DR-477 AURORA: Forecasting Treatment Resistance and Evolutionary Trajectories with Multi-Modal Data

AURORA discovers novel biomarkers that predict treatment resistance and guide therapy selection in metastatic cancers. Leveraging Hartwig’s pan-cancer whole-genome sequencing data with treatment outcomes, we identify genomic signatures—including complex mutational patterns, structural variants, and clonal architectures—that presage resistance mechanisms. Our machine learning approaches extract interpretable biomarker panels that capture how multiple genomic alterations synergistically determine treatment response. Unlike single-gene markers, we discover combinatorial biomarkers specific to metastatic disease that generalize across cancer types. This addresses the critical challenge that over 90% of cancer deaths result from metastatic disease, where current biomarkers inadequately predict resistance. Our discoveries will enable precision medicine for metastatic cancer patients.

Regina Barzilay, Massachusetts Institute of Technology, United States of America