KRASAVA-An Expert System for Virtual Screening of KRAS G12D Inhibitors.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• KRAS G12D inhibitors are crucial for cancer therapy.
• Existing quantitative structure-activity relationship (QSAR) models lack applicability domain determination and virtual screening.
• This limits their utility in drug discovery and development.

Purpose of the Study:

• To develop novel QSAR models for KRAS G12D inhibitors.
• To integrate these models into a user-friendly computational framework.
• To identify and validate potential new KRAS G12D inhibitor drug candidates.

Main Methods:

• Employed various molecular descriptors and machine learning algorithms to build regression QSAR models.
• Developed a consensus model with high predictive accuracy (Q²=0.70) and defined applicability domain.
• Integrated the model into the KRASAVA Python framework for activity prediction, rule-based filtering, and molecular docking (GNINA).

Main Results:

• The consensus QSAR model achieved a Q² test value of 0.70 on an external test set.
• The KRASAVA platform successfully predicted inhibitory activity, assessed bioavailability rules, and identified undesirable chemical structures.
• Molecular docking studies validated the proposed inhibitors, showing consistency with QSAR predictions and reference compound MRTX1133 (RMSD 0.76 Å).

Conclusions:

• The developed QSAR models and KRASAVA framework offer a robust tool for KRAS G12D inhibitor discovery.
• The study successfully identified promising novel inhibitor candidates with potential therapeutic applications.
• The findings highlight the synergy between QSAR modeling and molecular docking for efficient drug design.