DR-541 Multi-Cohort Transcriptomic Prediction of Anti-PD-1 Immunotherapy Response and Overall Survival in Skin Cutaneous Melanoma Using Pre-Treatment Bulk RNA-seq

Cutaneous melanoma treated with anti-PD-1 immunotherapy achieves response in only 33–44% of patients, yet no validated pre-treatment biomarker currently guides patient selection. This project develops a machine learning framework that integrates pre-treatment tumour transcriptomic data, specifically pre-specified biological axis scores and literature-curated gene biomarkers, to jointly predict immunotherapy response and overall survival across three melanoma cohorts (N=155). This framework employs low-dimensional, biologically grounded features with rigorous cross-cohort validation. Findings could enable clinicians to identify likely non-responders before treatment initiation, reducing unnecessary toxicity and informing personalised treatment decisions.

Erika Molteni, University College London, United Kingdom