Traditional biomarkers often fail to capture the dynamic nature of disease states, highlighting the need for advanced tools to predict patient responses to KRAS therapies. This study evaluates the ResponderID™ KRAS, an RNA-based sequencing classifier that integrates KRAS biology with machine learning to predict therapeutic responses. Re-implementation of the K20 model, which includes gene expression patterns and KRAS mutation status, showed poor reproducibility. In contrast, ResponderID KRAS demonstrated robust predictive capabilities, emphasizing the efficiency of focusing on essential KRAS biology. Our two-feature model showed comparable ROC AUC to more complex models. ResponderID KRAS supports real-time drug development and clinical decisions, promoting personalized medicine in KRAS inhibitor therapies and enhancing the efficacy of multi-drug combinations, making it a valuable tool for cancer treatment.

Published at 5th Annual RAS-Targeted Drug Development Summit 2023.

Joshua Wheeler, Anže Lovše, Klemen Žiberna, Janez Kokošar, Miha Štajdohar, Rafael Rosengarten, Aditya Pai, Daniel Pointing, Luka Ausec, Mark Uhlik

Genialis Inc., Boston MA, United States

Share this story, choose your platform!