Israel Cabeza de Vaca, Ph.D., from Yale University, is a candidate for a faculty position in the Department of Chemistry, part of the cluster hiring initiative in the BioInspired Institute.
Abstract: Computational drug design methods are widely used in industry and academia to reduce time and expenses in drug discovery. Statistical mechanics approaches are the most accurate methods to determine the thermodynamic properties of molecular systems such as free energies. The accuracy of the predictions relies on two main factors: accuracy of the energy estimation and proper sampling. In general, molecular mechanics force fields are able to produce satisfactory free energy estimation if a large enough subset of the configurational space is sampled. However, in large biomolecular systems, the vastness of this space may produce insufficient sampling, thereby generating an inexact representation of the thermodynamic properties. Traditionally, molecular dynamics (MD) has been the most frequently used method to perform the sampling. The main MD limitation is the small time step needed (1-2 fs), which requires a significant amount of sampling to jump between different local energy minima. In this presentation, I will introduce a Monte Carlo (MC) method that overcomes this issue by adapting the attempted movement sizes to perform a more extensive exploration of alternative local energy minima. The MC algorithm uses the free energy perturbation (FEP) method to estimate protein-ligand binding free energies through alchemical ligand transformations. In addition, the recent creation FEP method will be presented to reduce the sampling demand in free energy estimations. This method takes advantage of the asymmetry between forward and backward directions in any alchemical transformation to estimate the right free energy difference with significantly less sampling. Finally, I will discuss the LigParGen server which provides OPLS-AA force field parameters in combination with CM1A/CM1A-LBCC charge models for organics molecules. Both charge models haven been optimized to reproduce accurate hydration free energies and heat of vaporization and are expected to be excellent charge models for computational drug design.
This event was published on January 7, 2020.