Machine Learning Classifiers for HER2-Low Breast Cancer Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying HER2 status in breast cancer, such as immunohistochemistry (IHC) and the PAM50 transcriptomic system, face reproducibility issues and do not adequately account for low HER2-expressing samples, which is particularly problematic given the recent approval of targeted therapies for HER2-low BCs.
Innovation Solution
A method using trained machine learning classifiers to identify HER2-low breast cancer from RNA expression data of tumor samples, involving multiple sets of genes and classifiers to determine molecular subtypes, including Basal, HER2-high, and HER2-low subtypes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If immunohistochemistry (IHC) analysis is used to evaluate HER2 status, then the assessment can be performed, but reproducibility issues arise
Solution Approach 1:
The patent replaces the manual immunohistochemistry (IHC) analysis method with an automated machine learning-based computational system. The ML model processes RNA expression data to determine HER2 status, eliminating the reproducibility issues associated with manual IHC interpretation while maintaining measurement precision through consistent algorithmic evaluation.
Solution Approach 2:
The patent introduces RNA expression data as an intermediary medium between the tumor sample and the final HER2 status determination. Instead of directly analyzing protein expression through IHC, the system uses RNA expression levels processed by machine learning algorithms to infer HER2 status, thereby improving reproducibility while preserving diagnostic accuracy.
2Adaptability or versatility
If the PAM50 transcriptomic system is used for molecular subtyping, then subtyping can be performed, but low HER2-expressing samples are not adequately accounted for
Solution Approach 1:
The patent segments the HER2 expression spectrum into distinct categories (HER2-high, HER2-low, and HER2-negative) using machine learning classification. This segmentation allows the system to specifically identify and characterize HER2-low expressing samples that were previously not adequately accounted for in traditional PAM50 subtyping, while maintaining precision through targeted classification algorithms.
Solution Approach 2:
The patent changes the parameter space by using machine learning models that can detect subtle expression patterns across multiple genes, rather than relying on fixed thresholds. This allows for precise identification of HER2-low subtypes by capturing continuous variations in gene expression that traditional binary classification systems miss.
3Measurement precision
If multiple machine learning classifiers are used to determine molecular subtypes, then accurate identification of HER2-low breast cancer can be achieved, but the system complexity increases
Solution Approach 1:
The patent divides the classification task into multiple specialized machine learning classifiers, each trained to identify specific molecular subtypes (Basal, HER2-high, HER2-low, Luminal). This segmentation allows each classifier to focus on specific gene expression patterns, improving overall accuracy while organizing system complexity into manageable, specialized components rather than one monolithic complex system.
Data Source
AI summary
Aspects of the disclosure relate to methods, systems, and computer-readable storage media, which are useful for characterizing subjects having certain cancers, for example breast cancer. The disclosure is based, in part, on methods for determining the molecular breast cancer (BC) type of a subject and identifying the subject's prognosis and/or one or more therapeutic agents for treating the subject based upon the molecular BC type determination.