Precision medicine workflows for drug repurposing in oncology: DNA polymerase ε and sparc as proof-of-concept targets

Precision oncology requires integrating genomic information with structural and functional data to translate patient-specific variants into mechanistic insight and therapeutic opportunities. A major challenge is bridging the gap between omics-level observations and molecular-level explanations of how mutations alter protein dynamics, enzymatic activity, and drug response. This doctoral work addresses this challenge through two independent projects focused on DNA polymerase ε (Polε) and the matricellular protein SPARC, combining molecular modeling, multi-microsecond molecular dynamics (MD) simulations, pocket communication analysis, ensemble-based virtual screening, and experimental validation. The first project originated from a precision oncology case involving a young colorectal cancer patient carrying the POLE p.F695V variant. This substitution lies within the P-domain of Polε, regulating access to the catalytic site and mediating communication within the Polε:DNA:PCNA complex. MD simulations demonstrated that F695V perturbs inter-domain dynamics and reshapes allosteric pathways linking the DNA-binding groove to PCNA-associated regions. Functional assays confirmed reduced DNA binding affinity and catalytic efficiency, identifying residue 695 as a critical regulatory hotspot. To explore mutation-selective therapeutic strategies, an ensemble structure-based virtual screening campaign targeted a mutation-associated allosteric cavity identified by MLCE analysis. Three dominant conformational clusters served as independent receptor states to account for pocket plasticity. Screening of 8,742 approved or clinical-stage compounds from DrugBank and ChEMBL identified COH-29 as a consensus hit, showing dose-dependent inhibition of the F695V POLE1 N-terminal domain with a stronger effect than wild-type, supporting a mutation-preferential activity profile. The second project focused on SPARC, a matricellular protein involved in extracellular matrix remodeling and tumor progression. Three independent 1 μs atomistic MD simulations identified druggable pockets connected to the collagen-binding Phe-rich region. Ensemble docking of authorized drugs yielded three candidate repurposing compounds. Collectively, this work establishes an integrated computational-experimental strategy to translate genomic variants and cancer targets into pharmacological hypotheses for mutation-informed therapeutic intervention.