Quantum Chemistry for Drug Design: From Theory to Applications
September 8, 2025 – September 10, 2025
Location: Prague, Czech Republic

Oral Presentation Title: Advancing CADD and AI/ML campaigns through QM/MM-based X-ray & Cryo-EM refinement, tautomer analysis, and ligand sampling

Speaker: Lance M. Westerhoff

Abstract:
Accurate protein:ligand structure determination is critical for structure-based drug discovery (SBDD), computer-aided drug design (CADD), and free-energy methods (e.g., MovableType, FEP). Precise modeling of protein:ligand complexes—including protonation states and explicit solvent—underpins workflows such as docking, thermodynamic calculations, lead optimization, and AI/ML predictions. However, traditional X-ray and Cryo-EM refinement methods rely on geometric restraints that can neglect key interactions (hydrogen bonding, polarization, charge transfer). These omissions introduce structural inaccuracies often addressed post hoc with molecular mechanics (MM) or quantum mechanics (QM) corrections.

To overcome this, QuantumBio has integrated the DivCon semiempirical quantum mechanics (SE-QM) engine into crystallographic refinement to enable automated, density-driven structure preparation (protonation), model completion (gap and truncation sampling), and QM/MM refinement. This approach improves agreement with experimental data while yielding chemically accurate models. It reduces ligand strain, clarifies protein:ligand interactions, and supports the development of high-quality AI/ML training datasets.

Beyond refinement, these methods resolve essential biochemical features, including tautomeric and protomeric states, chiral centers, rotamer conformations, and solvation effects. With XModeScore, SE-QM refinement determines protonation states and stereoisomers even at resolutions where experimental methods struggle.

This talk will explore how density-driven refinement, tautomer analysis, and ligand sampling enhance SBDD and CADD workflows. By bridging experimental and computational techniques, these methods produce structurally precise, chemically robust models that advance AI/ML predictive power and drive more effective drug discovery campaigns.