PD-L1 Status Prediction Using RNA Expression Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining PD-L1 status in cancer cells, such as IHC staining, FISH, and RPPA, are time-consuming, require expensive reagents and trained technicians, and are subjective, leading to inaccuracies and discomfort for patients due to tissue sampling needs.
Innovation Solution
A computer-implemented method aligns unlabeled gene expression data with a trained PD-L1 predictive model to predict PD-L1 expression status, generating a clinical decision support information report that includes treatment recommendations for immune checkpoint blockade therapy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If IHC staining, FISH, or RPPA are used to detect PD-L1 status, then treatment-related molecule detection is achieved, but the process requires time, trained technicians, expensive equipment and reagents
Solution Approach 1:
The patent replaces manual IHC staining, FISH, and RPPA procedures with a computational machine learning model that processes RNA expression data. This substitution eliminates the need for complex laboratory equipment, trained technicians, and expensive reagents while maintaining or improving detection accuracy through algorithmic analysis of gene expression patterns associated with PD-L1 status
Solution Approach 2:
The invention creates a computational model that learns from labeled training data containing RNA expression profiles and corresponding PD-L1 status. The trained model then copies this knowledge to predict PD-L1 status in unlabeled samples, replacing the need for direct immunological or molecular detection methods with a data-driven prediction system
2Productivity
If IHC staining is performed by analysts, then PD-L1 status can be determined, but analysts do not have enough time to count all cancer cells leading to inaccurate estimates
Solution Approach 1:
The patent replaces manual cell counting and visual assessment by analysts with an automated machine learning model that processes RNA expression data. This computational approach eliminates time constraints and human fatigue, providing consistent and accurate predictions without requiring analysts to manually examine and count individual cancer cells
Solution Approach 2:
The system enables self-service prediction where the computational model autonomously analyzes RNA expression data and determines PD-L1 status without human intervention. The model independently processes data, applies learned patterns, and generates predictions, eliminating the need for analyst time and subjective judgment
3Quantity of substance
If multiple slices of tumor tissue are collected for IHC, FISH, or RPPA assays, then sufficient sample material is obtained, but patients experience discomfort and inconvenience from repeated biopsies
Solution Approach 1:
The invention uses RNA expression data as a surrogate copy that contains information about PD-L1 status without requiring additional tissue samples. By analyzing computational features from available RNA data, the model predicts PD-L1 status, eliminating the need to collect multiple tissue slices through repeated biopsies and thereby reducing patient discomfort
Solution Approach 2:
The patent makes RNA expression data serve multiple functions: it is used for both general gene expression analysis and specifically for predicting PD-L1 status. This multi-functionality allows the same RNA data to provide treatment-related information without requiring separate tissue sampling procedures, reducing patient burden while maintaining diagnostic capability
4Measurement precision
If IHC, FISH, or RPPA assays are performed, then PD-L1 status is determined, but the cost of reagents, equipment, and trained personnel is expensive
Solution Approach 1:
The patent replaces expensive immunological and molecular assays with a computational model that processes existing RNA expression data. This substitution eliminates costs associated with specialized reagents, equipment maintenance, and trained technician salaries, while maintaining detection accuracy through algorithmic analysis of gene expression patterns
Solution Approach 2:
The invention uses computational algorithms as disposable, easily replicable solutions instead of expensive physical assays. The machine learning model can be trained once and then applied indefinitely to new samples at minimal computational cost, replacing the need for repeated purchases of expensive reagents and equipment usage
Data Source
AI summary
Provided herein are computer-implemented methods of identifying programmed-death ligand 1 (PD-L1) expression status of a subject's sample comprising a cancer cell. In exemplary embodiments, the method comprises receiving an unlabeled expression data set for the subject's sample; aligning the unlabeled expression data set to labeled expression data according to a trained PD-L1 predictive model, wherein the trained PD-L1 predictive model has been trained with a plurality of labeled expression data sets, each labeled expression data set comprising expression data for a sample of a labeled cancer type and a labeled PD-L1 expression status; wherein aligning the unlabeled gene expression data set to labeled expression data according to the trained PD-L1 predictive model identifies PD-L1 expression status for the subject's sample. Further provided are related methods of preparing a clinical decision support information (CDSI) report and methods of determining treatment for a subject. Additionally provided are CDSI reports and computing devices.