Molecular pharmacogenomics-guided development of novel targeted therapies for drug-resistant human cancers

Abstract

Despite unprecedented therapeutic advances, multiple myeloma remains incurable, with nearly all patients eventually developing drug resistance and disease relapse. This urgent clinical challenge demands innovative approaches to identify and validate novel therapeutic targets in relapsed/refractory disease. This dissertation implements a machine learning-guided preclinical platform for rational drug target identification and validates its predictive power through experimental confirmation of multiple therapeutic candidates for relapsed/refractory multiple myeloma (RRMM). We developed an integrated computational-experimental pipeline combining multi-omics datasets, large-scale cytotoxicity screening, and machine learning algorithms to systematically identify dysregulated pathways in B-cell malignancies. This approach successfully predicted several high-priority therapeutic targets for RRMM, which we subsequently validated through rigorous pharmacological testing. Our experimental validation focused on three key findings. First, we addressed the historically "undruggable" RAS-MAPK pathway, found to be mutated in two-thirds of RRMM patients, by developing CRISPR-engineered Ras-mutant models and screening novel inhibitors across >50 human myeloma cell lines, revealing RAS and RAC1 as viable downstream targets. Second, we identified convergent evidence from both direct omics screening and the SecDrug predictive algorithm identifying BIRC5 (Survivin) as a potent therapeutic target, with MCL-1 co-inhibition producing synergistic pro-apoptotic effects. Third, we traced these effects upstream to DDX5, a transcriptional regulator, and characterized novel analogs with improved pharmacological properties. This work demonstrates that machine learning-guided target identification coupled with systematic experimental validation can successfully predict effective therapies for treatment-resistant cancers. The validated targets, particularly RAS pathway components, BIRC5, and DDX5, represent viable candidates for clinical development in RRMM and potentially other B-cell malignancies, offering new hope for patients who have exhausted current treatment options.