative of molecular profiles (310b, c) associated with individuals who have a cancerous and/or pre-cancerous disease; - Determining, based on the classification of the molecular profile, a cancer risk score that indicates a probability of cancer occurring in the individual.

Description

15

20

25

30

Our Ref.: W07717 AT/CHP

DETERMINING A CANCER RISK SCORE

TECHNICAL FIELD

[0001] The present invention relates generally to the field of computer-implemented oncology. In particular, the present invention relates to a computer-implemented method for determining a cancer risk for an individual. The present invention further relates to a computer device configured to carry out the method, a computer program instructing the computer device to carry out the method

to carry out, and a computer-readable medium storing such a computer program.

BACKGROUND

[0002] Cancer, or cancer in general, affects millions of people and is one of the leading causes of death worldwide. The most common or widespread cancers are breast, lung, colon, and rectum/prostate cancer, but almost any organ in the human body can be affected by cancer, such as the liver, pancreas, kidneys, and others. Typically, the mortality rate from cancer can be reduced if it is detected and treated early. When cancer is detected early, it is more likely to respond to treatment and less likely to spread to other tissue, which can lead to a higher probability of survival with lower morbidity, as well as less costly treatment.

[0003] In many cases, cancer, cancers, or precancerous conditions can be detected or diagnosed based on laboratory test results of an individual's or patient's biofluid sample, such as blood, urine, and cerebrospinal fluid (CSF). Currently established and widely used methods for analyzing biofluid samples include blood chemistry tests, which measure the amount of certain substances in a blood sample, and complete blood count tests, which measure the number of red and white blood cells and platelets in a blood sample. Other cancer detection or analysis methods include tumor marker tests, which measure substances produced by cancer cells or other body cells in response to cancer, and urinalysis, which describes the color and content of a urine sample. Other approaches use liquid biopsy in combination with next-generation sequencing technologies for cancer detection.

[0004] In recent years, many developments have also been made towards computer-implemented or automated cancer detection or assessment methods in order to

For example, data-driven and/or individualized patient care based on

CHP:CHP 7/44

15

20

25

30

35

2 of laboratory test results of a biofluid sample from an individual, as described above. In general, artificial intelligence (AI) and machine learning (ML), a subset of AI that enables computers to learn from training data, have proven to be useful tools in oncology and can demonstrably improve the accuracy of medical care and patient outcomes. [0005] However, the AI- and ML-based methods currently used to detect cancer based on measurements of biofluid samples usually focus on specific substances present in a biofluid sample and are therefore usually limited to the detection of specific types of cancer. Furthermore, at least with some of the cancer detection methods currently in use, early detection of cancers or even precancerous lesions is hardly possible because, for example, the corresponding cancer signal in the

Early stages of cancer may not be detectable.

SUMMARY

[0006] It may therefore be desirable to provide an improved computer-implemented method and corresponding computer device for determining a cancer risk, which indicates the probability of cancer occurring in an individual. The method and computer device described herein may, in particular, allow or enable early detection of a cancerous condition (cancer) and/or a precancerous condition (precancerous stage).

[0007] This is achieved by the subject matter of the independent claims, with further embodiments being included in the dependent claims and the following description. [0008] Aspects of the present disclosure relate to a computer-implemented method for determining a cancer risk score for an individual, to a computer device configured to perform a method for determining a cancer risk score for an individual, to a corresponding computer program, and to a computer-readable medium on which such a computer program is stored. Any disclosure presented herein with reference to one aspect of the present disclosure also applies to any other aspect of the present disclosure.

[0009] In one aspect of the present disclosure, a computer-implemented method for determining, calculating, and/or estimating a cancer risk score for an individual is provided. The method comprises:

- Obtaining hydrogen 1-nuclear magnetic resonance, 1H NMR spectroscopy data for a biofluid sample of an individual on a computing device, wherein the 1H NMR spectroscopy data is indicative of an NMR signal intensity as a function of chemical shift;

- Evaluating the 1H NMR spectroscopy data with at least one trained machine learning model of the computer device with respect to a molecular profile of the biofluid sample,

where the molecular profile includes all hydrogen peaks in the 1H NMR spectroscopy data

15

20

25

30

35

3 which can be associated with and/or are associated with one or more of a metabolite, a protein, an amino acid, a micromolecule and a macromolecule contained in the biofluid sample; - classifying the molecular profile based on the evaluation with the at least one trained machine learning model into at least a first class and a second class of molecular profiles, wherein the first class is representative of molecular profiles associated with healthy individuals and the second class is representative of molecular profiles associated with individuals who have a cancerous and/or pre-cancerous disease; and - determining, based on the classification of the molecular profile, a cancer risk score which indicates a probability of cancer occurring in the individual. [0010] Cancer can affect the overall molecular composition of one or more of an individual's biofluids, such as the composition of blood, urine, cerebrospinal fluid and other biofluids. This influence may include changes in the proportion of specific proteins, metabolites, micro- and/or macromolecules that are either directly released by the cancer cells or whose release is directly or indirectly induced by them. [0011] The method described here uses highly detailed 1H NMR (nuclear magnetic resonance) spectroscopy data, also called high-frequency NMR spectroscopy data or proton NMR spectroscopy data, obtained from a biofluid sample of a patient or individual, in combination with machine learning (ML) to analyze and/or evaluate the molecular profile of the biofluid sample, for example, to detect the above-mentioned cancer-induced changes in the biofluid sample. The 1H NMR spectroscopy data and/or the molecular profile contained therein can provide information about the chemical environment of each hydrogen atom (1H) in the biofluid sample. This means that, depending on the structure of the molecule containing the hydrogen atom, a different chemical environment may be present, changing the chemical shift of that specific hydrogen atom. For these reasons, high-frequency NMR or 1H NMR spectroscopy data from a biofluid sample can provide an extremely high information density that can be used to distinguish between healthy and sick patients with a cancerous and/or pre-cancerous condition. [0012] Pre-cancerous conditions generally refer to health conditions or lesions that have an increased risk of developing into cancer but are not yet classified as cancer. The accompanying cellular changes are often abnormal but not completely uncontrolled, as is the case with cancer. An example of a precancerous condition is dysplasia of the cervix, known as cervical intraepithelial neoplasia (CIN), which can lead to cervical cancer. An example of a cancer is invasive cervical cancer, in which the cancer cells spread uncontrollably.

divide and penetrate the surrounding tissue.

15

20

25

30

35

4 [0013] Accordingly, the term precancerous disease in the context of the present disclosure may refer to diseases that are either direct precursors of cancer, such as IPMN, a tumor (growth) of the pancreas that often later develops into cancer, or to diseases that increase the risk of cancer, such as chronic pancreatitis. [0014] While currently used cancer detection methods are generally limited to individual signals that can be attributed to a specific tumor marker or specific signals in the respective measurement, such as a specific molecule, metabolite, or lipoprotein, the method according to the present disclosure takes into account all signatures in the 1H NMR spectroscopy data. In this way, the entire molecular profile of an individual's biofluid sample can be considered, analyzed, and evaluated, enabling reliable discrimination between healthy and cancer-stricken patients or individuals.

[0015] Due to the high information content of 1H NMR spectroscopy data and the fact that cancer can affect the molecular profile and metabolism of the individual at a very early stage, various changes in the molecular profile caused by or associated with cancer can be detected at an early stage of cancer or even pre-cancerous disease, thus enabling early detection and treatment.

[0016] Furthermore, the method described here is not limited to a specific type of cancer or cancerous disease, but can be advantageously used to detect various types of cancer and/or even to distinguish between different types of cancer, cancerous diseases and/or precancerous lesions.

[0017] Furthermore, the computer-implemented method described herein can be considered or used as a screening method that can be integrated into a normal or routine examination of patients or individuals, e.g., when a biofluid sample is taken from the individual during a routine checkup at a healthcare provider or in the hospital. Furthermore, 1H NMR is also well suited to high-throughput screening approaches, is relatively inexpensive, and has high reproducibility. Therefore, the method described here can assist healthcare providers in the early detection or identification of individuals at risk for cancer and/or precancerous conditions. As a result, the method described here can enable reliable and accurate risk stratification of an individual with respect to having cancer, having precancerous conditions, and/or being healthy or not suffering from cancer. This, in turn, can enable better and more efficient use of healthcare resources as well as cost savings.

[0018] In the context of the present disclosure, the term "individual" may refer generally to a vertebrate, including animals and humans. The term "individual"

can be used here synonymously or interchangeably with subject or patient.

15

20

25

30

35

5 [0019] The 1H NMR spectroscopy data used herein may refer to data acquired and/or generated using a high-frequency NMR spectrometer, e.g., at a frequency of about 500 MHz or higher, preferably at 600 MHz or higher. The 1H NMR spectroscopy data, also referred to as a 1H NMR spectrum, may be acquired, for example, using a pulse sequence similar to CPMGPR1D. [0020] In general, the 1H NMR spectroscopy data may include or indicate the free induction decay (FID), which may be measured as raw 1H NMR spectroscopy data in a 1H NMR spectroscopy of the biofluid sample. The measured FID can optionally be converted into a spectrum by Fourier transformation, and further optionally referenced and/or converted into a chemical shift in parts per million (ppm), for example, based on normalization to a reference peak in the spectrum, such as the peak of lactate or anomeric D-glucose doublet. Accordingly, the term "1H NMR spectroscopy data indicative of an NMR signal intensity as a function of chemical shift" includes both 1H NMR spectroscopy data expressed as a function of chemical shift and 1H NMR spectroscopy data expressed as a function of FID or another quantity from which the NMR signal intensity as a function of chemical shift can be derived. [0021] As used herein, obtaining the 1H NMR spectroscopy data may comprise accessing and/or retrieving the 1H NMR spectroscopy data. For example, the 1H NMR spectroscopy data may be accessed and/or retrieved in at least one memory or data storage of the computing device performing the method of the present disclosure, from a memory or data storage of another computing device, and/or from a remote data storage (a database, further storage, cloud storage, or the like). Accordingly, retrieving the 1H NMR spectroscopy data may optionally comprise downloading the 1H NMR spectroscopy data from an external computing device or data storage. [0022] Additionally or alternatively, obtaining the 1H NMR spectroscopy data may comprise receiving the 1H NMR spectroscopy data, e.g., from a computer or device. B. from a different computing device than the computing device accessing the data. Accordingly, obtaining the 1H NMR spectroscopy data may comprise one or more of the following steps: receiving the 1H NMR spectroscopy data, storing the 1H NMR spectroscopy data in the memory or data storage of the computing device, and retrieving the 1H NMR spectroscopy data by the computing device. [0023] As used herein, evaluating the 1H NMR spectroscopy data with the at least one trained machine learning model of the computing device with respect to the molecular profile of the biofluid sample may comprise processing and/or analyzing the 1H NMR spectroscopy data with the aid of the at least one trained ML model with respect to all hydrogen peaks contained in the 1H NMR spectroscopy data. Such evaluation of the

1H-NMR spectroscopy data related to the molecular profile can facilitate the evaluation and/or

15

20

25

30

35

6 Analysis of the molecular profile of the biofluid sample with respect to one or more molecular reference profiles of one or more reference individuals. For example, the at least one ML model can be trained based on 1H NMR spectroscopy reference data of one or more reference individuals. [0024] Based on the evaluation of the molecular profile and/or the 1H NMR spectroscopy data of the biofluid sample, the at least one trained machine learning model can perform a binary classification or multi-class classification into at least two classes of molecular profiles, namely at least the first class of molecular profiles associated with healthy individuals and the second class of molecular profiles associated with a cancer and/or a precancerous condition. Accordingly, the at least one machine learning model can refer to a classification model or algorithm trained according to a machine learning approach. [0025] As used herein, determining the cancer risk score may also comprise calculating and/or assessing the risk and/or probability of the occurrence and/or presence of a cancerous and/or pre-cancerous condition in the individual. This may comprise assessing, calculating and/or determining the probability of the occurrence and/or presence of cancer or a cancerous and/or pre-cancerous condition in the individual. In other words, determining the cancer risk score may comprise, for example, calculating and/or determining the probability of the occurrence of cancer in the individual based on or using the at least one trained machine learning model of the computing device. [0026] The cancer risk score used here may refer to a numerical measure that indicates the determined risk and/or probability of the occurrence and/or presence of a cancerous and/or pre-cancerous condition in an individual. The cancer risk score may be given on an arbitrary scale that ranges from a minimum value, e.g. B. zero or 0, up to a maximum value, e.g. one or 100. Any other scale, including relative and absolute scales, may be used to represent the cancer risk score. It should be noted that a variety of cancer risk scores may be calculated, e.g. using a variety of different machine learning models or algorithms, as described in more detail below. [0027] Optionally, the cancer risk score may be output by the computing device, for example, on a user interface of the computing device. Further, optionally, additional or contextual information associated with the cancer risk score may be determined by the computing device, which may optionally be provided as output from the computing device. Such contextual information may include a risk level or stage, such as low risk, moderate risk, high risk, for a cancerous and/or pre-cancerous condition occurring and/or present in the individual. Alternatively or additionally, the contextual information may be an indication of a

cancer type and/or an estimated cancer stage.

15

20

25

30

35

7 [0028] The computing device described here can refer to any data processing device having one or more processors for data processing. The computing device can be embodied as a standalone computing device, as a server, and/or as a computer network with a plurality of cooperating computing devices, such as a cloud computing system or server system. Alternatively or additionally, the computing device can be embodied at least partially as a mobile device, such as a smartphone, a tablet computer, a notebook, or the like. [0029] According to one embodiment, classifying the molecular profile comprises determining and/or calculating a classification result indicating a probability for at least one of the first class and the second class, wherein the cancer risk score is determined based on the classification result. Accordingly, the cancer risk score can comprise or correspond to the classification result calculated or generated by the at least one ML model. Classifying the molecular profile may, for example, involve drawing a conclusion as to whether the molecular profile of the biofluid sample belongs to the first class, the second class, and optionally to one or more further classes. Alternatively or additionally, classifying the molecular profile may comprise calculating a probability for the molecular profile of the biofluid sample to belong to the first class of molecular profiles, to the second class of molecular profiles, and optionally to one or more further classes of molecular profiles. In other words, the at least one machine learning model may be trained to determine a probability for a binary classification and/or multiclass classification of the molecular profile of the biofluid sample. It should be noted that in the case of a binary classification, only one classification result of one of the first and second classes may be calculated, and the classification result of the other of the first and second classes may be calculated based thereon, e.g., based on the subtraction of the determined classification result from one or 100%. [0030] According to one embodiment, the biofluid sample is selected from the group consisting of a blood serum sample, a blood plasma sample, a blood sample, a urine sample, and a cerebrospinal fluid sample. In other words, the biofluid sample may comprise one or more blood serum samples, blood plasma samples, blood samples, urine samples, and cerebrospinal fluid samples. [0031] Depending on the type of cancer or cancerous disease, the composition of one or more biofluids of an individual may be altered, modified, and/or transformed, wherein such changes may optionally be induced at different stages of the cancerous disease. In the majority of cancers and cancerous diseases, it is assumed that the composition or molecular composition of an individual's blood changes, for example due to altered metabolism and protein composition, thereby altering the proportion of one or more proteins, metabolites,

amino acids, micro- and/or macromolecules that are either directly derived from the

15

20

25

30

35

8 Cancer cells are released into the blood or their release is induced directly or indirectly by the cancer, e.g., via biological networks. In these cases, a blood sample, a blood serum sample, and/or a blood plasma sample can be used as a biofluid sample to generate the 1H NMR spectroscopy data and determine the cancer risk score. Other cancer types or cancer diseases, such as tumors of the bladder, kidney, or urinary tract, can, on the other hand, alter the molecular composition of the urine in the early cancer stage, and a urine sample can be used as a biofluid sample to generate the 1H NMR spectroscopy data and determine the cancer risk score, which can enable early cancer detection. Other cancer types or cancer diseases, such as tumors of the bladder, kidney, or urinary tract, can, However, certain cancers, such as brain tumors, can alter the molecular composition of the cerebrospinal fluid, and a cerebrospinal fluid sample can be used as a biofluid sample to generate the 1H NMR spectroscopy data and determine the cancer risk score. Accordingly, the use of one or more blood serum samples, blood plasma samples, blood samples, urine samples, and cerebrospinal fluid samples can enable the earliest possible diagnosis or cancer detection. [0032] According to one embodiment, the 1H NMR spectroscopy data is indicative of an NMR signal intensity in binned increments of chemical shift, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. It should be noted that the aforementioned ppm values and ranges apply to each incremental chemical shift bin discussed or disclosed herein. In general, the use of such fine-grained binning of the 1H NMR spectrum can increase the information content of the 1H NMR spectroscopy data and thus enable the identification of fine-grained changes in the molecular profile caused by cancer or cancerous disease at a very early stage of the cancer. [0033] According to one embodiment, the method further comprises converting the 1H NMR spectroscopy data into a binned data structure based on the figure and/or assigning the NMR signal intensity to binned chemical shift increments, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. Alternatively or additionally, the method may further comprise binning the 1HNMR spectroscopy data into a plurality of incremental chemical shift bins, each bin having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. Optionally, the binned 1H NMR spectroscopy data may be normalized to a median of at least a subset of the bins. An advantage of such normalization may be a reduction in the variability between

different pre-analytical approaches.

15

20

25

30

35

9 [0034] In an exemplary implementation, a biofluid sample may be analyzed with a high-frequency NMR spectrometer, e.g., at or above about 500 MHz with a pulse sequence similar to CPMGPR1D. The free induction decay (FID) may be measured and Fourier transformed into a spectrum, and the Fourier transformed spectrum may be normalized to a reference peak, such as lactate, anomeric D-glucose doublet, or another peak. Further, the baseline may be corrected, and the spectrum may then be binned increments of less than or equal to 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. This makes it possible to obtain a significant amount of data for deciphering the molecular profile of the biofluid sample and provides information about all hydrogen atoms present, where each bin may represent a specific position in the 1H NMR spectrum or in the spectroscopy data. [0035] According to one embodiment, the at least one trained machine learning model is a machine-learned classifier configured to process, as input data, the 1H NMR spectroscopy data in a data structure of binned chemical shift increments. In other words, the at least one ML model may be trained or machine-learned to receive and/or process the 1H NMR spectroscopy data in a data structure of binned chemical shift increments and to provide, as output, a classification result and/or the at least one cancer risk score, which may represent, include, and/or correspond to the classification result. [0036] According to one embodiment, the at least one machine learning model is trained and/or configured to identify at least one molecular signature of a cancer and/or a pre-cancerous condition in the molecular profile of the biofluid sample and/or in the 1H NMR spectroscopy data. In general, the ML model can be trained and/or configured to identify and/or determine one or more molecular signatures of cancer and/or pre-cancerous conditions in the 1H NMR spectroscopy data. The one or more molecular signatures can be identified at any position or location in the spectrum, for example, in one or more chemical shift bins. Furthermore, a molecular signature of a cancer and/or pre-cancerous condition can comprise any alteration and/or change in the molecular profile of the biofluid sample compared to one or more molecular profiles of healthy individuals. Such a molecular signature or such a change and/or alteration of the molecular profile compared to the molecular profiles of healthy individuals may, for example, include an additional hydrogen peak in the 1H NMR spectroscopy data caused by a cancer and/or a pre-cancerous condition, a hydrogen peak with increased height or size, a hydrogen peak with reduced height or size, a change in the width of a hydrogen peak, an overlap of several hydrogen peaks, the removal of a hydrogen peak, a shift in ppm of one or

several hydrogen peaks or a combination thereof.

15

20

25

30

35

10 [0037] According to one embodiment, the determined cancer risk score can be used as a computational biomarker that indicates a pathogenic molecular signature and/or molecular signatures of a cancerous and/or pre-cancerous disease in the molecular profile of the biofluid sample. In other words, the determined cancer risk score can be a computer-implemented biomarker that informs about pathogenic molecular signatures in the individual's biofluid sample. As described above, one or more different molecular signatures can be used to calculate cancer risk. Therefore, different computational biomarkers in the form of different cancer risk scores for different types of cancer, cancers, and/or pre-cancerous conditions can also be calculated according to the method described herein. Accordingly, the method described herein can be applied or used for many different types of cancers and/or pre-cancerous conditions, thereby providing a versatile approach or method for cancer detection and/or cancer risk assessment. [0038] According to one embodiment, identifying the at least one molecular signature comprises identifying one or more bins in the 1H NMR spectroscopy data that are associated with one or more hydrogen peaks and/or contain one or more hydrogen peaks that indicate a cancer-induced alteration of the molecular profile compared to molecular profiles of healthy individuals and/or reference individuals. Based on the identified one or more bins in the 1H NMR spectroscopy data, the cancer risk score can be calculated. Accordingly, the at least one ML model can be trained to identify one or more cancer-related alterations in one or more hydrogen peaks with respect to one or more molecular profiles of healthy individuals. [0039] According to one embodiment, identifying the at least one molecular signature comprises identifying one or more bins and/or a chemical shift range in the 1H NMR spectroscopy data that are associated with an overlap of hydrogen peaks that can be attributed to a variety of metabolites, proteins, amino acids, micromolecules, and macromolecules in the biofluid sample. In other words, the ML model can be trained to identify an overlap of hydrogen peaks that can be caused by a cancer and/or a precancerous condition and that can comprise one or more chemical shift bins in the 1H NMR spectroscopy data. This can make it possible to identify complex patterns of molecular signatures caused by the cancer and/or the precancerous condition and/or to consider complex patterns of molecular signatures when calculating the cancer risk. This can increase the overall versatility of the method, for example, to enable the determination of different types of cancer and/or precancerous conditions. [0040] According to one embodiment, identifying the at least one molecular

Signature the identification of one or more bins and/or a range of chemical

15

20

25

30

35

11 Shift in the 1H NMR spectroscopy data that cannot be assigned to individual metabolites, proteins, amino acids, micromolecules, and macromolecules in the biofluid sample. Accordingly, molecular signatures that cannot be assigned to and/or are not associated with individual metabolites, proteins, amino acids, micromolecules, and macromolecules in the biofluid sample can also be taken into account when calculating the cancer risk score. This can also increase the overall versatility of the method, for example, to enable the determination of various types of cancer and/or precancerous lesions. [0041] According to one embodiment, the at least one molecular signature is associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H NMR spectroscopy data that are associated with one or more metabolites, proteins, amino acids, micromolecules, and macromolecules in the biofluid sample relative to a biofluid sample from one or more healthy individuals. A change in the proportion of a hydrogen peak can relate to one or more of the following quantities: height of the peak, size of the peak, integral of the peak, area of the peak, mean value of the peak, center of the peak, location or position of the peak in the 1H NMR spectrum, or a combination thereof. In general, by identifying changes in the proportion of one or more hydrogen peaks to classify the molecular profile into at least the first and second classes, the cancer risk score can be accurately and reliably determined. [0042] According to one embodiment, evaluating the 1H NMR spectroscopy data with respect to the molecular profile comprises analyzing all hydrogen peaks in the 1H NMR spectroscopy data that are associated with one or more metabolites, proteins, amino acids, micromolecules, and macromolecules in the biofluid sample. By analyzing all hydrogen peaks in the molecular profile, several or all molecular signatures induced by the cancerous and/or pre-cancerous condition in a molecular profile can be taken into account when determining the cancer risk score, thereby improving the accuracy and robustness of the determination and increasing the versatility of the determination. [0043] According to one embodiment, evaluating the 1H NMR spectroscopy data with respect to the molecular profile comprises detecting a pattern of hydrogen peaks associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H NMR spectroscopy data compared to a biofluid sample of one or more healthy individuals. Alternatively or additionally, the at least one machine learning model can be trained to detect a pattern of hydrogen peaks associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H NMR spectroscopy data compared to a biofluid sample of one or more healthy individuals. For example, complex correlations between cancers and molecular signatures in an individual’s molecular profile can occur, leading to patterns of hydrogen peaks in the 1H NMR spectroscopy data that can be used to determine the

Cancer risk scores can be detected. By detecting such patterns of

15

20

25

30

35

12 hydrogen peaks, the cancer risk score can be calculated with high accuracy, reliability, and robustness. [0044] According to one embodiment, obtaining the 1H NMR spectroscopy data comprises acquiring raw 1H NMR spectroscopy data (1H NMR spectroscopy raw data) with an NMR spectrometer at a frequency above about 500 MHz. The raw 1H NMR spectroscopy data may include or indicate the free induction decay as measured in 1H NMR spectroscopy. Accordingly, acquiring raw 1H NMR spectroscopy data with an NMR spectrometer at a frequency above about 500 MHz may comprise measuring the free induction decay with an NMR spectrometer. [0045] In an exemplary implementation, the method performs further spectral processing of the 1H NMR spectroscopy raw data and/or peak alignment of the 1H NMR spectroscopy raw data. This may allow for correcting shifts in the position or chemical shifts of hydrogen peaks of different biofluid samples, for example, such that hydrogen peaks associated with a particular metabolite but measured on different biofluid samples are located at the same position in the corresponding 1H NMR spectroscopy data. In this way, the comparability of 1H NMR spectroscopy data of different biofluid samples can be improved. [0046] According to one embodiment, the spectral processing of the 1H NMR spectroscopy raw data comprises one or more of chemical shift referencing, phase adjustment, and baseline correction of the 1H NMR spectroscopy raw data. Accordingly, one or more of these preprocessing steps can be applied to the 1H NMR spectroscopy raw data, which can increase the comparability of the 1H NMR spectroscopy data of different biofluid samples and improve the robustness and reproducibility of the determination of the cancer risk score with respect to different biofluid samples. [0047] According to one embodiment, the method further comprises one or more of: normalization, scaling, binning, and filtering of the 1H NMR spectroscopy raw data and/or the 1H NMR spectroscopy data. This can also increase the comparability of 1H NMR spectroscopy data of different biofluid samples and improve the robustness and reproducibility of the determination of the cancer risk score with respect to different biofluid samples. [0048] According to one embodiment, the at least one machine learning model for classification is trained with one or more statistical machine learning algorithms. In general, statistical machine learning algorithms can use statistical methods to develop models that can learn from data and make predictions or decisions based on the learning. Example and non-limiting statistical machine learning algorithms include a voting classifier, an ensemble method, and logistic regression. [0049] A voting classifier is a machine learning model that is based on an ensemble

numerous models trained and an output or at least one class based on the

15

20

25

30

35

13 highest probability of the selected class as output. Ensemble methods, on the other hand, are techniques that aim to improve the accuracy of model results by combining multiple models instead of using a single model. The combined models can thus increase the accuracy of the classification results obtained. Furthermore, logistic regression refers to a statistical method that can be used to build machine learning models where the dependent variable is dichotomous and/or binary. Accordingly, logistic regression can be used particularly for binary classification into two classes, such as the first and second classes of molecular profiles. [0050] According to one embodiment, classifying the molecular profile into at least the first class and the second class of molecular profiles comprises classifying the molecular profile into at least one healthy class of molecular profiles associated with healthy individuals and a non-healthy or diseased class of molecular profiles associated with non-healthy or diseased individuals based on the evaluation of the 1HNMR spectroscopy data regarding the molecular profile of the biofluid sample with a first trained machine learning model. Furthermore, the method comprises, after determining that the molecular profile of the biofluid sample is associated with or lies within the non-healthy or diseased class of molecular profiles, classifying the molecular profile into a non-cancer class of molecular profiles associated with non-cancerous individuals and a cancer class of molecular profiles associated with cancerous and/or pre-cancerous individuals based on the evaluation of the 1H NMR spectroscopy data with respect to the molecular profile of the biofluid sample with a second trained machine learning model. The first trained machine learning model and the second machine learning model differ from each other in this respect. In particular, the first and second machine learning models may differ from each other with respect to the training of the respective models. For example, the first ML model can be trained to provide a binary classification of the molecular profile of the biofluid sample into the healthy class of molecular profiles associated with healthy individuals and the unhealthy or diseased class of molecular profiles associated with unhealthy or diseased individuals, while the second ML model can be trained to provide a binary classification of the molecular profile of the biofluid sample into the non-cancer class of molecular profiles associated with non-cancerous individuals and the cancer class of molecular profiles associated with cancerous and/or pre-cancerous individuals. Alternatively or additionally, different types of ML models can be used for the first and second ML models, such as different statistical ML models or algorithms. Alternatively or additionally, different classifications can be used for the first and second ML models, such as a classification related to the cancer stage, e.g. B. metastatic and non-metastatic, and/or a classification in relation to a sub-disease. Alternatively or additionally, the

first and second ML model in one or more parameters, the selected algorithm

14 / 44

15

20

25

30

35

14 and the feature engineering methods. For example, principal component analysis (PCA) and/or t-distributed stochastic neighbor embedding (tSNE) can be used to generate features based on unsupervised learning for the first and/or second ML model. [0051] Accordingly, different ML models can be used sequentially or consulted to determine the cancer risk score. Applying a multi-stage approach to determining the cancer risk score can generally increase the accuracy and robustness of the overall method for determining the cancer risk score. The versatility of the overall method for calculating the cancer risk score can also be increased. [0052] According to one embodiment, the method further comprises determining, based on the classification of the molecular profile of the biofluid sample with the first trained machine learning model, a first cancer risk score indicating a probability that the molecular profile is associated with or lies in the healthy class and/or the non-healthy or diseased class. The method further comprises determining, based on the classification of the molecular profile of the biofluid sample with the second trained machine learning model, a second cancer risk score indicating a probability that the molecular profile is associated with or lies in the non-cancer class and/or the cancer class of the molecular profile, and determining the cancer risk score based on the first and second cancer risk scores. For example, the classification result and/or the second cancer risk score of the second ML model can be used as or form the cancer risk score. Alternatively or additionally, both the first and the second cancer risk score can be combined, e.g., in accordance with a predefined metric, to determine the cancer risk score. Alternatively or additionally, a calibration classifier can be used to provide refined probabilities for the classification results or the first and/or second cancer risk scores. [0053] According to one embodiment, the method further comprises, after determining that the molecular profile of the biofluid sample is associated with or lies in the cancer class of molecular profiles, classifying the molecular profile into a plurality of classes of molecular profiles based on the evaluation of 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a third trained machine learning model, each class being associated with a particular type of cancer and/or pre-cancer. Optionally, a third cancer risk score can be determined, which indicates a probability that the molecular profile is associated with a specific type of cancerous and/or precancerous disease. Accordingly, a type of cancer, cancerous and/or precancerous disease can be identified using the method described here.

procedure can be determined.

15

20

25

30

35

15 [0054] The third ML model may differ from one or both of the first and second ML models. In particular, the third ML model may differ from the first and/or second ML models in terms of training. In particular, the third ML model may be trained to provide a multi-class classification result for classifying the molecular profile into multiple classes of molecular profiles, each class being associated with a specific type of cancerous and/or pre-cancerous condition, while the first and second ML models may be trained to provide a binary classification result, as described above. Optionally, a different type of ML model may be used for the third ML model than for the first and second ML models. Alternatively or additionally, different classifications may be used for the first, second and third ML models, such as a classification related to the cancer stage, e.g. B. metastatic and non-metastatic, and/or a classification with respect to a sub-disease. Alternatively or additionally, the first, second and third ML models may differ in one or more parameters, the chosen algorithm and the feature engineering methods. For example, principal component analysis (PCA) and/or t-distributed stochastic neighborhood embedding (tSNE) may be used to generate features based on unsupervised learning for at least one of the first, second and third ML models. [0055] Another aspect of the present disclosure relates to a computing device configured to perform the steps of the method as described above and below. It is emphasized that each feature, function, element and/or step described herein with reference to the method may be a feature, function and/or element of the computing device and vice versa. [0056] The computing device comprises one or more processors for data processing. Furthermore, the computing device may comprise a data store and/or a memory for storing data, such as the 1H NMR spectroscopy data, one or more cancer risk scores, and/or other data. Furthermore, the data store and/or the memory may store software instructions and/or a computer program that, when executed by a computing device, instructs the computing device to perform steps of the method described herein and hereinafter. [0057] Furthermore, the at least one machine learning model, e.g., one or more of the first ML model, the second ML model, and the third ML model, may be implemented in software and/or hardware on the computing device. [0058] In an exemplary implementation, the computing device may comprise one or more communication interfaces configured to communicate with one or more remote devices, e.g., an external data store or database, and/or one or more remote computing devices. At least part of the 1H NMR spectroscopy data can be

Communication interface from an external data storage or database

15

20

25

30

35

16. Alternatively or additionally, the calculated cancer risk score or related information may be transmitted to one or more remote devices or stored in the external data storage or database. [0059] Another aspect of the present disclosure relates to a computer program that, when executed by a computing device, instructs the computing device to perform steps of the method described above and below. [0060] Another aspect of the present disclosure relates to a non-transitory computer-readable medium having stored thereon a computer program that, when executed by a computing device, instructs the computing device to perform steps of the above and below

described procedure.

BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Exemplary embodiments are further described with reference to the figures, wherein:

[0062] Figure 1 shows a computer device according to an exemplary embodiment; [0063] Figure 2 shows 1H NMR spectroscopy data illustrating the steps of a computer-implemented method for determining a cancer risk score;

[0064] Figure 3A shows a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to an exemplary embodiment;

[0065] Figures 3B to 3D each show 1H NMR spectroscopy data that can be used as input for the method shown in Figure 3A; and

[0066] Figures 4 to 6 each show a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to exemplary embodiments.

[0067] The figures are only schematic and not to scale. In principle, identical or similar parts, elements and/or steps in the figures are represented by identical or similar

provided with reference numbers.

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS

[0068] Figure 1 shows a computer device or computer system 100 configured to determine a cancer risk score according to an exemplary embodiment.

[0069] The computer device 100 includes a processing circuit 110 with one or more processors 112 for data processing.

[0070] The computing device 100 further includes at least one machine learning model 114. The ML model 114 may be implemented in software and/or hardware and may be part of the

Processing circuit 110 of the computing device 100. The computing device 100 may

15

20

25

30

35

17 optionally comprise a plurality of machine learning models 114, such as a first, a second and a third ML model, as described in more detail with reference to Figure 6 below, for example. The optional plurality of ML models are shown together in Figure 1 with the reference number 114. [0071] The computing device 100 further comprises at least one data store 116 and/or a memory 116 for storing data and/or software instructions, for example in the form of a computer program, for instructing the computing device 100 to carry out steps of the method for determining the cancer risk score, as described above and below. [0072] The computing device shown in Figure 1 further comprises a user interface 118 for outputting data and/or information, such as the cancer risk score, to a user of the computing device 100 and/or for controlling the computing system 100 by the user. [0073] Furthermore, the computer system 100 comprises a communication interface 120 for communicatively and/or operatively coupling the computer device 100 to an external computer device 500 and/or for coupling the computer system 100 to one or more external data sources 500, for example, to receive or obtain 1H NMR spectroscopy data from them and/or to transmit data to them, such as the cancer risk score. [0074] Figure 2 shows 1H NMR spectroscopy data in 200 to illustrate the steps of a computer-implemented method for determining a cancer risk score. In particular, Figure 2 shows the NMR signal intensity in arbitrary units as a function of the chemical shift in parts per million, ppm. [0075] Figure 2 shows an exemplary molecular profile 210a of a healthy individual (solid line) compared to a molecular profile 210b of an individual with cancer and/or pre-cancer (dashed line). The molecular profiles 210a, b comprise or consist of all hydrogen peaks of the 1H NMR spectroscopy data 200. [0076] Furthermore, the molecular profiles 210a, b shown in Figure 2 comprise groups or peaks 220 that can be assigned to metabolites, proteins, amino acids, micromolecules, and macromolecules that are assigned to and/or associated with the hydrogen peaks shown in the respective molecular profiles 210a, b. [0077] As can be seen from the comparison of the molecular profile 210a with the molecular profile 210b, the molecular profiles 210a, b of healthy individuals and individuals with a cancerous and/or pre-cancerous disease differ in several ranges of the chemical shift. Such differences may be referred to herein as molecular signatures associated with a cancerous and/or pre-cancerous disease. Based on these differences and/or molecular signatures, the ML model 114 and/or the computing device 100 may determine the cancer risk score, as described in more detail above and below.

described.

15

20

25

30

35

18 [0078] For example, a peak 222 attributable to choline phospholipid is present in the molecular profile 210b of the cancerous individual, whereas this peak is absent in the profile 210a of the healthy individual. Also, the proportions of the taurine peak 220, the glutamine peak 224, the glutamine and glutamate peak 226, the amino acid peaks 228, the lipid and fatty acid peaks 230, 232, and others differ between the molecular profile 210a of the healthy individual and the molecular profile 210b of the cancerous and/or pre-cancerous individual. One or more of such differences may be used by the ML model 114 and/or the computing device 100 to determine the cancer risk score, as described in more detail above and below. [0079] Figure 3A shows a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to an exemplary embodiment. Figures 3B to 3D each show 1H NMR spectroscopy data 300a, 300b, 300c that can be used as input to the method of Figure 3A. In particular, Figures 3B to 3D each show the NMR signal intensity in arbitrary units as a function of chemical shift in parts per million, ppm. Therein, the 1H NMR spectroscopy data 300a of Figure 3B corresponds to or shows a molecular profile 310a associated with a healthy individual. The 1H NMR spectroscopy data 300b of Figure 3C corresponds to or shows a molecular profile 310b associated with an individual with a pre-cancerous condition, such as cancer. B. pancreatitis, and the 1H NMR spectroscopy data 300c of Figure 3D correspond to or show a molecular profile 310c associated with an individual with a cancerous disease, such as pancreatic cancer. In the following, reference may be made generally to Figures 1 to 3D. [0080] The method comprises a step S1 of obtaining 1H NMR spectroscopy data 300a-c for a biofluid sample of an individual at a computing device 100, wherein the 1H NMR spectroscopy data 3100a-c indicate an NMR signal intensity as a function of chemical shift. Optionally, step S1 may comprise accessing the 1H-NMR spectroscopy data at and/or retrieving the 1H-NMR spectroscopy data 300a-c from a data storage 116 of the computing device 100 and/or from an external computing device 500. [0081] In a further step S2, the method comprises evaluating and/or analyzing the 1H-NMR spectroscopy data 300a-c with at least one trained machine learning model 114 of the computing device with respect to a molecular profile 310a-c of the biofluid sample, wherein the molecular profile 310a-c contains all hydrogen peaks in the 1H-NMR spectroscopy data 300a-c that can be assigned to one or more of a metabolite, a protein, an amino acid, a micromolecule, and a macromolecule contained in the biofluid sample. [0082] The method further comprises a step S3 of classifying the molecular profile 310a-c on the basis of the evaluation with the at least one trained machine learning model

114 in at least a first class and a second class of molecular profiles 310a-c,

15

20

25

30

35

19 wherein the first class is representative of molecular profiles 310a associated with healthy individuals, and the second class is representative of molecular profiles 310b, c associated with individuals having a cancerous and/or pre-cancerous condition. [0083] In a further step S4, the method comprises determining, based on the classification of the molecular profile 310a-c, a cancer risk score indicating a probability of cancer occurrence in the individual. Optionally, step S4 may comprise calculating a probability of cancer occurrence in the individual. Furthermore, the cancer risk score and/or the context information may optionally be output from the computing device 100, e.g., at the user interface 118. [0084] As mentioned above, each of the 1H NMR spectroscopy data 300a-c shown in Figures 3B-3D may be used as input to determine a corresponding cancer risk score for the corresponding individual and/or biofluid sample. Using the spectrum 300a of a healthy individual, the cancer risk score may indicate a high probability, greater than 50%, that the molecular profile 310a is associated with the first class of molecular profiles of a healthy or non-cancerous individual. On the other hand, using the 1H NMR spectroscopy data 300b, c of Figures 3C and 3D, the cancer risk score may indicate a high probability of over 50% that the molecular profile 310b, c is associated with the second class of molecular profiles of an individual with a cancer and/or pre-cancerous condition. Optionally, different classes for cancer and pre-cancerous conditions may be used, and the cancer risk score may indicate whether the individual has a cancer, such as pancreatic cancer (Figure 3D), or a pre-cancerous condition, such as pancreatitis (Figure 3C). For example, a cancer risk score above 0.5 may indicate a cancer, while a cancer risk score below 0.5 may indicate a healthy individual and/or an individual with a pre-cancerous condition. [0085] With the method described here, for example, stage I-IV pancreatic cancer can be differentiated from chronic pancreatitis in a binary classification with an accuracy of over 95% with 10-fold cross-validation. [0086] The method for determining the cancer risk score, related aspects, and advantages are summarized below. The method described here can accurately detect cancer, cancers, and/or precancerous lesions that are usually only diagnosed when it is too late and survival rates are low. The method involves a combination of high-frequency 1H NMR (nuclear magnetic resonance) spectroscopy and machine learning applied to a biofluid sample, e.g., a blood sample. As a result, a cancer risk score is generated for the various cancers and/or precancerous lesions, enabling earlier detection compared to other known methods. In contrast to currently used or known techniques, the method described here can not only detect cancer-related signals or molecular signatures, but also

identified molecular signatures provide information about the individual's health status

20 / 44

15

20

25

30

35

20. Furthermore, the method described here is fast, scalable, and cost-effective, making it very suitable for industrial use. [0087] Cancer cells can influence the overall molecular composition of biofluids, such as blood, due to their altered metabolism and protein composition. This influence can include changes in the proportion of specific proteins, metabolites, micro- and/or macromolecules that are either released directly by the cancer cells or whose release is directly or indirectly induced by them, e.g., via biological networks, as described, for example, with reference to Figure 2. To detect this change, the high-frequency 1H NMR spectroscopy and/or the 1H NMR spectroscopy data 200, 300a-c are used, which can provide information about the chemical environment of each hydrogen atom in the biofluid sample and thus indirectly provide information about each individual molecule and its concentration. Depending on the structure of the molecule containing the corresponding hydrogen atom, a different chemical environment may be present, changing the chemical shift of that specific hydrogen atom. For these reasons, the 1H NMR spectroscopy data 200, 300a-c of the biofluid sample can provide an extremely high information density that can be used to distinguish between healthy individuals and individuals with a cancerous and/or pre-cancerous disease. [0088] Currently, analysis methods are limited to signals that can be assigned to a specific molecule, e.g., a metabolite or a lipoprotein. The great advantage of the method described here may be that all hydrogen peaks in the 1H NMR spectrum are taken into account, thus allowing the entire molecular profile 310a-c of the individual's biofluid sample to be considered. Even in early stages, cancer cells can influence the metabolism and molecular profile, which can be reliably detected using the method of the present disclosure. [0089] Early cancer detection is one of the greatest challenges in oncology and is therefore a dynamic, rapidly growing field. The method described here can be used to screen high-risk patients or individuals with predispositions to cancer characteristics. Therefore, patients or individuals can provide a biofluid sample, e.g., a blood sample, to a physician, healthcare provider, or hospital, which can then be analyzed using the method described here. [0090] For NMR analysis, the biofluid samples can be analyzed using a high-frequency NMR spectrometer with a pulse sequence similar to the CPMGPR1D. The free induction decay (FID) can be converted into a spectrum by Fourier transformation, normalized to a reference peak (lactate, anomeric D-glucose doublet, etc.), baseline corrected, and the spectrum converted into binned increments of less than or equal to 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm, as shown in Figures 3B to 3D. This allows a considerable amount of data to be obtained for the decoding of the

molecular composition or profile 310a-c of the biofluid sample and can

15

20

25

30

35

21 provide information about all hydrogen atoms present, where each bin can represent a specific position in the spectrum. [0091] While most other approaches to cancer or multicancer detection use liquid biopsy in combination with next-generation sequencing technologies, the method according to the present disclosure focuses not on nucleotide sequences, but on the molecular profile 310a-c of the individual's biofluid sample using 1H NMR spectroscopy, machine learning, and optionally biological reasoning. In particular, hydrogen peaks that cannot be attributed to specific metabolites, proteins, amino acids, micro- and/or macromolecules, but rather to an overlap of multiple molecules, proteins, nucleic acids, and/or other components or constituents of the biofluid sample, can be used to determine the cancer risk score. For this reason, the method according to the present disclosure can provide improved sensitivity and specificity of over 95% (in both preliminary and pilot study data), compared to approximately 60-80% for other approaches for detecting multiple cancers or tumor markers with even lower accuracy. [0092] The method according to the present disclosure can enable the discrimination of healthy and diseased patients or individuals with a cancer and/or pre-cancer disease. To train the at least one ML model 114, 1-HNMR spectroscopy data 200, 300a-c of a training dataset of preferably a plurality of individuals can be normalized. Subsequently, feature selection and engineering for features of the at least one ML model 114 can be performed. For example, one or more ranges in and/or bins of the chemical shift may be selected based on the normalized 1H NMR spectroscopy data 300a-c as features of the ML model 114 that provide a difference between healthy and cancer/pre-cancer samples and thus may represent the molecules, proteins, amino acids, micro- and/or macromolecules that reflect or indicate changes between healthy or non-cancerous individuals and cancer/pre-cancerous individuals. [0093] Optionally, methods such as principal components analysis (PCA) may be used to create additional constructed and/or extracted features for the at least one ML model 114. The identified features may include information from metabolites, amino acids, proteins, micro- and/or macromolecules. After feature selection, for example, by a p-value-based filter, the at least one ML model 114 may be trained using the selected, constructed and/or extracted features. [0094] A machine learning approach is used to train the at least one ML model 114 that can predict the molecular profile 310a-c of the healthy/non-cancerous and the diseased/cancerous/pre-cancerous patients or individuals. The use of the molecular profile 310a-c can enable the detection of cancer signals or signatures from the individual's biofluid sample. The at least one ML model 114

can be validated by cross-validation and optionally on a separate test data set of 1H-NMR

15

20

25

30

35

22 Spectroscopy data 200, 300a-c of one or more further individuals may be tested, which, for example, differ from the individuals considered in the training dataset. [0095] Accordingly, the capability of high-frequency NMR spectroscopy may be combined with a well-designed machine learning and feature identification approach to detect highly complex correlations in the 1H NMR spectroscopy data 200, 300a-c that may be characteristic of the individual's biofluid sample and/or the corresponding molecular profile 310a-c, which may enable the earliest possible detection of a cancerous and/or pre-cancerous disease. [0096] For example, by outputting and/or calculating one or more classification results and/or probabilities for the respective classes, a cancer risk score may be determined, which may provide information about the extent to which the molecular profile 310a-c resembles a particular health condition. Based on the cancer risk score, information can be provided about how likely the predicted health condition is. [0097] Optionally, the at least one ML model 114 can be adapted by adjusting the cancer risk score cut-off values for patient or individual groups and desired false positive and/or false negative rates, for example, based on an area-under-the-curve AUC receiver operator curve (AUC-ROC). [0098] Figure 4 shows a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to an exemplary embodiment. The method of Figure 4 can be performed by the computing device 100 described with reference to Figure 1. In particular, Figure 4 illustrates the training of the ML model 114 of the computing device 100. [0099] In a first step 400, raw 1H NMR spectroscopy data is acquired at an NMR frequency of about 500 MHz or above, e.g., with a pulse sequence similar to CPMG and/or CPMGPR1D. [0100] In step 402, the 1H NMR spectroscopy data may be aligned and/or spectral processing may be performed, such as one or more chemical shift referencing, phase adjustment or correction, and baseline correction of the raw 1H NMR spectroscopy data. [0101] In a further step 404, the 1H NMR spectroscopy raw data may be divided into incremental chemical shift bins, each bin having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. [0102] In step 406, the binned 1H NMR spectroscopy data may be normalized to a median of the spectroscopy data. [0103] In step 408, labels may be assigned to the binned 1H NMR spectroscopy data

be assigned, e.g. 0 and 1 for a binary classification.

23 / 44

15

20

25

30

35

23 [0104] In step 410, feature selection, identification, and/or extraction for features of the ML model 114 may be performed as described above. For example, KBest may be used for feature selection. [0105] In optional step 412, bin features may be discarded based on the ppm range, unsupervised clustering of features may be performed, undersampling and/or oversampling may be performed. [0106] In step 414, scaling may be performed, e.g., with StandardScaler. [0107] In step 416, the ML model 114 may be trained for binary or multi-class classification using a statistical machine learning algorithm, such as VotingClassifier, ensemble methods, logistic regression, or others. [0108] In step 418, the trained ML model can be used in the inference to generate a cancer risk score, which, for example, indicates a probability between 0 and 1 for a binary classification. The cancer risk score can act as a computer-implemented biomarker that informs about pathogenic molecular signatures in the biofluid sample of an individual. [0109] Figure 5 shows a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to an exemplary embodiment. The method of Figure 5 can be performed by the computing device 100 described with reference to Figure 1. In particular, Figure 5 illustrates the inference of the ML model 114 of the computing device 100. [0110] In a first step 500, 1H NMR spectroscopy raw data is acquired at an NMR frequency of about 500 MHz or above, e.g. B. with a pulse sequence similar to CPMG and/or CPMGPR1D. [0111] In step 502, the 1H NMR spectroscopy data may be aligned and/or spectral processing may be performed, such as one or more chemical shift referencing, phase adjustment or correction, and baseline correction of the 1H NMR spectroscopy raw data. [0112] In a further step 504, the 1H NMR spectroscopy raw data may be binned into incremental chemical shift bins, each bin having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example, about 0.00016 ppm. [0113] In step 506, the binned 1H-NMR spectroscopy data can be normalized to a median of the spectroscopy data. [0114] In step 508, scaling can be performed, e.g., with StandardScaler. [0115] In step 510, the trained ML model 114, which was trained, e.g., according to the method of Figure 4, can be used to generate a cancer risk score, which, e.g., indicates a probability between 0 and 1 for a binary classification. The cancer risk can serve as a computer-implemented biomarker that can be used to predict pathogenic molecular

signatures in an individual’s biofluid sample. For example, one of the

15

20

25

30

35

24 The 1H NMR spectroscopy data 310a-c shown in Figures 3B to 3D can be used as input in step 510 to calculate or determine the corresponding cancer risk score. [0116] Figure 6 shows a flowchart illustrating the steps of a computer-implemented method for determining a cancer risk score according to an exemplary embodiment. The method of Figure 6 can be performed by the computing device 100 described in Figure 1. Furthermore, the method shown in Figure 6 can be used to supplement step S3 of the method shown in Figure 3. Accordingly, the method of Figure 6 can comprise steps S1 to S4 of the method of Figure 3, wherein step S3 can be supplemented by the detailed steps shown and described in Figure 6. [0117] In step 600 of the method of Figure 6, step S3 of classifying the molecular profile 310a-c into at least the first class and the second class of molecular profiles 310a-c comprises classifying the molecular profile 310a-c into at least one healthy class of molecular profiles 310a associated with healthy individuals and one unhealthy or diseased class of molecular profiles 310b,c associated with unhealthy or diseased individuals based on evaluating the 1H NMR spectroscopy data 300a-c with respect to the molecular profile 310a-c of the biofluid sample with a first trained machine learning model. For clarity and simplicity, Figure 6 depicts a classification into the healthy class with reference number 602 and a classification into the unhealthy or diseased class with reference number 604. [0118] Furthermore, the method comprises, in step 606, after determining in step 604 that the molecular profile 310b, c of the biofluid sample is associated with or is in the non-healthy or diseased class of molecular profiles 310b, c, classifying, based on the evaluation of the 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a second trained machine learning model, the molecular profile into a non-cancer class 310a of molecular profiles associated with non-cancerous individuals and a cancer class of molecular profiles 310b, c associated with cancerous and/or pre-cancerous individuals. For clarity and simplicity, a classification into the non-cancer class is depicted in Figure 6 with reference number 608 and a classification into the cancer class with reference number 610. [0119] Optionally, in steps 600-604, a first cancer risk score can be determined, which indicates a probability that the molecular profile 310a is associated with the healthy class and/or the non-healthy or diseased class or lies in this class. Furthermore, optionally, in steps 606-610, a second cancer risk score can be determined, which indicates a probability that the molecular profile 310b, c is associated with the non-cancer class and/or the cancer class of the molecular profiles 310b, c or lies in this class. Furthermore, optionally, in steps 606-610, the cancer risk score or a final cancer risk score can be determined, e.g., on the basis of the first and second cancer risk scores.

Scores.

15

20

25

30

35

25 [0120] Furthermore, in step 612, after determining that the molecular profile of the biofluid sample is associated with or is in the cancerous class of molecular profiles, the method comprises classifying the molecular profile into a plurality of classes of molecular profiles based on the evaluating 1H NMR spectroscopy data related to the molecular profile of the biofluid sample with a third trained machine learning model, wherein each class is associated with a particular type of cancer and/or pre-cancerous condition, as indicated by reference numerals 614a, 614b, 614c in Figure 6. [0121] Optionally, a third cancer risk score may be determined that indicates the probability that the molecular profile is associated with a particular type of cancer, such as pancreatic cancer, and/or a pre-cancerous condition, such as pancreatitis. Accordingly, a type of cancer, a cancer disorder, and/or a precancerous condition can be determined using the method described herein. [0122] Although the invention is shown and described in detail in the drawings and the above description, these illustrations and descriptions are to be considered as illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. The invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments can be understood and practiced by those skilled in the art practicing the claimed invention, based on the drawings, the disclosure, and the claims. [0123] As used herein, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" does not exclude a plurality. The mere fact that certain measures are recited in dependent claims does not mean that a combination of these measures cannot be advantageous. Any reference signs in the claims should not be construed as limiting the scope of application.

[0124] As used herein, the phrase "indicative of" may mean "reflective" and/or "comprising." Accordingly, a unit/element referred to herein as "indicative of [...]" may be used synonymously or interchangeably with the unit/element "comprising [...]" or the unit/element "reflective [...]." [0125] Furthermore, the terms "first," "second," "third," or "a)," "b)," "c)," and the like are used in the description and claims to distinguish similar elements and not necessarily to describe a sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances, and that the embodiments of the invention described herein may be operated in orders other than those described or depicted herein. [0126] Within the scope of the present invention, each specified numerical value is typically associated with an accuracy interval that the person skilled in the art understands to be such that the technical effect of the feature in question is still ensured. Within the scope of the present invention, the deviation from the specified numerical value is in the range of + 10%, preferably +

5%. The above-mentioned deviation from the specified numerical interval of + 10%, preferably

26 / 44

+ 5%, is also indicated by the terms "approximately" and "approximately" used here in relation to a numerical value.

expressed as "approximately".

15

20

25

30

35

27 CLAIMS

1. A computer-implemented method for determining a cancer risk score for an individual, the method comprising:

Obtaining, at a computing device (100), hydrogen 1-nuclear magnetic resonance, 1HNMR spectroscopy data (300a-c) for a biofluid sample of an individual, wherein the 1H-NMR spectroscopy data (300a-c) is indicative of an NMR signal intensity as a function of chemical shift;

Evaluating, with at least one trained machine learning model (114) of the computing device (100), the 1H NMR spectroscopy data (300a-c) with respect to a molecular profile (310a-c) of the biofluid sample, wherein the molecular profile includes all hydrogen peaks in the 1H NMR spectroscopy data that can be assigned to one or more of a metabolite, a protein, an amino acid, a micromolecule, and a macromolecule contained in the biofluid sample;

Classifying, based on the evaluation with the at least one trained machine learning model (114), the molecular profile (310a-c) into at least a first class and a second class of molecular profiles, wherein the first class is representative of molecular profiles (310a) associated with healthy individuals and the second class is representative of molecular profiles (310b, c) associated with individuals having a cancerous and/or pre-cancerous disease; and

Determine, based on classifying the molecular profile, a cancer risk

Scores that indicate a probability of cancer occurring in the individual.

2. The method according to the preceding claim, wherein determining the cancer risk score comprises calculating and/or determining the probability of cancer occurrence in the individual based on the at least one trained machine learning model (114) of the computer device (100).

3. The method according to any one of the preceding claims, wherein classifying the molecular profile (310a-c) comprises determining and/or calculating a classification result indicative of a probability for the first class and/or the second class; and

where the cancer risk score is determined based on the classification result.

4. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained to determine a probability for a binary classification and/or multi-class classification of the molecular profile (310a-c) of the

biofluid sample.

28 / 44

15

20

25

30

35

28 5. The method according to any one of the preceding claims, wherein the biofluid sample is selected from the group consisting of a blood serum sample, a blood plasma sample, a blood sample, a

urine sample and a cerebrospinal fluid sample.

6. The method according to any one of the preceding claims, wherein the 1H-NMR spectroscopy data (300a-c) are indicative of an NMR signal intensity in binned increments of chemical shift, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or

equal to about 0.006 ppm, for example about 0.00016 ppm.

7. A method according to any one of the preceding claims, further comprising:

Converting the 1H NMR spectroscopy data (300a-c) into a binned data structure based on a map of NMR signal intensity to binned increments of chemical shift, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example

about 0.00016 ppm.

8. The method of any one of the preceding claims, further comprising: combining the 1H NMR spectroscopy data (300a-c) into a plurality of incremental chemical shift bins, each bin having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example about 0.00016 ppm; and optionally normalizing the binned 1H NMR spectroscopy data (300a-c) to a median of at least one

Subgroup of bins.

9. The method according to any one of the preceding claims, wherein the at least one trained machine learning model (114) is a machine-learned classifier configured to receive as input data the 1H-NMR spectroscopy data (300a-c) in a data structure of binned

to process chemical shift increments.

10. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained and/or configured to identify at least one molecular signature of a cancerous and/or pre-cancerous disease in the molecular profile (310a-

c) to identify the biofluid sample.

11. Method according to one of the preceding claims, wherein the determined cancer risk score can be used as a computational biomarker which represents a pathogenic molecular signature and/or a molecular signature of a cancerous and/or pre-cancerous disease in the

molecular profile (310a-c) of the biofluid sample.

29 / 44

15

20

25

30

35

29

12. The method according to any one of claims 10 and 11, wherein identifying the at least one molecular signature includes identifying one or more bins in the 1H-NMR spectroscopy data (300a-c) associated with and/or containing one or more hydrogen peaks that indicate a cancer-related alteration of the molecular profile (310b, c) with respect to molecular profiles (310a) associated with healthy individuals,

point out.

13. The method of any one of claims 10 to 12, wherein identifying the at least one molecular signature includes identifying one or more bins and/or a chemical shift range in the 1H NMR spectroscopy data (300a-c) associated with an overlap of hydrogen peaks attributable to a plurality of metabolites, proteins, amino acids, micromolecules, and macromolecules contained in the biofluid sample; and/or

wherein identifying the at least one molecular signature comprises identifying one or more bins and/or a range of chemical shift in the 1H-NMR spectroscopy data (300a-c) that cannot be assigned to individual metabolites, proteins, amino acids, micromolecules and macromolecules contained in the biofluid sample

are.

14. The method according to any one of claims 10 to 13, wherein the at least one molecular signature is associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) associated with one or more metabolites, proteins, amino acids, micromolecules and macromolecules contained in the biofluid sample with respect to a biofluid sample of one or more healthy

individuals.

15. The method according to any one of the preceding claims, wherein evaluating the 1H-NMR spectroscopy data (300a-c) with respect to the molecular profile (310a-c) comprises analyzing all hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) associated with one or more metabolites, proteins, amino acids, micromolecules, and macromolecules

contained in the biofluid sample.

16. The method according to any one of the preceding claims, wherein evaluating the 1H-NMR spectroscopy data (300a-c) with respect to the molecular profile (310a-c) comprises detecting a pattern of hydrogen peaks associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) with respect to a biofluid sample of one or more healthy individuals; and/or

wherein the at least one machine learning model is trained to generate a pattern of

15

20

25

30

35

30 hydrogen peaks, which is associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H NMR spectroscopy data (300a-c) with respect to a

Biofluid sample of one or more healthy individuals.

17. The method according to any one of the preceding claims, wherein obtaining the 1H-NMR spectroscopy data (300a-c) comprises acquiring raw 1H-NMR spectroscopy data with an NMR

spectrometer at a frequency above about 500 MHz.

18. The method according to the preceding claim, further comprising spectral processing of the 1H NMR spectroscopy raw data and/or peak alignment of the 1H NMR spectroscopy raw data.

19. The method according to the preceding claim, wherein the spectral processing comprises one or more of a chemical shift referencing, a phase adjustment and a

Baseline correction of the 1H NMR spectroscopy raw data is included.

20. The method of any preceding claim, further comprising one or more of normalizing, scaling, binning, and filtering the 1H NMR spectroscopy raw data and/or the 1H NMR spectroscopy data.

21. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained for classification using one or more statistical machine learning algorithms, which are preferably based on one or more

based on a voting classifier, an ensemble procedure and logistic regression.

22. The method according to any one of the preceding claims, wherein classifying the molecular profile (310a-c) into at least the first class and the second class of molecular profiles comprises: f

Classifying, based on the evaluation of the 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a first trained machine learning model, the molecular profile into at least one healthy class of molecular profiles (310a) associated with healthy individuals and a non-healthy class of molecular profiles (310b, c) associated with non-healthy individuals; and

after determining that the molecular profile of the biofluid sample is associated with or in the unhealthy class of molecular profiles (310b, c), classifying, based on the evaluation of the 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a second trained machine learning model, the molecular profile into a non-cancerous class of molecular profiles (310a) associated with non-cancerous individuals

and a cancer class of molecular profiles associated with cancerous and/or

15

20

25

30

35

31 pre-cancerous individuals (310b, c); where the first trained machine learning model and the second machine learning model

Learning models differ from each other.

23. The method according to the preceding claim, further comprising:

Determining, based on the classification of the molecular profile (310a-c) of the biofluid sample with the first trained machine learning model, a first cancer risk score indicating a probability that the molecular profile (310a-c) is associated with or lies in the healthy class and/or the non-healthy class;

Determining, based on the classification of the molecular profile of the biofluid sample with the second trained machine learning model, a second cancer risk score indicating a probability that the molecular profile (310a-c) is associated with or located within the non-cancer class and/or the cancer class of molecular profiles; and

Determination of the cancer risk score based on the first and second cancer risk

Scores.

24. The method according to any one of claims 22 and 23, further comprising:

after determining that the molecular profile of the biofluid sample is associated with or is within the cancer class of molecular profiles, classifying the molecular profile (310a-c) into a plurality of classes of molecular profiles based on the evaluated 1HNMR spectroscopy data relating to the molecular profile of the biofluid sample with a third trained machine learning model, each class being associated with a particular type of cancerous and/or pre-cancerous condition; and

optionally determining a third cancer risk score that indicates a probability that the molecular profile is associated with a specific type of cancerous and/or

precancerous disease. 25. A computer program which, when executed by a computer device (100), instructs the computer device (100) to carry out the method according to any one of the preceding claims

to carry out.

26. Non-transitory computer-readable medium on which a computer program is embodied in accordance with

previous claim is stored.

27. Computer device (100) adapted to carry out the method according to one of the claims

1 to 24 is set up.

Claims (27)

CHANGED CLAIMS 1. A computer-implemented method for determining a cancer risk score for an individual, the method comprising: Obtaining, at a computing device (100), hydrogen 1-nuclear magnetic resonance, 1HNMR spectroscopy data (300a-c) for a biofluid sample of an individual, wherein the 1H-NMR spectroscopy data (300a-c) is indicative of an NMR signal intensity as a function of chemical shift; Evaluating, with at least one trained machine learning model (114) of the computing device (100), the 1H NMR spectroscopy data (300a-c) with respect to a molecular profile (310a-c) of the biofluid sample, wherein the molecular profile includes all hydrogen peaks in the 1H NMR spectroscopy data that can be assigned to one or more of a metabolite, a protein, an amino acid, a micromolecule, and a macromolecule contained in the biofluid sample; Classifying, based on the evaluation with the at least one trained machine learning model (114), the molecular profile (310a-c) into at least a first class and a second class of molecular profiles, wherein the first class is representative of molecular profiles (310a) associated with healthy individuals and the second class is representative of molecular profiles (310b, c) associated with individuals having a cancerous and/or pre-cancerous disease; and Determine, based on classifying the molecular profile, a cancer risk Scores that indicate a probability of cancer occurring in the individual. 2. The method according to the preceding claim, wherein determining the cancer risk score comprises calculating and/or determining the probability of cancer occurrence in the individual on the basis of the at least one trained machine learning model (114) the computer device (100). 3. The method according to any one of the preceding claims, wherein classifying the molecular profile (310a-c) comprises determining and/or calculating a classification result indicative of a probability for the first class and/or the second class; and where the cancer risk score is determined based on the classification result. 4. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained to determine a probability for a binary classification and/or multi-class classification of the molecular profile (310a-c) of the biofluid sample, wherein for training the at least one ML model (114) 1- 4 MOST RECENTLY SUBMITTED CLAIMS HNMR spectroscopy data (200, 300a-c) of a training data set from a variety of individuals are normalized. 5. The method according to any one of the preceding claims, wherein the biofluid sample is selected from the group consisting of a blood serum sample, a blood plasma sample, a blood sample, a urine sample and a cerebrospinal fluid sample. 6. The method according to any one of the preceding claims, wherein the 1H-NMR spectroscopy data (300a-c) are indicative of an NMR signal intensity in binned increments of chemical shift, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example about 0.00016 ppm. 7. A method according to any one of the preceding claims, further comprising: Converting the 1H NMR spectroscopy data (300a-c) into a binned data structure based on a map of NMR signal intensity to binned increments of chemical shift, each increment having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example about 0.00016 ppm. 8. The method of any preceding claim, further comprising: combining the 1H NMR spectroscopy data (300a-c) into a plurality of incremental chemical shift bins, each bin having a width of less than or equal to about 0.02 ppm, preferably less than or equal to about 0.01, more preferably less than or equal to about 0.006 ppm, for example about 0.00016 ppm; and normalizing the binned 1H NMR spectroscopy data (300a-c) to a median of at least one Subgroup of bins. 9. The method according to claim 7, wherein the at least one trained machine learning model (114) is a machine-learned classifier configured to receive as input data the 1HNMR spectroscopy data (300a-c) in a data structure of binned increments of the chemical shift. 10. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained and/or configured to identify at least one molecular signature of a cancerous and/or pre-cancerous disease in the molecular profile (310a-c) of the biofluid sample. 11. The method according to claim 10, wherein the determined cancer risk score is used as a calculated 2 40 / 44 MOST RECENTLY SUBMITTED CLAIMS Biomarker can be used that contains a pathogenic molecular signature and/or a molecular signature of a cancerous and/or pre-cancerous disease in the molecular profile (310a-c) of the biofluid sample. 12. The method according to any one of claims 10 and 11, wherein identifying the at least one molecular signature includes identifying one or more bins in the 1H-NMR spectroscopy data (300a-c) associated with and/or containing one or more hydrogen peaks that indicate a cancer-related alteration of the molecular profile (310b, c) with respect to molecular profiles (310a) associated with healthy individuals associated with. 13. The method of any one of claims 10 to 12, wherein identifying the at least one molecular signature includes identifying one or more bins and/or a chemical shift range in the 1H NMR spectroscopy data (300a-c) associated with an overlap of hydrogen peaks attributable to a plurality of metabolites, proteins, amino acids, micromolecules, and macromolecules contained in the biofluid sample; and/or wherein identifying the at least one molecular signature comprises identifying one or more bins and/or a range of chemical shift in the 1HNMR spectroscopy data (300a-c) that cannot be assigned to individual metabolites, proteins, amino acids, micromolecules and macromolecules present in the Biofluid sample. 14. The method according to any one of claims 10 to 13, wherein the at least one molecular signature is associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) associated with one or more metabolites, proteins, amino acids, micromolecules and macromolecules contained in the biofluid sample with respect to a biofluid sample of one or more several healthy individuals. 15. The method according to any one of the preceding claims, wherein evaluating the 1HNMR spectroscopy data (300a-c) with respect to the molecular profile (310a-c) comprises analyzing all hydrogen peaks in the 1H NMR spectroscopy data (300a-c) associated with one or more metabolites, proteins, amino acids, micromolecules, and macromolecules contained in the biofluid sample. 16. Method according to one of the preceding claims, wherein the evaluation of the 1H- NMR spectroscopy data (300a-c) related to the molecular profile (310a-c) detecting a pattern of hydrogen peaks associated with a cancer-related change in 3 MOST RECENTLY SUBMITTED CLAIMS proportion of one or more hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) with respect to a biofluid sample of one or more healthy individuals; and/or wherein the at least one machine learning model is trained to recognize a pattern of hydrogen peaks that is associated with a cancer-related change in the proportion of one or more hydrogen peaks in the 1H-NMR spectroscopy data (300a-c) with respect to a Biofluid sample of one or more healthy individuals. 17. The method according to any one of the preceding claims, wherein obtaining the 1H-NMR spectroscopy data (300a-c) comprises acquiring raw 1H-NMR spectroscopy data with a NMR spectrometer at a frequency above about 500 MHz. 18. The method according to the preceding claim, further comprising spectral processing of the 1H NMR spectroscopy raw data and/or peak alignment of the 1H NMR spectroscopy raw data. 19. The method according to the preceding claim, wherein the spectral processing comprises one or more of a chemical shift referencing, a phase adjustment and a baseline correction of the 1H NMR spectroscopy raw data. 20. The method of claim 17, further comprising one or more of normalizing, scaling, binning and filtering the 1H NMR spectroscopy raw data and/or the 1H NMR spectroscopy data. 21. The method according to any one of the preceding claims, wherein the at least one machine learning model (114) is trained for classification using one or more statistical machine learning algorithms, which are preferably based on one or more of a voting classifier, an ensemble method and logistic based on regression. 22. The method according to any one of the preceding claims, wherein classifying the molecular profile (310a-c) into at least the first class and the second class of molecular profiles comprises: Classifying, based on the evaluation of the 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a first trained machine learning model, the molecular profile into at least one healthy class of molecular profiles (310a) associated with healthy individuals and a non-healthy class of molecular profiles (310b, c) associated with non-healthy individuals; and after determining that the molecular profile of the biofluid sample is consistent with or in the unhealthy class of molecular profiles (310b, c), classifying based on 4 42 / 44 MOST RECENTLY SUBMITTED CLAIMS evaluating the 1H NMR spectroscopy data with respect to the molecular profile of the biofluid sample with a second trained machine learning model, classifying the molecular profile into a non-cancer class of molecular profiles (310a) associated with non-cancerous individuals and a cancer class of molecular profiles associated with cancerous and/or pre-cancerous individuals (310b, c); where the first trained machine learning model and the second machine Learning models differ from each other. 23. The method according to the preceding claim, further comprising: Determining, based on the classification of the molecular profile (310a-c) of the biofluid sample with the first trained machine learning model, a first cancer risk score indicating a probability that the molecular profile (310a-c) is associated with or lies in the healthy class and/or the non-healthy class; Determining, based on the classification of the molecular profile of the biofluid sample with the second trained machine learning model, a second cancer risk score indicating a probability that the molecular profile (310a-c) is associated with or located within the non-cancer class and/or the cancer class of molecular profiles; and Determination of the cancer risk score based on the first and second cancer risk Scores. 24. The method according to any one of claims 22 and 23, further comprising: after determining that the molecular profile of the biofluid sample is associated with or is within the cancer class of molecular profiles, classifying the molecular profile (310a-c) into a plurality of classes of molecular profiles based on the evaluated 1H NMR spectroscopy data relating to the molecular profile of the biofluid sample with a third trained machine learning model, each class being associated with a particular type of cancerous and/or pre-cancerous condition; and Determining a third cancer risk score that indicates a probability that the molecular profile is associated with a specific type of cancerous and/or precancerous disease is associated. 25. A computer program which, when executed by a computer device (100), instructs the computer device (100) to carry out the method according to any one of the preceding to carry out claims. 26. Non-transitory computer-readable medium on which a computer program is embodied in accordance with the previous claim is stored. 5 43 / 44 MOST RECENTLY SUBMITTED CLAIMS 27. Computer device (100) which is used to carry out the method according to one of the Claims 1 to 24 are set up. 6 44 / 44 MOST RECENTLY SUBMITTED CLAIMS