Pan-cancer multi-omic integration for identification of clinically aggressive and potentially actionable molecular modules in breast, endometrial, and ovarian cancers.

Abstract

e12610 Background: Pan-cancer analysis enable the detection of molecular alterations that are individually infrequent within single tumor types but become detectable across harmonized multi-tumor datasets. Beyond tumor-type aggregation, an unmet need is to understand how coordinated multi-omic programs differ between clinically aggressive and non-aggressive phenotypes. We conducted a multi-omic integrative study across breast, endometrial, and ovarian cancers (Pan-GYN) to identify molecular programs associated with clinical aggressiveness and to prioritize candidate biomarkers and potentially actionable targets. Methods: Approximately 3,800 publicly available tumor profiles with paired somatic mutations, somatic copy number alterations, and gene expression data were integrated across breast, endometrial, and ovarian cancers. Samples were prefiltered using stringent quality control criteria. Patients were classified into aggressive and non-aggressive phenotypes based on recurrence, metastasis, or death versus absence of these events.Multi-omic integration and module discovery were performed using the ModulOmics algorithm. Molecular modules were calculated within each clinical class across all three tumor types, enabling identification of coordinated molecular programs associated with clinical phenotype. A noise-aware framework was applied, and two complementary strategies were implemented: (A) a descriptive strategy for group-level enrichment and (B) an observational and predictive strategy based on individualized oncogenic risk scores for patient-specific prioritization. Results: Joint pan-cancer module discovery identified 143 molecular modules in the aggressive phenotype. Based on clinical association, 42 modules comprising 28 genes were prioritized as associated with aggressive disease. These included well-established oncogenes and tumor suppressors, as well as underreported candidates. Prioritized modules were enriched for oncogenic processes relevant to tumor progression and clinical outcomes. Importantly, prioritized modules were significantly associated with clinical outcomes, and individualized oncogenic risk scores improved discrimination of aggressive phenotypes. Actionability analyses mapped prioritized genes and modules to drug–gene interaction networks, supporting the identification of candidate pharmacologic modulators and potential therapeutic targets. Conclusions: This pan-cancer, multi-omic framework captures differences between aggressive and non-aggressive phenotypes across breast, endometrial, and ovarian cancers, supporting systems-level approaches to identify convergent oncogenic programs linked to clinical aggressiveness, prioritize biomarkers, and inform network-based therapeutic and drug repositioning strategies.