Accelerating reliable multiscale quantum refinement of protein–drug systems enabled by machine learning

• DOI: 10.1038/s41467-024-48453-4
• Publication Date: 2024-05-16T19:01:53Z
• Authors: Zeyin Yan; Dacong Wei; Xin Li; Lung Wa Chung
• Keywords: 0301 basic medicine; SARS-CoV-2; Science; Q; Molecular Dynamics Simulation; Antiviral Agents; Article; COVID-19 Drug Treatment; Machine Learning; 03 medical and health sciences; Quantum Theory; Humans; Coronavirus 3C Proteases

Biomacromolecule structures are essential for drug development and biocatalysis. Quantum refinement (QR) methods, which employ reliable quantum mechanics (QM) methods in crystallographic refinement, showed promise improving the structural quality or even correcting structure of biomacromolecules. However, vast computational costs complex mechanics/molecular (QM/MM) setups limit QR applications. Here we incorporate robust machine learning potentials (MLPs) multiscale ONIOM(QM:MM) schemes to describe core parts (e.g., drugs/inhibitors), replacing expensive QM method. Additionally, two levels MLPs combined first time overcome MLP limitations. Our unique MLPs+ONIOM-based achieve QM-level accuracy with significantly higher efficiency. Furthermore, our refinements provide evidence existence bonded nonbonded forms Food Drug Administration (FDA)-approved nirmatrelvir one SARS-CoV-2 main protease structure. This study highlights that powerful accelerate QRs protein-drug complexes, promote broader applications more atomistic insights into development.