Speaker
Description
Molecular dynamics simulations of protein–ligand systems are central to computer-aided drug discovery, but their reliability rests on the underlying force field. Metalloproteins are a particularly hard case: classical force fields describe metal coordination poorly, and reliable parameters are often unavailable. Quantum mechanics/molecular mechanics (QM/MM) molecular dynamics treats such sites accurately, yet reaches only picosecond time scales, far short of those relevant to ligand binding and drug design. Force matching (FM) addresses this gap, turning short QM/MM trajectories into system-specific classical force fields that retain reference accuracy at classical cost. In this talk I will present my work on force matching within the MiMiC multiscale framework. I will first describe MiMiCPy-FM, an automated implementation of QM/MM force matching that derives atomic charges and bonded parameters from MiMiC (CPMD/GROMACS) simulations and writes GROMACS topologies ready for classical MD. Its application to the Mg-based enzyme IDH1, a drug target in brain cancer, shows a seamless transition from picosecond QM/MM to microsecond classical dynamics that preserves the QM active-site structure. Fixed-charge force fields, however, cannot respond to a changing electrostatic environment, a limitation for metal sites and charged ligands. I will therefore present work extending force matching to polarizable force fields, coupling it to AMOEBA with Tinker-HP, validated on acetone in water.