Explainable machine learning identifies features and thresholds predictive of immunotherapy response
Machine learning identifies specific immune markers and thresholds that predict immunotherapy response in melanoma, guiding next-generation precision treatment studies.
Integrating clinical, DNA, and RNA sequencing data from 229 melanoma samples, this study developed an explainable random forest model using SHAP for immunotherapy response prediction, identifying LAG3 expression, mutation burden features, and immune cell abundance as key drivers. SHAP-inferred numerical thresholds for these features may guide future prospective biomarker studies for precision immunotherapy selection.
What the study was
- Study design
- Retrospective ML model development + independent cohort validation
- Population
- Melanoma patients; train n=138, test n=53, additional stable disease n=15, non-cutaneous n=23
- Sample size
- 229
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- Scientific Reports
Why it surfaced
Multi-omics ML for immunotherapy response in melanoma; LAG3 as key feature is novel angle. Test cohort validation is independent (n=53) but small. QIMR Berghofer + Cambridge collaboration. Scored 6.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.