Our studies demonstrate the potential for predictive and generative artificial intelligence and machine learning approaches to address critical challenges multi-omics, multi-modal biomarker design to support clinical decision-making and enhance the clinical utility of multi-omics tumor profiling. Improving therapy assignment through more robust pharmacogenomic biomarkers, which can improve outcomes and reduce ineffective therapy assignment. These novel biomarkers can be built both to expand labeling for existing drugs and support novel drug development.