A Bayesian Framework for Discovering Novel Cancer Drug Combinations via Mechanistic Pathways in the Absence of primary trials, with Statistical and Histopathological Validation.

Combinational therapy is the cornerstone of modern oncology, yet the process of selecting which drug combinations should advance to clinical trials remains largely subjective and resource-constrained. This study presents a Bayesian probabilistic framework designed to model cancer mechanistic pathways,specifically angiogenesis,to identify optimal drug combinations using only existing randomized control trial (RCT) data. The model maps each drug to its biomarker target and simulates combinations using a Bayesian network, optimizing for survival and biomarker suppression while minimizing resistance and adverse effects. Dimensionality reduction via Partial Least Squares Regression (PLSR), followed by survival analysis and statistical validation in SPSS, confirms predicted synergies. Ex vivo histopathological validation further supports the translational relevance of the predictions. A cost-benefit analysis in glioblastoma multiforme illustrates how this approach dramatically reduces time and expense compared to traditional clinical trial pipelines. This framework provides a data-driven decision-support tool to prioritize promising combinations for future clinical testing, with potential applicability across 23 cancers where angiogenesis plays a critical role.