Methods and systems for determining responders to treatment

Abstract

To provide methods, systems and apparatuses for classifying a patient as a responder or a non-responder.SOLUTION: A method comprises: determining first gene data associated with a plurality of genes; determining second gene data associated with the genes, where the genes are sequenced from a plurality of tumor samples, where each of the tumor samples is labeled as a responder or a non-responder; determining, on the basis of the first gene data and the second gene data, a plurality of features for a predictive model; training, on the basis of a first portion of the second gene data, the predictive model according to the features; testing, on the basis of a second portion of the second gene data, the predictive model; and outputting, on the basis of the testing, the predictive model.SELECTED DRAWING: Figure 1

Description

Background Art

[0001] Cross - reference to Related Applications This application claims the benefit of U.S. Provisional Patent Application No. 62 / 990,814, filed Mar. 17, 2020, the entire contents of which are incorporated herein by reference.

[0002] One of the biggest problems facing the use of machine learning is the lack of availability of large annotated datasets. Data annotation is not only expensive and time - consuming but also highly dependent on the availability of expert observers. When the amount of training data is limited, the performance of supervised machine - learning algorithms, which often require very large amounts of data to train to avoid overfitting, can be inhibited. So far, much effort has been directed at extracting as much information as possible from the available data. In particular, one area where large annotated datasets are lacking is the analysis of biological data, such as gene expression data (e.g., clinical data). The ability to analyze gene data (e.g., gene expression data) and predict a patient's response to treatment is important for patient care. However, in many cases, insufficient data is available to train machine - learning algorithms to accurately predict a patient's response.

[0003] Thus, there is a need for improved systems and methods for determining and utilizing relevant datasets for use in machine - learning applications. Accordingly, an object of the present invention is to provide a computer - implemented system and method having an improved ability to determine and utilize gene datasets for training machine - learning applications for making predictions, including predicting a patient's response to treatment.

Summary of the Invention

[0004] It should be understood that both the following general description and the following detailed description are exemplary and explanatory only and not restrictive.

[0005] In one embodiment, the disclosed method includes determining first gene data relating to a plurality of genes, determining second gene data relating to a plurality of genes, wherein the plurality of genes are sequenced from a plurality of tumor samples, and each tumor sample of the plurality of tumor samples is labeled as a responder or a non-responder, determining a plurality of features for a predictive model based on the first and second gene data, training the predictive model according to the plurality of features based on a first portion of the second gene data, testing the predictive model based on a second portion of the second gene data, and outputting the predictive model based on this test.

[0006] In one embodiment, the disclosed method includes receiving baseline genetic data relating to multiple genes of a subject, wherein the multiple genes are sequenced from the tumor of the subject; providing the baseline genetic data to a predictive model; and determining, based on the predictive model, that the subject is a candidate for a therapeutic treatment.

[0007] In one embodiment, the disclosed method includes determining baseline gene expression data associated with a plurality of genes, wherein the plurality of genes are associated with a plurality of tumor samples, and each tumor sample of the plurality of tumor samples is labeled as a responder or a non-responder; determining transcription factor gene data based on the plurality of genes; generating a transcription factor (TR) network based on the transcription factor gene data and the plurality of genes; determining an enrichment score associated with each transcription factor gene in the set of transcription factor genes based on the TR network and baseline gene expression data; and determining one or more predicted transcription factor genes in the set of transcription factor genes based on the enrichment score.

[0008] Additional advantages may be described in the following description or may be known through practice. These advantages will be realized and achieved by the elements and combinations specifically indicated in the attached claims. [Brief explanation of the drawing]

[0009] The accompanying drawings incorporated herein and forming part of this description serve to illustrate the principles of the methods and systems described herein: [Figure 1] Figure 1 shows an exemplary method. [Figure 2] Figure 2 shows an exemplary machine learning system. [Figure 3] Figure 3 shows an exemplary machine learning method. [Figure 4] Figure 4 shows an exemplary timeline for obtaining baseline and treatment-dependent gene expression data. [Figure 5] Figure 5 shows standardized immunomarker gene expression. [Figure 6A] Figure 6A shows differentially expressed genes, determined by comparing baseline gene expression data with gene expression during treatment for all patients (responders and non-responders). [Figure 6B] Figure 6B shows differentially expressed genes, determined by comparing baseline gene expression data with treatment-dependent gene expression in pairs for responders only. [Figure 6C] Figure 6C shows differentially expressed genes, determined by comparing baseline gene expression data with treatment-dependent gene expression in pairs for non-responders only. [Figure 7] Figure 7 shows that the heatmap on the right displays the top 50 differentially expressed genes with duplicate expression from responder pairs only. [Figure 8] Figure 8 shows the differentially expressed genes between baseline responders and baseline non-responders. [Figure 9] Figure 9 shows selected gene set data that are unrelated to disease. [Figure 10] Figure 10 shows the predicted genes identified using only selected, disease-independent gene sets. [Figure 11] Figure 11 shows an exemplary, high-performing gene signature. [Figure 12] Figure 12 shows the performance of an exemplary, highest-performing gene signature. [Figure 13] Figure 13 illustrates an exemplary systems biology method for identifying predictive transcription factor genes. [Figure 14] Figure 14 shows exemplary predictive transcription factor gene data identified using systems biology methods. [Figure 15] Figure 15 shows a block diagram of an exemplary computing device. [Figure 16] Figure 16 shows an exemplary method. [Figure 17] Figure 17 shows an exemplary method. [Figure 18] Figure 18 shows an exemplary method. [Modes for carrying out the invention]

[0010] Where used herein and in the appended claims, the singular forms “a,” “an,” and “the” refer to multiple references unless it is evident from the context that they should be interpreted in a different sense. Herein, ranges may be expressed as “about” one particular value and / or “about” another particular value. Where such ranges are expressed, another configuration includes one particular value and / or another particular value. Similarly, where values ​​are expressed as approximations, the use of the antecedent “about” will be understood to mean that a particular value forms another configuration. It will be further understood that each endpoint of these ranges is significant in relation to and independently of the other endpoints.

[0011] "Any" or "optionally" means that the event or situation described thereafter may or may not occur, and that this description includes both the case where the event or situation occurs and the case where it does not occur.

[0012] Throughout this specification and the claims, the word "comprise" and variations of this word, such as "comprising" and "comprises", mean "including but not limited to", and are not intended to exclude, for example, other components, integers, or steps. "Exemplary" means "an example of", and is not intended to convey a preferred or ideal configuration. "Such as" is not used in a limiting sense, but for illustrative purposes.

[0013] When combinations of components, subsets, interactions, groups, etc. are disclosed, specific references to each of the various individual and collective combinations and permutations of these components may not be explicitly described, but each is understood to be specifically contemplated and described herein. This applies to all parts of this application, including but not limited to the steps in the described method. Thus, if there are various additional steps that can be performed, it is understood that each of these additional steps may be performed with any particular configuration or combination of configurations of the described method.

[0014] As will be understood by those skilled in the art, hardware, software, or a combination of software and hardware can be implemented. Further, a computer program product on a computer-readable storage medium (e.g., non-transitory) has processor-executable instructions (e.g., computer software) embodied therein. Any suitable computer-readable storage medium can be utilized, including a hard disk, CD-ROM, optical storage device, magnetic storage device, memresistor, non-volatile random access memory (NVRAM), flash memory, or a combination thereof.

[0015] Throughout this application, reference is made to block diagrams and flowcharts. It will be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, can be implemented by processor-executable instructions. These processor-executable instructions may be loaded onto a general purpose computer, a special purpose computer, or other programmable data processing apparatus, thereby creating a device for implementing the functions specified in the blocks of the flowchart by processor-executable instructions that execute on the computer or other programmable data processing apparatus.

[0016] These processor-executable instructions may also be stored in a computer-readable memory that can instruct a computer or other programmable data processing apparatus to function in a particular manner, thereby generating a manufactured article that includes processor-executable instructions stored in the computer-readable memory for implementing the functions specified in the blocks of the flowchart. The processor-executable instructions may also be loaded onto a computer or other programmable device to cause a series of operational steps to be performed on the computer or other programmable data processing apparatus to generate a computer-implemented process, thereby providing steps for implementing the functions specified in the blocks of the flowchart by processor-executable instructions that execute on the computer or other programmable device.

[0017] The blocks of the block diagrams and flowcharts support combinations of devices for performing the specified functions, combinations of steps for performing the specified functions, and program instruction means for performing the specified functions. It will also be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, can be implemented by a special purpose hardware-based computer system for performing the specified function or functions, or by a combination of special purpose hardware and computer instructions.

[0018] This paper describes a method and system for generating machine learning classifiers for predicting drug treatment responses to diseases. Machine learning (ML) is a subfield of computer science that gives computers the ability to learn without explicit programming. Machine learning platforms include, but are not limited to, naive Bayesian classifiers, support vector machines, decision trees, and neural networks. In one embodiment, baseline (pre-treatment) gene expression data may be obtained for multiple patients before treatment, and intra-treatment gene expression data may be obtained for multiple patients during treatment. Patients who responded to the treatment and patients who did not can be determined. In one embodiment, baseline gene expression data and / or intra-treatment gene expression data may be analyzed to determine one or more predictor genes. One or more predictor genes may predict whether a patient is a responder or non-responder to the drug. In one embodiment, one or more predictor genes may be determined by analyzing baseline gene expression data, intra-treatment gene expression data, and / or selected gene set enriched data from one or more other studies. In one embodiment, one or more predictor genes may be determined by analyzing gene expression related to one or more metabolic pathways.

[0019] One embodiment (shown in Figure 1) discloses a method 100 for generating a predictive model, which includes determining first gene data related to a plurality of genes in 110, determining second gene data related to a plurality of genes in 120, determining a plurality of features for a predictive model based on the first and second gene data in 130, and generating a predictive model based on the plurality of features in 140.

[0020] The first genetic data may include one or more of the following: a list of multiple genes, sequence data associated with the list of genes, enrichment data, and / or similar. Multiple genes in the first genetic data may be associated with the first multiple tumor samples. Each tumor sample in the first multiple tumor samples may be labeled as a responder or non-responder to the treatment.

[0021] The first genetic data can be called a selected, disease-free gene set data, because, as described below, the selected, disease-free gene set data may be associated with the same treatment as the second genetic data, but may also be associated with the same or different diseases. In one embodiment, the selected, disease-free gene set data may not be associated with the same treatment or the same disease as the second genetic data described below, but may be associated with one or more categories of gene sets, such as immune cell type / functional gene sets, tumor microenvironment component and signaling gene sets, or cancer cell proliferation and DNA repair gene sets. The selected, disease-free gene set data may contain at least one gene common to the second genetic data.

[0022] Determining the primary gene data in 110 may involve downloading / acquiring / receiving a curated set of disease-free gene data, which may be obtained from various sources, including recent publications and / or publicly available databases. The curated set of disease-free gene data may include multiple gene datasets that may be associated with different conditions (e.g., melanoma, breast cancer, lung cancer, ovarian cancer, etc.) and may be generated from various data types and / or platforms (e.g., bulk RNA-seq, single-cell RNA-seq, NanoString, etc.). The methods described herein may utilize curated set of disease-free gene data to improve the identification of predictor genes.

[0023] The second genetic data may include one or more of the following: a list of multiple genes, sequence data associated with the list of genes, enrichment data, and / or similar. Multiple genes in the second genetic data may be sequenced from a second set of tumor samples. Each tumor sample in the second set of tumor samples may be labeled as a responder or a non-responder. Determining the second genetic data associated with multiple genes in 120 may include determining the baseline (pre-treatment) gene expression level for each tumor associated with the second set of tumor samples. Each tumor may be treated with a therapeutic agent, and after treatment, it may be determined which tumors are responders or non-responders to the therapeutic agent. The baseline (pre-treatment) gene expression level for each tumor may then be labeled as a responder or a non-responder and stored as the second genetic data. In one embodiment, the baseline gene expression data and the gene expression data during treatment may include one or more of the following: RNA-Seq data, TCR-Seq data, DNA-Seq data, and / or imaging data. RNA-Seq data may indicate the presence and amount of RNA in a biological sample. TCR-seq data can indicate the presence and quantity of T cell receptors in a biological sample. DNA-seq data can indicate the presence and quantity of DNA and / or mutations in a biological sample.

[0024] Figures 2 and 3 describe how, based on the first and second gene data, several features are determined for the predictive model at 130, and based on these features, a predictive model at 140 is generated.

[0025] In one embodiment, a predictive model (e.g., a machine learning classifier) ​​may be generated to classify patients as responders or non-responders based on an analysis of their baseline gene expression data. The predictive model may be trained according to a first set of gene data (e.g., a curated, disease-independent gene set) and a second set of gene data (e.g., baseline gene expression data and / or gene expression data during treatment). The baseline gene expression data and the gene expression data during treatment may relate to a single trial containing the same patient cohort treated with the drug / treatment. The curated, disease-independent gene set may contain at least one gene common to the baseline gene expression data and may relate to one or more different trials containing different patient cohorts treated with the same or different drugs / treatments and having the same or different diseases. In one embodiment, one or more features of the predictive model may be extracted from one or more of the baseline gene expression data, the gene expression data during treatment, and / or the curated, disease-independent gene set. In one embodiment, one or more features of the predictive model may be extracted from one or more combinations of a portion of baseline gene expression data and / or a portion of a selected, disease-independent gene set data.

[0026] As shown in Figure 2, the System 200 described herein is configured to train at least one machine learning-based classifier 230, which is configured to classify baseline gene expression data as being associated with responders or non-responders based on an analysis of one or more training datasets 210A-210B by a training module 220 using machine learning techniques. In one embodiment, training dataset 210A (e.g., first gene data) may include a selection of disease-independent gene sets from one or more tests (e.g., one or more gene lists). In one embodiment, training dataset 210A may include only the selection of disease-independent gene sets, or only a portion of the selection of disease-independent gene sets. In one embodiment, training dataset 210B (e.g., second gene data) may include labeled baseline gene expression data. In one embodiment, training dataset 210B may include only the labeled baseline gene expression data, or only a portion of the labeled baseline gene expression data. The labeling may include responders and non-responders.

[0027] Secondary genetic data for each patient may be randomly assigned to either the training dataset 210B or the test dataset. In some implementations, the assignment of data to the training dataset or the test dataset may not be entirely random. In this case, one or more criteria may be used during assignment, such as ensuring that a similar number of patients with different responder / non-responder states are present in each of the training and test datasets. Generally, any preferred method may be used to assign data to the training dataset or the test dataset, while ensuring that the distribution of responder / non-responder states is somewhat similar in the training and test datasets. In one embodiment, 75% of the labeled baseline gene expression data may be assigned to the training dataset 210B, and 25% of the labeled baseline gene expression data may be assigned to the test dataset.

[0028] In one embodiment, the training module 220 may train a machine learning-based classifier 230 by extracting a feature set from a first gene data set (e.g., a curated set of disease-free genes) in the training dataset 210A according to one or more feature selection techniques. In one embodiment, the training module 220 may further define the feature set obtained from the training dataset 210A by applying one or more feature selection techniques to a second gene data set (e.g., labeled baseline gene expression data) in the training dataset 210B, which includes statistically significant features of positive examples (e.g., responders) and statistically significant features of negative examples (e.g., non-responders).

[0029] In one embodiment, the training module 220 may extract feature sets from the training dataset 210A and / or the training dataset 210B in various ways. The training module 220 may perform feature extraction multiple times in each iteration using different feature extraction techniques. In one embodiment, each of the feature sets generated using different techniques may be used to generate different machine learning-based classification models 240. In one embodiment, the feature set with the highest quality metric may be selected for use in training. The training module 220 may use the feature sets to construct one or more machine learning-based classification models 240A-240N configured to indicate whether new data relates to a responder or a non-responder.

[0030] In one embodiment, the training dataset 210B may be analyzed to determine any dependencies, associations, and / or correlations between measured gene expression levels and the responder / non-responder states of patients within the training dataset 210B. Identified correlations may take the form of a list of genes differentially expressed for samples associated with different responder / non-responder states. In one embodiment, the training dataset 210A may be analyzed to determine one or more lists of genes that have at least one gene in common with the training dataset 210B. Genes may be considered features (or variables) in the context of machine learning. As used herein, the term “feature” may refer to any feature of an item of data that can be used to determine whether an item of data falls into one or more specific categories. As an example, features described herein may include one or more genes.

[0031] In one embodiment, the feature selection technique may include one or more feature selection rules. These feature selection rules may include genegenetic rules. The genegenetic rules may include determining which genes occur a threshold number of times in the training dataset 210A, and identifying those genes that satisfy the threshold as candidate features. For example, any gene that appears two or more times in the training dataset 210A may be considered a candidate feature. Any gene that appears less than two times may be excluded from consideration as a feature.

[0032] In one embodiment, one or more feature selection rules may include expression level rules. Expression level rules may include determining which genes in the baseline gene expression data in the training dataset 210B have expression levels above an expression threshold, and identifying those genes that meet the threshold as candidate features. For example, any gene with an expression level of 2 transcripts / million (TPM) or higher may be considered a candidate feature. Any gene with an expression level of less than 2 TPM may be excluded from consideration as a feature.

[0033] In one embodiment, one or more feature selection rules may include significance rules. Significance rules may include determining responder gene expression data and non-responder gene expression data from baseline gene expression data in the training dataset 210B. Since the baseline gene expression data in the training dataset 210B is labeled as responder or non-responder, the labeling may be used to determine responder gene expression data and non-responder gene expression data. The gene expression levels of genes in the responder gene expression data may be compared to the gene expression levels of those same genes in the non-responder gene expression data. Genes with statistically significant (e.g., p-value) differential expression may be determined based on the comparison. For example, genes with differential expression having a p-value below a threshold may be selected as candidate features. The threshold may be, for example, 0.1. Genes with differential expression having a p-value above the threshold may be excluded from consideration as features.

[0034] In one embodiment, one or more feature selection rules may include tumor mutational load (TMB) rules. A TMB rule may include determining the TMB value for each gene included in training dataset 210A and / or training dataset 210B. The TMB values ​​may be used as features.

[0035] In one embodiment, a single feature selection rule may be applied to select features, or multiple feature selection rules may be applied to select features. In one embodiment, the feature selection rules may be applied in a cascading manner, where the feature selection rules are applied in a specific order and applied to the results of previous rules. For example, gene development rules may be applied to the training dataset 210A to generate a first list of genes. Expression level rules may be applied to the genes in the first list to determine which genes in the first list satisfy the expression level rule in the training dataset 210B, and generate a second list of genes. Significance rules may be applied to the genes in the second list to determine which genes in the second list satisfy the significance rule in the training dataset 210B, and generate a final list (features) of candidate genes.

[0036] The final list of candidate genes may be analyzed according to additional feature selection techniques to determine one or more candidate gene signatures (e.g., a set of genes that can be used to predict whether a patient is a responder or non-responder). Candidate gene signatures may be identified using any feature selection technique, such as filtering methods, wrapping methods, and / or embedding methods, using any suitable computational technique. In one embodiment, one or more candidate gene signatures may be selected according to a filtering method. Filtering methods include, for example, Pearson correlation, linear discriminant analysis, analysis of variance (ANOVA), chi-squared, and combinations thereof. Feature selection according to a filtering method is independent of any machine learning algorithm. Alternatively, features may be selected based on their scores in various statistical tests for correlation with an outcome variable (e.g., responder / non-responder).

[0037] In one embodiment, one or more candidate gene signatures may be selected according to a wrapper method. The wrapper method may be configured to use a subset of features and to train a machine learning model using the subset of features. Features may be added to and / or removed from the subset based on inferences drawn from previous models. Wrapper methods include, for example, forward feature selection, backward feature reduction, recursive feature reduction, and combinations thereof. In one embodiment, forward feature selection may be used to identify one or more candidate gene signatures. Forward feature selection is an iterative method that starts with no features in the machine learning model. In each iteration, features that best improve the model are added until the performance of the machine learning model no longer improves by adding new variables. In one embodiment, backward reduction may be used to identify one or more candidate gene signatures. Backward reduction is an iterative method that starts with all features in the machine learning model. In each iteration, the lowest-ranking features are removed until no improvement is observed when features are removed. In one embodiment, recursive feature reduction may be used to identify one or more candidate gene signatures. Recursive feature reduction is a greedy optimization algorithm that aims to find the best-performing feature subset. Recursive feature reduction iteratively constructs a model, setting aside the best-performing or worst-performing features at each iteration. It continues building the next model with remaining features until all features have been exhausted. Finally, the features are ranked based on the order in which they were removed.

[0038] In one embodiment, one or more candidate gene signatures may be selected according to an embedding method. The embedding method combines the qualities of a filtering method and a wrapping method. Embedding methods include, for example, the least absolute contraction and selection operator (LASSO) and ridge regression, which implement penalty functions to reduce overfitting. For example, LASSO regression implements L1 regularization, which adds a penalty equivalent to the absolute value of the coefficient magnitude, and ridge regression implements L2 regularization, which adds a penalty equivalent to the square of the coefficient magnitude.

[0039] After the training module 220 generates a feature set, the training module 220 may generate a machine learning-based classification model 240 based on the feature set. A machine learning-based classification model can refer to a complex mathematical model for data classification generated using machine learning techniques. In one example, this machine learning-based classifier may include a map of support vectors representing boundary features. In this example, the boundary features may be selected from and / or represent the highest-ranking features in a given feature set.

[0040] In one embodiment, the training module 220 may use feature sets extracted from the training dataset 210A and / or the training dataset 210B to construct machine learning-based classification models 240A-240N for each classification category (e.g., responder, non-responder). In some examples, the machine learning-based classification models 240A-240N may be combined into a single machine learning-based classification model 240. Similarly, the machine learning-based classifier 230 may represent a single classifier containing one or more machine learning-based classification models 240, and / or multiple classifiers containing one or more machine learning-based classification models 240.

[0041] The extracted features (e.g., one or more candidate genes and / or candidate gene signatures derived from the final list of candidate genes) may be combined in a classification model trained using machine learning approaches, such as discriminant analysis; decision trees; nearest neighbor (NN) algorithms (e.g., k-NN models, replicator NN models, etc.); statistical algorithms (e.g., Bayesian networks, etc.); clustering algorithms (e.g., k-means, mean shifts, etc.); neural networks (e.g., reservoir networks, artificial neural networks, etc.); support vector machines (SVM); logistic regression algorithms; linear regression algorithms; Markov models or chains; principal component analysis (PCA) (e.g., for linear models); multilayer perceptron (MLP) ANN (e.g., for nonlinear models); reservoir network replication (e.g., for nonlinear models, usually for time series); random forest classification; combinations thereof and / or similar. The resulting machine learning-based classifier 230 may include decision rules or mappings that assign patients to classes (responder / non-responder) using the gene expression levels in the candidate gene signatures.

[0042] Candidate gene signatures and a machine learning-based classifier 230 may be used to predict the responder / non-responder status of test samples in a test dataset. In one example, the result for each test sample includes a confidence level corresponding to the likelihood or probability that the corresponding test sample belongs to a predicted responder / non-responder status. The confidence level may be a value between 0 and 1 and represents the likelihood that the corresponding test sample belongs to a responder / non-responder status. In one example, if there are two states (e.g., responder and non-responder), the confidence level may correspond to a value p, which indicates the likelihood that a particular test sample belongs to the first state. In this case, the value 1-p may indicate the likelihood that a particular test sample belongs to the second state. In general, if there are states greater than 2, multiple confidence levels may be provided for each test sample and for each candidate gene signature. The best-performing candidate gene signature may be determined by comparing the results obtained for each test sample with the known responder / non-responder status for each test sample. Generally, the best-performing candidate gene signature will have results that closely match known responder / non-responder states.

[0043] The most effective candidate gene signature may be used to predict an individual's responder / non-responder status. For example, baseline gene expression data for potential patients may be determined / received. Baseline gene expression data for potential patients may be provided to a machine learning-based classifier 230, which may classify potential patients as responders or non-responders based on the most effective candidate gene signature. If classified as a responder, the candidate patient may be treated with a drug / treatment. If classified as a non-responder, an alternative treatment may be offered to the potential patient.

[0044] Figure 3 is a flowchart illustrating an exemplary training method 300 for generating a machine learning-based classifier 230 using the training module 220. The training module 220 can implement supervised, unsupervised, and / or semi-supervised (e.g., augmentation-based) machine learning-based classification models 240. The method 300 illustrated in Figure 3 is an example of a supervised learning method; variations of this example of the training method are considered below, however other training methods can be implemented similarly to train unsupervised and / or semi-supervised machine learning models.

[0045] Training method 300 may determine (e.g., access, receive, acquire, etc.) primary genetic data (e.g., a list of genes, expression data, etc.) for one or more populations of patients, and secondary genetic data (e.g., access, receive, acquire, etc.) for one or more other populations of patients. The primary genetic data may include one or more datasets, each dataset relating to a specific trial. Each trial may include one or more genes common to the secondary genetic data. Each trial may include or not include the same drug / treatment, and may or may not relate to the same or different disease / condition. Each trial may include different patient populations, but it is intended that there may be some overlap in patients. In one embodiment, each dataset may include a list of differentially expressed genes. The secondary genetic data may include one or more datasets, each dataset relating to a specific trial and different from that of the primary genetic data. Each trial may include one or more genes common to the primary genetic data. Each trial may include or not include the same drug / treatment, and may or may not relate to the same or different disease / condition. Each study may include different patient populations, but some overlap in patients is intended. In one embodiment, each dataset may include a labeled list of differentially expressed genes. In another embodiment, each dataset may include labeled baseline gene expression data. In yet another embodiment, each dataset may further include labeled gene expression data during treatment. Labeling may include responder or non-responder. Gene expression data may include whole exome sequencing data, whole genome sequencing data, RNA-seq data, or combinations thereof. Gene expression data may include identification and expression levels of genes present in the patient's biological sample. For example, in the case of RNA-seq data, the amount and sequence of RNA in the biological sample may be determined using next-generation sequencing (NGS).

[0046] Training method 300 may generate training datasets and test datasets in 320. Training datasets and test datasets may be generated by randomly assigning labeled gene expression data from individual patients from second genetic data to either the training dataset or the test dataset. In some implementations, the assignment of patients as training or test samples may not be entirely random. In one embodiment, training datasets and test datasets may be generated using only labeled baseline gene expression data for a particular test. In one embodiment, training datasets may be generated using the majority of labeled baseline gene expression data for a particular test. For example, 75% of the labeled baseline gene expression data for a particular test may be used to generate the training dataset, and 25% may be used to generate the test dataset. In another embodiment, training datasets and test datasets may be generated using only labeled in-treatment gene expression data for a particular test.

[0047] The training method 300 may determine (e.g., extract, select, etc.) one or more features that can be used by a classifier to distinguish between different classifications (e.g., responders vs. non-responders). One or more features may include a set of genes. In one embodiment, the training method 300 may determine a set of features from a first set of gene data. In another embodiment, the training method 300 may determine a set of features from a second set of gene data. In another embodiment, the set of features may be determined from gene data from a different test than the test associated with the labeled gene data of the training and test datasets. In other words, gene data from a different test (e.g., selected, disease-free gene data) may be used for feature determination rather than for training a machine learning model. In one embodiment, the training dataset may be used in conjunction with gene data from a different test to determine one or more features. Gene data from a different test is used to determine an initial set of features, which may be further reduced using the training dataset.

[0048] Training method 300 allows for the training of one or more machine learning models in 340 using one or more features. In one embodiment, the machine learning models may be trained using supervised learning. In another embodiment, other machine learning techniques, including unsupervised and semi-supervised learning, may be used. The machine learning models trained in 340 may be selected based on different criteria depending on the problem to be solved and / or the data available in the training dataset. For example, machine learning classifiers may be subject to different degrees of bias. Thus, more than one machine learning model can be trained in 340, optimized, improved, and cross-validated in 350.

[0049] Training method 300 may select one or more machine learning models to build a predictive model in 360 (e.g., a machine learning classifier). The predictive model may be evaluated using a test dataset. The predictive model may analyze the test dataset and generate classification values ​​and / or predictive values ​​in 370. The classification values ​​and / or predictive values ​​may be evaluated in 380 to determine whether such values ​​have achieved the desired level of accuracy. The performance of the predictive model can be evaluated in numerous ways based on the classification of a number of true positives, false positives, true negatives, and / or false negatives of multiple data points indicated by the predictive model. For example, false positives in the predictive model may refer to the number of times the predictive model incorrectly classified a patient as a responder when the patient was actually a non-responder. Conversely, false negatives in the predictive model may refer to the number of times the machine learning model classified one or more patients as non-responders when the patient was actually a responder. True negatives and true positives may refer to the number of times the predictive model correctly classified one or more patients as responders or non-responders. The concepts of recall and accuracy are relevant to these measurements. Generally, recall refers to the ratio of true positives to the sum of true positives and false negatives, thereby quantifying the sensitivity of a predictive model. Similarly, precision refers to the ratio of true positives to the sum of true positives and false positives.

[0050] If such a desired level of accuracy is reached, the training phase may be terminated and the predictive model may be output at 390; however, if the desired level of accuracy is not reached, subsequent iterations of the training method 300 may be carried out, starting at 310 with variations, for example, by considering a larger collection of gene expression data.

[0051] Figure 4 shows gene expression data (e.g., RNA-seq data) obtained from a cohort of patients treated with a drug for disease (CSCC data). The patient cohort was treated with cemiplimab for 48 weeks for the treatment of cutaneous squamous cell carcinoma (CSCC). All patients in the cohort underwent baseline pre-treatment screening before initiating treatment. During baseline (pre-treatment) screening, a biopsy sample of each patient's tumor was taken and each biopsy sample was sequenced using next-generation sequencing (NGS) technology. Baseline gene expression data for each patient was thus obtained pre-treatment (e.g., day 1). After initiating treatment, another biopsy sample of each patient's tumor was taken and each biopsy sample was sequenced using NGS technology to obtain gene expression data during treatment. Gene expression data during treatment for each patient was thus obtained during the treatment period (e.g., day 29). While described as being determined in association with cemiplimab and CSCC, it should be understood that the methods and systems described herein may be applicable to any treatment and any condition. Therefore, baseline and treatment-dependent gene expression data can be determined for any drug / treatment and for any disease / condition. Baseline and treatment-dependent gene expression data may include one or more of the following: RNA-Seq data, TCR-Seq data, DNA-Seq data, and / or imaging data.

[0052] After treatment, patients were classified as responders or non-responders. Patients could be counted as responders if they exhibited a reduction of more than 30% in tumor volume. Patients may be classified as responders or non-responders by varying the percentage reduction in tumor volume (e.g., 10%, 20%, 40%, 50%, 60%, 70%, 80%, 100%) using other techniques. Baseline and treatment-dependent gene expression data for each patient were then labeled as responders or non-responders.

[0053] The therapeutic effect of cemiprimab is an increase in the expression of specific immune cell marker genes. As shown in Figure 5, putative immune marker gene expression in CSCC suggests that cemiprimab tends to increase the infiltration of immune cell subsets, and this is more pronounced in responders. Figure 6A shows differentially expressed genes, determined by comparing baseline gene expression data with gene expression during treatment for all patients (responders and non-responders). Figure 6B shows differentially expressed genes, determined by comparing baseline gene expression data with gene expression during treatment for responders only. Figure 6C shows differentially expressed genes, determined by comparing baseline gene expression data with gene expression during treatment for non-responders only. Figures 6B and 6C show that responders have larger gene expression changes than non-responders.

[0054] One or more predictive genes can be determined by comparing and analyzing labeled baseline gene expression data and / or labeled in-treatment gene expression data. Figure 7 shows the top 50 pharmacodynamic genes with overlap between responders and non-responders. Of the 252 identified predictive genes for responders and 14 identified predictive genes for non-responders, only two predictive genes were common between responders and non-responders. As shown in Figure 8, attempts to identify genes differentially expressed between baseline responders and baseline non-responders have revealed very few statistically significant genes. The inability to determine sufficient predictive genes using baseline gene expression data results from heterogeneous baseline samples. For example, tumor purity is often not quantified, and biopsy sites are often inconsistent between patients (e.g., skin, lung, head, neck, etc.). The result is the identification of differentially expressed genes that are tissue-specific. Consequently, it is difficult to generate a baseline machine learning classifier according to baseline gene expression data from this single test.

[0055] In one embodiment, selected, disease-free gene set data from other studies involving the same drug / treatment may be analyzed to improve the identification of predictive genes. Selected, disease-free gene set data may be obtained from various sources, including recent publications. Selected, disease-free gene set data may include multiple gene set data associated with different conditions (e.g., melanoma, breast cancer, lung cancer, ovarian cancer, etc.) and may be generated from various data types and / or platforms (e.g., bulk RNA-seq, single-cell RNA-seq, NanoString, etc.). The selected, disease-free gene set includes at least one gene common to the CSCC data.

[0056] In this example, the selected, disease-free gene sets were determined from one or more of the following publications: ·

[2018] [Journal of ImmnoTherapy of Cancer][Turan T.et al][Immune oncology immune responsiveness and the theory of everything] ·

[2005] [Richard D.Wood et al.][Human DNA repair genes,2005] ·

[2017] [Cell Reports][Charoentong P.et al.][Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade] ·

[2017] [Wouter Hendrickx et al.][Identification of genetic determinants of breast cancer immune phenotypes by integrative genome scale analysis] ·

[2012] [CancerImmImmunotherapy][Ji R.et al][An immune-active tumor microenvironment favors clinical response to ipilimumab] ·

[2013] [JCO][Ulloba-Montoya F.et al][Predictive Gene Signature in MAGE-A3 Antigen-Specific Cancer Immunotherapy] ·

[2018] [Nature Medicine][Peng Jiang][Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response Nat Medicine Aug 2018] ·

[2018] [Nature Medicine][Noam Auslander][Robust prediction of response to immune checkpoint blockade therapy in metastatic melanoma Nat Medicine Aug 2018] ·

[2018] [NatMedicine][Savas.P.et al][Single-cell profiling of breast cancer T cells reveals a tissue-resident memory subset associated with improved prognosis] ·

[2018] [Cell][Jerby-Arnon L.et al.][Signature of T cell exclusion and ICI resistance Cancer Cell 2018]

[0057] Figure 9 shows data from selected, disease-free gene sets. These selected, disease-free gene sets may be categorized. For example, disease-free gene sets could be categorized as immune cell type / function gene sets, tumor microenvironment components and signaling gene sets, and cancer cell proliferation and DNA repair gene sets. Similar to the initially obtained baseline gene expression data and gene expression data during treatment, attempts to identify genes differentially expressed between baseline responders and baseline non-responders using only selected, disease-free gene sets reveal few statistically significant genes. Figure 10 shows that predictive genes identified using only selected, disease-free gene sets only partially explain the clinical outcomes of the CSCC cohort.

[0058] The machine learning techniques described herein were applied to baseline gene expression data and selected, disease-independent gene set data. Figure 11 shows the best-performing gene signatures generated using the machine learning techniques described above. Figure 11 shows the standardized gene expression of the best-performing predicted gene signature. Patient samples are identified as R (responder) or NR (non-responder) at the top of Figure 11. Figure 11 shows that patients with higher expression of the best-performing predicted gene signature (dark red in Figure 11) are more likely to be responders (orange), while patients with lower expression of the best-performing predicted gene signature (dark blue in Figure 11) are more likely to be non-responders (light blue).

[0059] Figure 12 shows the performance of gene signatures in classifying patients during machine learning model training, applied to the test dataset based on cross-validation. The area under the receiver operating curve (ROC) (AUC) represents the performance of the classification method.

[0060] In one embodiment, Figure 13 shows a method 1300 for a systems biology approach to identify predictive transcription factor genes.

[0061] A transcription factor network can be generated in 1310. Transcription factor data can be obtained from a gene ontology (GO) resource. The transcription factor data may include a list of genes identified as transcription factor genes, and any genes that may affect the transcription of other genes annotated in GO. The transcription factor network can be generated, for example, by ARACNE (an algorithm for reconstructing gene regulatory networks in mammalian cell contexts). The transcription factor network may include multiple nodes (each node being a gene (transcription factor gene or target gene)) and multiple edges, where an edge between two nodes may indicate a relationship. This relationship may indicate a transcription factor gene associated with one or more target genes. This relationship may include, for example, "is a transcription factor of" or "transcription is regulated by". In one embodiment, baseline gene expression data may be used to filter the transcription factor data. Genes present in both the gene expression data and the target genes in the transcription factor data may be identified. A transcription network may be generated using the identified genes and associated transcription factor genes. In one embodiment, a cross-information-based method is constructed to determine the relationship between transcription factor genes and any other genes in gene expression data, such that a transcriptional network is built connecting transcription factor genes and their target genes.

[0062] The transcription factor network may be refined at 1320. Refining the transcription factor network may involve removing one or more edges that are likely to have occurred by chance. Refinement may be performed based on the number of samples in the gene expression data used to construct the network, and the calculation of the probability of each network connection can be reliably found given the sample size. For example, network connections may be randomly rearranged and the probability of observing that network connection can be determined. Any network connection with a statistically insignificant probability (e.g., higher than the p-value) may be removed.

[0063] In step 1330, sequentially or in parallel, the genes for each target in the baseline gene expression data may be ranked by their expression derived from the baseline gene expression data. In step 1340, the transcription factor genes that target the ranked genes may be determined based on the transcription factor network and the list of ranked genes. Across the transcription network, nodes associated with sets of target genes also found in the gene ranking list may be identified. From these nodes, edges may be determined that identify transcription factor genes associated with sets of target genes.

[0064] In 1350, for each subject in the baseline gene expression data, an enrichment score can be determined for each transcription factor gene associated with that subject. The enrichment score for a transcription factor gene may be based on the gene expression order of its transcription target gene identified in the transcriptional network.

[0065] The enrichment score for each transcription factor gene may be compared using a scale of 1360. For example, the ratio of enrichment scores for transcription factor genes between baseline responders and non-responders can be determined.

[0066] One or more predictive transcription factor genes can be determined by 1370. One or more predictive transcription factor genes can be determined by evaluating the statistical significance of the ratio of enrichment scores for a given transcription factor gene. Transcription factor genes with statistically significant ratios of enrichment scores can be identified as predictive transcription factor genes.

[0067] Candidates for therapeutic interventions can be identified using one or more predictive transcription factor genes. Baseline gene expression data can be obtained from new subjects. Baseline gene expression data were ranked, target genes of network-based predictive transcription factor genes were collected, and then enrichment scores were calculated to identify the activity of predictive transcription factor genes. If a subject has high enrichment scores for one or more predictive transcription factor genes, then the subject is a candidate for therapeutic intervention.

[0068] Figure 14 shows exemplary top predictive transcription factor genes and their enrichment scores, determined using baseline samples from the previously described CSCC cohort.

[0069] Figure 15 is a block diagram depicting an environment 1500 including a non-limiting example of computing devices 1501 and servers 1502 connected via a network 1504. In one embodiment, some or all steps of any described method can be performed on the computing devices described herein. The computing device 1501 may include one or more computers configured to store one or more of the following: a training module 220, training data 210 (e.g., labeled baseline gene expression data, labeled gene expression data during treatment, and / or selected, disease-independent gene set data). The server 1402 may include one or more computers configured to store gene data 1524 (e.g., selected, disease-independent gene set data). Multiple servers 1502 can communicate with the computing device 1501 via the network 1504.

[0070] The computing device 1501 and server 1502 may be digital computers, with respect to their hardware architecture, generally including a processor 1508, a memory system 1510, an input / output (I / O) interface 1512, and a network interface 1514. These components (1508, 1510, 1512, and 1514) are communicatively connected via a local interface 1516. The local interface 1516 may be, but is not limited to, one or more buses or other wired or wireless connections known in the art. The local interface 1516 may have additional elements (omitted for simplicity) to enable communication, such as controllers, buffers (caches), drivers, repeaters, and receivers. Furthermore, the local interface may include address, control, and / or data connections to enable appropriate communication between the aforementioned components.

[0071] The processor 1508 may be a hardware device for executing software, particularly stored in the memory system 1510. The processor 1508 can be any custom-made or commercially available processor, a central processing unit (CPU), an auxiliary processor among several processors associated with the computing device 1501 and server 1502, a semiconductor-based microprocessor (in the form of a microchip or chipset), or any device in general for executing software instructions. When the computing device 1501 and / or server 1502 are operating, the processor 1508 may be configured to execute software stored in the memory system 1510 to communicate data to and from the memory system 1510, and to generally control the operation of the computing device 1501 and server 1502 according to the software.

[0072] The I / O interface 1512 can be used to receive user input from and / or provide system output to one or more devices or components. User input may be provided, for example, via a keyboard and / or mouse. System output may be provided via a display device and / or printer (not shown). The I / O interface 1512 may include, for example, a serial port, a parallel port, a Small Computer System Interface (SCSI), an infrared (IR) interface, a radio frequency (RF) interface, and / or a Universal Serial Bus (USB) interface.

[0073] The network interface 1514 can be used to send and receive data from the computing device 1501 and / or from the server 1502 on the network 1504. The network interface 1514 may include, for example, a 10BaseT Ethernet® adapter, a 100BaseT Ethernet adapter, a LAN PHY Ethernet adapter, a Token Ring adapter, a wireless network adapter (e.g., WiFi, cellular, satellite), or any other suitable network interface device. The network interface 1514 may include address, control, and / or data connections to enable proper communication on the network 1504.

[0074] The memory system 1510 may include one or a combination of volatile memory elements (e.g., random access memory (RAM such as DRAM, SRAM, SDRAM, etc.)) and non-volatile memory elements (e.g., ROM, hard drive, tape, CD-ROM, DVD-ROM, etc.). Furthermore, the memory system 1510 may incorporate electronic, magnetic, optical, and / or other types of storage media. It should be noted that the memory system 1510 may have a distributed architecture in which various components are located apart from each other but can be accessed by the processor 1508.

[0075] The software in the memory system 1510 may include one or more software programs, each of which includes an ordered list of executable instructions for performing a logical function. In the example in Figure 15, the software in the memory system 1510 of computing device 1501 may include a training module 220 (or its subcomponents), training data 220, and a preferred operating system (O / S) 1518. In the example in Figure 15, the software in the memory system 1510 of server 1502 may include gene data 1524 and a preferred operating system (O / S) 1518. The operating system 1518 essentially controls the execution of other computer programs and provides scheduling, input-output control, file and data management, memory management, and communication control, as well as related services.

[0076] For illustrative purposes, application programs and other executable program components, such as the operating system 1518, are illustrated herein as separate blocks, but it is recognized that such programs and components may reside at different times in different storage components of the computing device 1501 and / or server 1502. Implementations of the training module 220 may be stored on or transmitted on some form of computer-readable medium. Any of the methods of this disclosure can be executed by computer-readable instructions embodied on the computer-readable medium. The computer-readable medium can be any available medium accessible by a computer. For example, and not intended to limit, the computer-readable medium may include “computer storage medium” and “communication medium.” “Computer storage medium” may include volatile and non-volatile removable and non-removable media, implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data. Exemplary computer storage media may include RAM, ROM, EEPROM, flash memory or other storage technologies, CD-ROM, digital versatile disk (DVD) or other optical storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or any other media that can be used to store desired information and can be accessed by a computer.

[0077] In one embodiment, the training module 220 may be configured to perform method 1600 as shown in Figure 16. Method 1600 may be performed entirely or partially by a single computing device, multiple electronic devices, and so on. Method 1600 may include determining a first genetic data set associated with multiple genes in 1610. The first genetic data set may consist of genetic data from multiple different datasets. The first genetic data set may be obtained from publicly available data sources, and the multiple genes may include one or more of the following: an immune cell type / function gene set, a tumor microenvironment component and signaling gene set, or a cancer cell proliferation and DNA repair gene set.

[0078] Method 1600 may include determining second genetic data related to multiple genes in 1620. The multiple genes may be sequenced from multiple tumor samples. Each tumor sample from the multiple tumor samples may be labeled as a responder or a non-responder. Determining second genetic data related to multiple genes may include determining baseline gene expression levels for each tumor related to the multiple tumor samples, treating each tumor related to the multiple tumor samples with a therapeutic agent, determining after treatment which tumors related to the multiple tumor samples are responders or non-responders to the therapeutic agent, labeling the baseline gene expression levels of each tumor related to the multiple tumor samples as responders or non-responders, and generating second genetic data based on the labeled baseline gene expression levels.

[0079] Method 1600 may include determining multiple features of a predictive model based on first and second genetic data in 1630. Determining multiple features of a predictive model based on first and second genetic data may include determining, from the first genetic data, a first set of candidate genes present in two or more of several different datasets as a first set of candidate genes; from the second genetic data, a second set of candidate genes, a first set of candidate genes expressed at 2 transcripts / million (TPM) or more in at least half of several tumor samples; and from the second genetic data, a third set of candidate genes, a second set of candidate genes with a statistically significant increase in expression levels between responders and non-responders, wherein the multiple features include determining the third set of candidate genes. Determining multiple features of a predictive model based on first and second genetic data may include determining tumor mutation burden (TMB) values ​​for each of several tumors associated with a third set of candidate genes, and determining a fourth set of candidate genes based on the TMB values, wherein the multiple features include the fourth set of candidate genes.

[0080] Method 1600 may include training a multi-feature predictive model in 1640 based on the first portion of the second gene data. By training a multi-feature predictive model based on the first portion of the second gene data, it is possible to determine the gene signature that indicates a responder.

[0081] Method 1600 may include testing the predictive model in 1650 based on a second portion of the second gene data. Method 1600 may also include outputting the predictive model in 1660 based on the test.

[0082] In one embodiment, the training module 220 may be configured to perform method 1700 as shown in Figure 17. Method 1700 may be performed in whole or in part by a single computing device, multiple electronic devices, and so on. Method 1700 may include receiving baseline genetic data related to multiple genes for a subject in 1710. The multiple genes may be sequenced from the tumor of the subject. Method 1700 may include providing the baseline genetic data to a predictive model in 1720. Method 1700 may include determining, based on the predictive model, that the subject is a candidate for therapeutic treatment in 1730. Method 1700 may further include treating the subject with the therapeutic treatment.

[0083] Method 1700 may further include training a predictive model.

[0084] Training a predictive model may include determining first gene data related to multiple genes, determining second gene data related to multiple genes, wherein the multiple genes are sequenced from multiple tumor samples, and each tumor sample is labeled as a responder or non-responder, determining multiple features of the predictive model based on the first and second gene data, training the predictive model with the multiple features based on a first portion of the second gene data, training the predictive model with the multiple features based on a second portion of the second gene data, and outputting the predictive model based on tests.

[0085] The first genetic data may be obtained from publicly available data sources, and multiple genes may include one or more of the following: an immune cell type / function gene set, a tumor microenvironment component and signaling gene set, or a cancer cell proliferation and DNA repair gene set. The first genetic data may consist of genetic data from multiple different datasets.

[0086] Determining multiple features of a predictive model based on first and second genetic data may include determining, from the first genetic data, a first set of candidate genes that are present in two or more different datasets; determining, from the second genetic data, a second set of candidate genes that are expressed at 2 transcripts / million (TPM) or more in at least half of multiple tumor samples; and determining, from the second genetic data, a third set of candidate genes that are statistically significant increases in expression levels between responders and non-responders, with the multiple features including the third set of candidate genes.

[0087] Determining multiple features of a predictive model based on first and second genetic data may include determining tumor mutation burden (TMB) values ​​for each of several tumors associated with a third set of candidate genes, and determining a fourth set of candidate genes based on the TMB values, wherein the multiple features include the fourth set of candidate genes.

[0088] Determining a second set of gene data related to multiple genes may include determining baseline gene expression levels for each tumor related to multiple tumor samples, treating each tumor related to the multiple tumor samples with a therapeutic agent, determining which tumors related to the multiple tumor samples are responders or non-responders to the therapeutic agent after treatment, labeling the baseline gene expression levels of each tumor related to the multiple tumor samples as responders or non-responders, and generating a second set of gene data based on the labeled baseline gene expression levels.

[0089] By training a predictive model with multiple features based on the first portion of the second set of genetic data, it becomes possible to determine the gene signature that indicates a responder.

[0090] In one embodiment, the training module 220 may be configured to carry out method 1800 as shown in Figure 18. Method 1800 may be carried out entirely or partially by a single computing device, multiple electronic devices, and so on. Method 1800 may include determining baseline gene expression data associated with multiple genes in 1810. The multiple genes may be associated with multiple tumor samples, and each tumor sample of the multiple tumor samples may be labeled as a responder or non-responder to the therapeutic agent / treatment. Determining baseline gene expression data may include determining the baseline gene expression level for each tumor associated with the multiple tumor samples, treating each tumor associated with the multiple tumor samples with the therapeutic agent, determining which tumors associated with the multiple tumor samples are responders or non-responders to the treatment after treatment, labeling the baseline gene expression levels for each tumor associated with the multiple tumor samples as responders or non-responders, and generating baseline gene expression data based on the labeled baseline gene expression levels.

[0091] Method 1800 may include determining transcription factor gene data in 1820 based on multiple genes. Determining transcription factor gene data based on multiple genes may include searching a gene ontology database for any gene with transcriptional function, determining one or more transcription factor genes and associated target genes based on queries, and generating transcription factor gene data based on one or more transcription factor genes and associated target genes.

[0092] Method 1800 may include generating a transcription factor (TR) network in 1830 based on transcription factor gene data and multiple genes. Generating a TR network based on transcription factor gene data and multiple genes may include generating multiple nodes, each node representing either a transcription factor gene or a target gene, connecting two or more of the multiple nodes with one or more edges, each edge representing a relationship between a transcription factor gene and a target gene, and storing the multiple nodes and one or more edges as a TR network. This relationship may indicate that a transcription factor gene regulates the transcription of a target gene. Method 1800 may further include refining the TR network. Refining the TR network may include removing one or more edges that are likely to have occurred by chance.

[0093] Method 1800 may include determining an enrichment score associated with each transcription factor gene in a set of transcription factor genes in 1840, based on the TR network and baseline gene expression data. The enrichment score associated with each transcription factor gene in a set of transcription factor genes may be based on one or more enrichment scores associated with one or more genes in the baseline gene expression data associated with the transcription factor genes.

[0094] Method 1800 may include determining one or more predicted transcription factor genes in a set of transcription factor genes based on enrichment scores in 1850. Determining one or more predicted transcription factor genes in a set of transcription factor genes based on enrichment scores may include determining the enrichment score ratio of responders to non-responders for each transcription factor gene in the set of transcription factor genes, and determining one or more predicted transcription factor genes from the set of transcription factor genes that have a statistically significant association with responders.

[0095] Method 1800 may include determining additional baseline gene expression data for a subject, determining the presence of one or more predictive transcription factor genes in the additional baseline gene expression data, and determining whether the subject is a candidate for therapeutic treatment based on the presence of one or more predictive transcription factor genes in the additional baseline gene expression data.

[0096] Embodiment 1: A method comprising determining first gene data related to multiple genes, determining second gene data related to multiple genes, wherein the multiple genes are sequenced from multiple tumor samples, and each tumor sample is labeled as a responder or a non-responder, determining multiple features of a predictive model based on the first and second gene data, training the predictive model according to the multiple features based on a first part of the second gene data, testing the predictive model based on a second part of the second gene data, and outputting the predictive model based on this test.

[0097] Embodiment 2: An embodiment according to any one of the prior embodiments, wherein determining a first gene data related to multiple genes includes obtaining the first gene data from a publicly available data source.

[0098] Embodiment 3: An embodiment according to any one of the prior embodiments, wherein multiple genes include one or more of the following: an immune cell type / functional gene set, a tumor microenvironment component and signaling gene set, or a cancer cell proliferation and DNA repair gene set.

[0099] Embodiment 4: An embodiment according to any one of the above embodiments, wherein determining first gene data relating to multiple genes includes determining multiple genes based on second gene data, determining one or more gene datasets containing at least one of the multiple genes based on the multiple genes, and generating first gene data based on one or more gene datasets.

[0100] Embodiment 5: An embodiment according to any one of the preceding embodiments, wherein the first gene data consists of gene data from multiple different gene datasets.

[0101] Embodiment 6: An embodiment according to any one of the prior embodiments, wherein determining second genetic data related to multiple genes includes determining baseline gene expression levels for each tumor related to multiple tumor samples, treating each tumor related to multiple tumor samples with a therapeutic agent, determining which tumors related to multiple tumor samples are responders or non-responders to the therapeutic agent after treatment, labeling the baseline gene expression levels for each tumor related to multiple tumor samples as responders or non-responders, and generating second genetic data based on the labeled baseline gene expression levels.

[0102] Embodiment 7: An embodiment according to any one of Embodiments 5 to 6, wherein determining multiple features of a predictive model based on first and second gene data includes determining, from the first gene data, genes present in two or more of several different gene datasets as the first set of candidate genes; from the second gene data, genes of the first set of candidate genes expressed at 2 transcripts / million (TPM) or more in at least half of several tumor samples as the second set of candidate genes; and from the second gene data, genes of the second set of candidate genes showing a statistically significant increase in expression levels between responders and non-responders as the third set of candidate genes, wherein the multiple features include determining the third set of candidate genes.

[0103] Embodiment 8: An embodiment according to any one of Embodiments 5 to 7, wherein determining multiple features for a predictive model based on first and second genetic data includes determining tumor mutation burden (TMB) values ​​for each of several tumors associated with a third set of candidate genes, and determining a fourth set of candidate genes based on the TMB values, wherein the multiple features include the fourth set of candidate genes.

[0104] Embodiment 9: An embodiment according to any one of the prior embodiments, wherein the gene signature indicating a responder is determined by training a predictive model with multiple features based on a first portion of a second gene data set.

[0105] Embodiment 10: A method comprising receiving baseline genetic data relating to multiple genes of a subject, wherein the multiple genes are sequenced from the tumor of the subject; providing the baseline genetic data to a predictive model; and determining, based on the predictive model, that the subject is a candidate for therapeutic treatment.

[0106] Embodiment 11: An embodiment according to Embodiment 10, further comprising training a predictive model.

[0107] Embodiment 12: An embodiment according to any one of embodiments 10 to 11, further comprising training a predictive model.

[0108] Embodiment 13: An embodiment according to any one of embodiments 10 to 12, wherein training a predictive model includes determining first gene data related to multiple genes, determining second gene data related to multiple genes, wherein the multiple genes are sequenced from multiple tumor samples, and each tumor sample of the multiple tumor samples is labeled as a responder or a non-responder, determining multiple features of the predictive model based on the first and second gene data, training the predictive model according to the multiple features based on the first part of the second gene data, testing the predictive model based on the second part of the second gene data, and outputting the predictive model based on this test.

[0109] Embodiment 14: The embodiment according to Embodiment 13, wherein determining first gene data related to multiple genes includes determining multiple genes based on second gene data, determining one or more gene datasets containing at least one of the multiple genes based on the multiple genes, and generating first gene data based on the one or more gene datasets.

[0110] Embodiment 15: Embodiments according to Embodiments 13-14, wherein the first gene data consists of gene data from multiple different gene datasets.

[0111] Embodiment 16: Embodiments of claims 13-15, wherein determining second genetic data related to multiple genes includes determining baseline gene expression levels for each tumor related to multiple tumor samples, treating each tumor related to multiple tumor samples with a therapeutic agent, determining which tumors related to multiple tumor samples are responders or non-responders to the treatment after treatment, labeling the baseline gene expression levels for each tumor related to multiple tumor samples as responders or non-responders, and generating second genetic data based on the labeled baseline gene expression levels.

[0112] Embodiment 17: Embodiments 14-16, which determine multiple features of a predictive model based on first and second gene data, include determining from the first gene data that genes present in two or more of several different gene datasets as the first set of candidate genes, determining from the second gene data that genes of the first set of candidate genes expressed at 2 transcripts / million (TPM) or more in at least half of several tumor samples as the second set of candidate genes, and determining from the second gene data that genes of the second set of candidate genes with a statistically significant increase in expression levels between responders and non-responders as the third set of candidate genes, wherein the multiple features include determining the third set of candidate genes.

[0113] Embodiment 18: An embodiment according to any one of embodiments 14 to 17, wherein determining multiple features for a predictive model based on first and second genetic data includes determining tumor mutation burden (TMB) values ​​for each of several tumors associated with a third set of candidate genes, and determining a fourth set of candidate genes based on the TMB values, wherein the multiple features include the fourth set of candidate genes.

[0114] Embodiment 19: Embodiments 10-18, wherein the gene signature indicating a responder is determined by training a predictive model with multiple features based on the first portion of the second gene data.

[0115] Embodiment 20: A method comprising determining baseline gene expression data associated with multiple genes, wherein the multiple genes are associated with multiple tumor samples, and each of the multiple tumor samples is labeled as a responder or a non-responder; determining transcription factor gene data based on the multiple genes; generating a transcription factor (TR) network based on the transcription factor gene data and the multiple genes; determining an enrichment score associated with each transcription factor gene in the set of transcription factor genes based on the TR network and baseline gene expression data; and determining one or more predicted transcription factor genes in the set of transcription factor genes based on the enrichment score.

[0116] Embodiment 21: An embodiment according to Embodiment 20, wherein determining baseline gene expression data includes determining the baseline gene expression level for each tumor associated with multiple tumor samples, treating each tumor associated with multiple tumor samples with a therapeutic agent, determining, after treatment, whether each tumor associated with multiple tumor samples is a responder or non-responder to the therapeutic agent, labeling the baseline gene expression level for each tumor associated with multiple tumor samples as responder or non-responder, and generating baseline gene expression data based on the labeled baseline gene expression levels.

[0117] Embodiment 22: An embodiment according to any one of embodiments 20 to 21, wherein determining transcription factor gene data based on multiple genes includes searching a gene ontology database for any gene having transcriptional function, determining one or more transcription factor genes and associated target genes based on the query, and generating transcription factor gene data based on one or more transcription factor genes and associated target genes.

[0118] Embodiment 23: An embodiment according to any one of embodiments 20 to 22, wherein generating a TR network based on transcription factor gene data and multiple genes includes generating multiple nodes, each node representing either a transcription factor gene or a target gene; connecting two or more of the multiple nodes with one or more edges, each edge representing a relationship between a transcription factor gene and a target gene; and storing the multiple nodes and one or more edges as a TR network.

[0119] Embodiment 24: An embodiment according to any one of embodiments 20 to 23, in which the relationship shows that a transcription regulator gene regulates the transcription of a target gene.

[0120] Embodiment 25: An embodiment according to any one of embodiments 20 to 24, further comprising refining the TR network.

[0121] Embodiment 26: Embodiment 25, which refines the TR network, includes removing one or more edges that are likely to have occurred by chance.

[0122] Embodiment 27: An embodiment according to any one of embodiments 20 to 26, wherein the enrichment score associated with each transcription factor gene in the set of transcription factor genes is based on one or more enrichment scores associated with one or more genes in baseline gene expression data related to the transcription factor genes.

[0123] Embodiment 28: An embodiment according to any one of embodiments 20 to 27, wherein determining one or more predicted transcription factor genes in a set of transcription factor genes based on enrichment scores includes determining the enrichment score ratio of responders to non-responders for each transcription factor gene in the set of transcription factor genes, and determining one or more predicted transcription factor genes in the set of transcription factor genes that have a statistically significant association with responders.

[0124] Embodiment 29: An embodiment according to any one of embodiments 20 to 28, further comprising determining additional baseline gene expression data for a subject, determining the presence of one or more predictive transcription factor genes in the additional baseline gene expression data, and determining that the subject is a candidate for therapeutic treatment based on the presence of one or more predictive transcription factor genes in the additional baseline gene expression data.

[0125] While methods and systems are described in relation to preferred embodiments and specific examples, their scope is not intended to be limited to the specific embodiments described herein. This is because the embodiments described herein are intended to be illustrative rather than restrictive in all respects.

[0126] Unless otherwise specified, no method described herein is intended to be construed as requiring its steps to be performed in a specific order. Therefore, if a claim of a method does not enumerate the order in which its steps should actually be followed, or unless otherwise specified in the claims or specification, no order should be presumed in any respect. This applies to all possible implicit basis for interpretation, including issues of logic regarding the arrangement of steps or the sequence of operations, the obvious meaning derived from grammatical organization or punctuation, and the number or type of embodiments described herein.

[0127] It will become apparent to those skilled in the art that various modifications and variations can be made without departing from the scope or spirit. Other embodiments will become apparent to those skilled in the art by considering the specification and practices disclosed herein. This specification and examples are for illustrative purposes only, and the true scope and spirit are intended to be shown by the claims below.

Claims

1. Method: Determining the primary gene data associated with multiple genes; Determining second gene data related to the plurality of genes, wherein the plurality of genes are sequenced from a plurality of tumor samples, and each of the plurality of tumor samples is labeled as a responder or a non-responder; Determining a plurality of features of a predictive model based on the first gene data and the second gene data, wherein determining the plurality of features includes generating transcription factor (TR) information based on the regulatory relationships between transcription factors and the plurality of genes; Training the predictive model based on the multiple features, using the first portion of the second gene data; Test the predictive model based on the second portion of the second gene data; and A method comprising outputting the predictive model based on the aforementioned test.

2. The method according to claim 1, wherein determining the first gene data relating to multiple genes includes retrieving the first gene data from a public data source.

3. The method according to any one of claims 1 to 2, wherein the plurality of genes include one or more of the following: an immune cell type / functional gene set, a tumor microenvironment component and signal transduction gene set, or a cancer cell proliferation and DNA repair gene set.

4. To determine the first gene data related to the aforementioned multiple genes: Determining the plurality of genes based on the second gene data; Based on the plurality of genes, determine one or more gene datasets that include at least one of the plurality of genes; and The method according to any one of claims 1 to 3, comprising generating the first gene data based on the one or more gene datasets.

5. The method according to any one of claims 1 to 4, wherein the first gene data is composed of gene data from a plurality of different gene datasets.

6. To determine the second gene data related to the aforementioned multiple genes: To determine the baseline gene expression levels for each tumor associated with the aforementioned multiple tumor samples; To determine which of the tumors associated with the aforementioned multiple tumor samples is a responder or a non-responder to the therapeutic agent; Labeling the baseline gene expression levels for each tumor associated with the plurality of tumor samples as responders or non-responders; and The method according to any one of claims 1 to 5, comprising generating the second gene data based on the labeled baseline gene expression level.

7. Based on the first and second gene data, the following multiple features of the predictive model can be determined: From the aforementioned first gene data, determine a first set of candidate genes consisting of genes present in two or more of the aforementioned multiple different gene datasets; From the second set of gene data, determine the genes of the first set of candidate genes that are expressed at a rate of 2 transcripts / million (TPM) or more in at least half of the multiple tumor samples as the second set of candidate genes; and This includes determining, from the second set of gene data, the genes of the second set of candidate genes that show a statistically significant increase in expression levels between responders and non-responders as the third set of candidate genes, The method according to claim 5, wherein the plurality of features include the third set of candidate genes.

8. Based on the first gene data and the second gene data, the plurality of features of the prediction model are determined. For the third set of candidate genes, to determine the tumor gene mutational load (TMB) value for each of the plurality of tumors associated with the third set of candidate genes; and This includes determining a fourth set of candidate genes based on the TMB value, The method according to claim 7, wherein the plurality of features include the fourth set of candidate genes.

9. The method according to any one of claims 1 to 8, wherein the gene signature indicating a responder is determined by training the predictive model with a plurality of features based on the first portion of the second gene data.

10. Receiving baseline genetic data related to multiple genes of a subject, wherein the multiple genes are sequenced from the tumor of the subject; and The method according to any one of claims 1 to 9, further comprising providing the baseline gene data to the prediction model.

11. The method according to claim 10, further comprising determining, based on the predictive model, that the subject is a candidate for therapeutic treatment.

12. The method according to claim 10, further comprising determining, based on the predictive model, that the subject is not a candidate for therapeutic treatment.

13. The generation of the transcription regulatory factor (TR) information is To generate a transcriptional regulatory network based on transcription factor genes and the aforementioned multiple genes, The method according to any one of claims 1 to 12, comprising determining an enrichment score for each transcription factor based on the transcription regulatory network and the second gene data.

14. The method according to claim 13, wherein generating the transcription factor information further comprises determining one or more predicted transcription factors based on the enrichment score.

15. One or more non-transient computer-readable media for storing processor-executable instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 14.

16. It is a system: A computing device configured to carry out the method described in any one of claims 1 to 14; and A system including a predictive model configured to receive multiple features.

17. It is a device: One or more processors; and A device comprising a memory that stores processor-executable instructions, which, when executed by one or more processors, cause the device to perform the method according to any one of claims 1 to 14.

Images