Systems and methods for detecting cellular pathway dysregulation in cancer specimens

Abstract

Disclosed herein are systems, methods, and compositions useful for determining cellular pathway disruption comprising the use of RNA expression level information. This determined level of disruption can assist in the identification of genetic variants that alter pathway activity, to correlate these variants with disease state and disease progression, and to identify those therapeutics most likely to be effective and which should be avoided.

1 . A method for generating a pathway disruption score based on transcriptome data from a patient tumor sample to detect dysregulation in a cellular pathway in the sample, the method comprising:
 (a) obtaining transcriptome data comprising whole transcriptome sequence data;   (b) providing the transcriptome data, in the form of a transcriptome value set comprising an identifier and an expression level for one or more genes in the transcriptome data, to one or more trained pathway disruption engines, the pathway disruption engines comprising a plurality of trained machine learning models;
 wherein the plurality of trained machine learning models comprises;
 i) at least one pathway level machine learning model trained to detect dysregulation in the input transcriptome value set associated with the cellular pathway; 
 ii) at least one module level machine learning model trained to detect dysregulation in the input transcriptome value set associated with the module; 
 iii) at least one gene level machine learning model trained to detect dysregulation in the input transcriptome value set associated with the gene; and 
 iv) at least one variant level machine learning model trained to detect dysregulation in the input transcriptome value set associated with the variant, or associated with a variant of unknown significance; 
 
   (c) selecting at least one trained machine learning model from (i)-(iii) and selecting at least one trained machine learning model (iv) from the at least one pathway disruption engine to apply to the transcriptome value set;
 wherein at least one selected variant level machine learning model is trained to detect dysregulation associated with a variant of unknown significance; 
   (d) applying the selected machine learning models to the transcriptome value set, to generate at least one pathway disruption score indicative of dysregulation in the cellular pathway.   
     
     
         2 . The method of  claim 1 ,
 wherein each pathway level trained machine learning model is trained based on training data comprising a plurality of positive control specimens and a plurality of negative control specimens, wherein each positive control specimen comprises genetic data, the positive control genetic data comprising at least one detectable, pathogenic variant in at least one gene included in the cellular pathway, wherein each negative control specimen comprises genetic data, the negative control genetic data comprising no pathogenic variants in any gene included in the cellular pathway;   wherein each module level trained machine learning model is trained based on training data comprising a plurality of positive control specimens and a plurality of negative control specimens, wherein each positive control specimen comprises genetic data, the positive control genetic data comprising at least one detectable, pathogenic variant in at least one gene included in the module, wherein each negative control specimen comprises genetic data, the negative control genetic data comprising no pathogenic variants in any gene included in the module;   wherein each gene level trained machine learning model is trained based on training data comprising a plurality of positive control specimens and a plurality of negative control specimens, wherein each positive control specimen comprises genetic data, the positive control genetic data comprising at least one detectable, pathogenic variant in the gene, wherein each negative control specimen comprises genetic data, the negative control genetic data comprising no pathogenic variants in the gene;   wherein each variant level trained machine learning model is trained based on training data comprising a plurality of positive control specimens and a plurality of negative control specimens, wherein each positive control specimen comprises genetic data, the positive control genetic data comprising the variant, wherein each negative control specimen comprises genetic data, the negative control genetic data comprising no pathogenic variants.   
     
     
         3 . The method of  claim 2 , wherein at least one trained pathway level model, at least one trained module level module, at least one trained gene level model, or at least one trained variant level model is trained based on training data comprising a plurality of positive control specimens and a plurality of negative control specimens, wherein the training comprises:
 calculating a plurality of differential metric values between the positive control specimens and the negative control specimens, each differential metric value being associated with at least one gene included in the pathway or the module, or associated with the gene or the variant, and selecting the gene(s) with differential metric score at or below a threshold value for inclusion in the model.   
     
     
         4 . The method of  claim 2 , wherein at least a portion of the positive control genetic data and the negative control genetic data comprises DNA data. 
     
     
         5 . The method of  claim 2 , wherein at least a portion of the positive control genetic data and the negative control genetic data comprises RNA data. 
     
     
         6 . The method of  claim 5 , wherein the RNA data comprises transcriptome data. 
     
     
         7 . The method of  claim 5 , wherein the detectable pathogenic variant comprises an RNA expression level. 
     
     
         8 . The method of  claim 5 , wherein the negative control RNA data comprises no detectable variation in expression level when compared to one or more wild-type samples for the expressed RNA. 
     
     
         9 . The method of  claim 8 , wherein the selected variant machine learning model is trained to detect dysregulation associated with the variant of unknown significance based on the expression level data. 
     
     
         10 . The method of  claim 9 , comprising:
 generating a pathway disruption report based on the at least one pathway disruption score; and presenting the pathway disruption report to at least one of a display or a memory.   
     
     
         11 . The method of  claim 10 , wherein the pathway disruption report comprises information associated with the at least one pathway disruption score, the information comprising at least one of: a) potential causative mutations; b) detection of dysregulation associated with one or more variants of unknown significance; c) one or more therapies recommended based on the pathway disruption score; d) a suggestion that an organoid be monitored after exposure to a treatment based on the pathway disruption score; e) matching at least one clinical trial to a patient associated with the specimen based on the pathway disruption score; and f) reference medical literature. 
     
     
         12 . The method of  claim 11 , where the recommended therapy of (c) or the clinical trial of (e) is not traditionally recommended or matched for the potential causative mutations associated with the pathway disruption score, and/or for the variant of unknown significance associated with the pathway disruption score. 
     
     
         13 . The method of  claim 1 , wherein the cellular pathway comprises 1 to 5 genes, 6 to 10 genes, 10 to 20 genes, or 20 to 100 genes. 
     
     
         14 . The method of  claim 1 , comprising:
 generating a pathway disruption report including a depiction of the cellular pathway, the depiction comprising a number of modules included in the cellular pathway, an indication of dysregulation in at least one of the modules included in the cellular pathway associated with the variant; and   presenting the pathway disruption report to at least one of a display or a memory.   
     
     
         15 . The method of  claim 1 , wherein the at least one trained pathway disruption engine comprises a rat sarcoma/receptor tyrosine kinase (RAS/RTK) pathway level machine learning model. 
     
     
         16 . The method of  claim 1 , wherein the at least one trained pathway disruption engine comprises a phosphatidylinositol-3kinase (PI3K) pathway level machine learning model. 
     
     
         17 . The method of  claim 1 , wherein the at least one trained pathway disruption engines comprises a custom pathway level machine learning model. 
     
     
         18 . The method of  claim 1 , wherein each trained model outputs a model score to the trained pathway disruption engine, and wherein the trained pathway disruption engine generates a pathway disruption score indicative of dysregulation in the cellular pathway. 
     
     
         19 . The method of  claim 1 , wherein the models are selected by the user.