Characterizing ligand-mediated conformational dynamics and state transitions of KRAS-G12C through machine learning and molecular dynamics simulations.

Abstract

KRAS mutations are major drivers of cancer initiation and progression, with KRAS G12C representing an important therapeutic target. In the present study, an integrated computational approach comprising machine learning-based virtual screening, molecular docking, molecular dynamics simulations, and Markov state modeling was utilized to identify potential allosteric inhibitors targeting the switch-II pocket of KRAS G12C. Supervised ML models were trained using a curated ChEMBL dataset, with the Random Forest classifier achieving the highest predictive performance for activity classification. The best-ranked compounds were evaluated using molecular docking. Triplicate 500 ns MD simulations demonstrated that complexes 5 and 115 maintained stable binding, exhibited reduced residue fluctuations, and formed persistent hydrogen-bond interactions. Our analysis demonstrated that the ligands strongly interacted with the key residues and reduced the flexibility of KRAS. Furthermore, MSM and tICA analyses showed that compound 115 stabilizes the coupled dynamics of the switch-I and switch-II regions through interactions with Thr36, Ile37, Glu38, Met68, Gln71, and Tyr72, promoting a highly populated inactive metastable state. Overall, compound 115 represents a promising lead scaffold for KRAS G12C inhibition and provides mechanistic insights for the rational design of next-generation anticancer therapeutics.

Abstract as supplied by the indexing source. © the authors / publisher.