In Silico Identification of Selective KRAS G12D Inhibitor via Machine Learning-Based Molecular Docking Combined with Molecular Dynamics Simulation

KRAS G12D mutation is prevalent in various cancers and is associated with poor prognosis. This study aimed to identify potential drug candidates targeting KRAS G12D using combined machine learning, virtual screening, molecular docking, and molecular dynamics (MD) simulations. The training and test sets are constructed based on a selection of inhibitors targeting the KRAS G12D mutant from the ChEMBL library. A random forest machine learning algorithm is developed to predict potential KRAS G12D binders. Molecular docking and the MM/PBSA binding energy are used to identify the lead compounds. The compound NPC489264 is identified as the top candidate, exhibiting favorable docking energy for the KRAS G12D mutant (-13.16 kcal mol-1). A hydrogen bond between the mutated Asp12 residue in the KRAS G12D mutant and NPC489264 is found to be a key interaction between these 2 molecules. MD simulations and MM/PBSA analysis revealed the strong binding affinity of NPC489264 to the G12D mutant (-5.49 kcal mol-1) compared to the wild type (10.17 kcal mol-1). These findings suggest that NPC489264 is a promising lead compound for further development of KRAS G12D-targeted cancer therapies. An integrated computational approach that combines ML-based quantitative structure-activity relationship (QSAR) modeling, structure-based virtual screening, and molecular dynamics simulations to identify and characterize promising lead compound inhibitors against the KRAS G12D mutation identify compound NPC489264 as the top candidate for the KRAS G12D mutant inhibitor. image