IQTC Seminar: Darío Estrin
Multi scale quantum-classical (QM-MM) simulation schemes: Basics and Applications
Tuesday November 12th 2024, 12:00h
Multiscale quantum (QM-MM) schemes, which consist of describing one portion of the system with quantum mechanics and the other using classical force fields, are ideal tools to describe chemical processes in complex environments, such as proteins or solutions.
The key issue of these schemes consists in modeling the coupling QM-MM contribution. This interaction is typically computed using a so called electrostatic embedding approach in which the QM density is calculated in the presence of the MM electrical field. In contrast a simpler mechanical embedding scheme in which the QM-MM coupling energy is computed by using a set of point charges on the QM region has also been implemented. We will show selected results obtained using our electrostatic embedding QM-MM implementation called LIO, regarding thiol chemistry in aqueous solution and in proteins as representative examples.
The computational cost of performing QM-MM simulations is mainly determined by the computational cost of the QM calculation and the extent of sampling necessary to describe the system. This makes QM-MM simulations very demanding from a computational point of view. For alleviating this issue, we present in the second part of the talk a novel scheme in which the QM region is described by using the ANI machine learning approach developed in the University of Florida in A. Roitberg’s group. ANI provides results of QM quality at a much lower computational cost and also partial charges. In order to be able to use the ANI approximation to the QM energy, a mechanical embedding scheme is proposed but adding a correction that includes the polarizability of the QM subsystem by the presence of the field generated by the MM subsystem. Our ML-MM approach is validated through simulations of solvation profiles, vibrational spectra, and torsion free energy profiles of small molecules in aqueous environments, as well as in protein-ligand interactions.
You can join the talk via zoom:
https://ub-edu.zoom.us/j/95829290146?pwd=LHUaKQzlcNLaNrpEywuAuLK3auXCTw.1
Meeting ID: 958 2929 0146
Passcode: 859283