Lung Cancer Mutation Profiles for CPI Resistance Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current biomarkers for predicting immune checkpoint inhibitor (CPI) therapy response in cancer, such as TMB and PDL-1 levels, provide only moderate predictive value and do not offer mechanistic insights into resistance mechanisms, hindering the development of targeted therapeutics and patient selection.
Innovation Solution
A computer-implemented method utilizing gene mutation profiles, specifically analyzing the presence or absence of mutations in genes like NF1, STK11, TSC2, BRCA2, BRAF, STAG2, U2AF1, BRIP1, PDGFRA, CTNNA1, PDK1, and FGF10, to classify a subject's mutation profile as matching a response or resistance signature, using machine learning classifiers trained on datasets like the Flatiron Health-Foundation Medicine NSCLC clinico-genomic database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current biomarkers (TMB, PDL-1 levels) are used for predicting CPI therapy response, then patient selection can be performed, but the predictive value is only moderate and mechanistic insights are lacking
Solution Approach 1:
The patent segments the monolithic TMB metric into multiple discrete gene mutation profiles (e.g., high TMB vs. low TMB categories, specific gene mutations like STK11, KEAP1, TP53). This segmentation allows for more granular prediction of CPI response while maintaining interpretability of which specific mutations drive resistance or sensitivity, thereby improving predictive value without losing mechanistic information.
Solution Approach 2:
The patent changes the parameter from a single continuous TMB value to multiple categorical parameters including TMB strata (high/low), specific gene mutation statuses (STK11, KEAP1, TP53, etc.), and their combinations. This parameter transformation enables more precise prediction by capturing non-linear relationships between mutations and CPI response that a single TMB value cannot represent.
2Loss of information
If extensive omics and computational approaches are used to identify predictive biomarkers, then understanding of CPI resistance mechanisms is improved, but the biomarkers do not generalize well
Solution Approach 1:
The patent develops a multi-functional biomarker system that evaluates multiple gene mutations (STK11, KEAP1, TP53, PBRM1, Pten, etc.) simultaneously, where each marker can function independently or in combination. This universal approach across multiple markers enables the model to generalize better to different patient populations and tumor types while maintaining mechanistic understanding of individual resistance pathways.
Solution Approach 2:
The patent creates a composite biomarker profile that combines multiple genetic markers (TMB, STK11 mutations, KEAP1 mutations, TP53 mutations, PBRM1 alterations, Pten status) into an integrated prediction model. This composite approach leverages the strengths of each individual marker while compensating for their individual limitations, resulting in improved generalizability across diverse cancer types and patient populations.
3Loss of information
If gene expression signatures are used to predict CPI response, then biological understanding is increased, but the signatures do not generalize to different cancer types
Solution Approach 1:
The patent transitions from continuous gene expression levels to categorical mutation presence/absence parameters for key genes (STK11, KEAP1, TP53, etc.). This parameter discretization creates more robust biomarkers that are less sensitive to technical variations in gene expression measurement across different platforms and cancer types, thereby improving generalizability while preserving biological interpretability.
Data Source
AI summary
The present invention provides a computer-implemented method for predicting the treatment response of a subject having a lung cancer to an immune checkpoint inhibitor (CPI) therapy, the method comprising: providing a mutation profile of the subject, said profile comprising the presence or absence of cancer-specific mutations at one or more locations in at least five genes selected from the group consisting of: NF1, STK11, TSC2, BRCA2, BRAF, STAG2, U2AF1, BRIP1, PDGFRA, CTNNA1, PDK1, FGF10, and FLT1; analysing the mutation profile to classify the profile as matching the mutation profile of a response signature or a resistance signature, wherein the subject is predicted to be likely to respond to the CPI therapy if the mutation profile for the subject is classified as matching the mutation profile of the response signature and is predicted to be likely not to respond to the CPI therapy if the mutation profile for the subject is classified as matching the mutation profile of the resistance signature. Also provided are related methods and systems for predicting the treatment response of a subject having a lung cancer to an immune checkpoint inhibitor (CPI) therapy.