Pan-Cancer Biomarker Analysis from the Cancer Dependency Map: A Blueprint for Precision Oncology

The diversity of therapeutic modalities and targets continue to grow, and the number of cancer patients eligible for targeted therapy has expanded accordingly. Despite advances, response rates to individual targeted therapy drugs remain variable, rarely achieving uniform tumor shrinkage across patients. Biomarkers are crucial for identifying patients likely to respond to therapy while sparing non-responders from toxicity and guiding them toward alternative treatments. We conducted a multi-omic analysis of biomarker-dependency relationships across 1,150 cancer molecular profiles to identify novel biomarkers for patient stratification. We first validated the Cancer Dependency Maps sensitivity and specificity in predicting therapeutic windows for targeted therapies. Next, we identified predictive biomarkers for single- and multi-gene dependencies, assessing their selectivity within biomarker-defined populations. Protein abundance analysis revealed potential immunohistochemistry (IHC) biomarkers for clinically deployed therapeutic compounds. We performed an analysis of lineage-enriched dependencies, highlighting new opportunities for target validation and drug development. Finally, we screen for associations between hotspot mutations, damaging mutations, and protein abundance, providing insights for developing heterobifunctional small molecules for induced proximity and protein degradation. These findings advance our understanding of cancer dependencies and inform biomarker-driven strategies to optimize therapeutic outcomes.