Machine Learning Ensemble for Homologous Recombination Deficiency Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting homologous recombination deficiency (HRD) in cancer tissues are limited, as they rely on DNA-based detection of biallelic loss of BRCA1 or BRCA2, missing patients who have not accumulated sufficient genetic lesions, and do not account for other factors contributing to HRD, leading to underidentification of HRD-positive patients who could benefit from PARP inhibitors and platinum-based therapies.
Innovation Solution
The use of machine-learning classifiers trained on RNA and DNA sequencing data from cancerous tissues to predict HRD status, incorporating genome-wide loss of heterozygosity and gene expression levels, allowing for the identification of HRD-positive cancers without biallelic BRCA1 or BRCA2 loss, and integrating data from both sequencing types in an ensemble model for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If DNA-based detection of biallelic loss of BRCA1 or BRCA2 is used to detect HRD, then the detection method is simple and well-established, but many HRD-positive patients are missed because they have not accumulated sufficient genetic lesions or have HRD caused by other factors
Solution Approach 1:
The patent combines multiple detection approaches including DNA sequencing, RNA sequencing, and machine learning analysis into an integrated HRD detection system. This merging of methods allows comprehensive detection of various HRD mechanisms beyond just biallelic BRCA loss, thereby improving measurement precision without requiring entirely new complex technology but rather synthesizing existing tools.
Solution Approach 2:
The patent introduces machine learning algorithms as an intermediary layer that processes and integrates data from DNA and RNA sequencing to predict HRD status. This intermediary computational approach enables the system to interpret complex genomic and transcriptomic patterns that indicate HRD, improving detection accuracy while managing complexity through automated analysis.
2Ease of operation
If only biallelic BRCA1 or BRCA2 loss is considered for HRD detection, then the diagnostic criteria are clear and easy to apply, but other important causes of HRD are not detected
Solution Approach 1:
The patent expands the diagnostic parameters from solely genetic (biallelic BRCA loss) to include transcriptomic parameters (gene expression profiles) and computational predictions (machine learning HRD scores). This parameter expansion allows detection of HRD through multiple biological mechanisms while maintaining clear operational criteria through standardized scoring systems and thresholds.
3Measurement precision
If comprehensive genomic and transcriptomic analysis is performed to improve HRD detection, then more HRD-positive patients are identified, but the complexity and cost of testing increases
Solution Approach 1:
The patent performs preliminary filtering and prioritization through machine learning models that assess the likelihood of HRD based on initial genomic and transcriptomic data. This preliminary action allows the system to focus computational resources and clinical attention on samples most likely to be HRD-positive, improving sensitivity while managing complexity through staged analysis rather than exhaustive evaluation of all possible markers.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Methods, systems, and software are provided for an ensemble model trained to distinguish between cancers with homologous recombination pathway deficiencies (HRD positive cancers) and cancers without homologous recombination pathway deficiencies (HRD negative cancers) based on nucleic acid sequencing data, e.g., both RNA and DNA sequencing data, generated from a cancerous tissue sample of the subject.