Identifying microRNAs associated with tumor immunotherapy response using an interpretable machine learning model

Area of Science:

• Oncology
• Immunology
• Genetics

Background:

• Predicting patient response to tumor immunotherapy, specifically immune checkpoint blockade (ICB), is critical for optimizing treatment efficacy and minimizing adverse effects.
• Current methods for preselecting patients likely to benefit from ICB therapy remain a significant challenge in clinical oncology.

Purpose of the Study:

• To investigate the potential of microRNAs (miRNAs) as predictive biomarkers for patient response to tumor immune checkpoint blockade (ICB) therapy.
• To develop and validate a machine learning model utilizing miRNA expression profiles for predicting ICB treatment outcomes across diverse cancer types.

Main Methods:

• Construction of random forest models to predict ICB therapy response based on miRNA expression data from 19 cancer types.
• Application of SHapley Additive exPlanations (SHAP) to interpret model predictions and identify the contribution of individual miRNAs.
• Analysis of pathways targeted by high-importance miRNAs to understand their role in immune response.

Main Results:

• Machine learning models using a limited set of high-importance miRNAs achieved predictive performance comparable to models using the entire miRNA expression profile.
• SHAP analysis identified specific miRNAs crucial for predicting ICB response.
• Genes targeted by these key miRNAs were significantly associated with tumor-related and immune-related biological pathways.

Conclusions:

• MicroRNA expression data holds significant potential for predicting patient responses to tumor immunotherapy.
• Selected informative miRNAs can serve as reliable biomarkers for assessing immunotherapy efficacy, advancing the understanding of underlying therapeutic mechanisms.