Co-Investigators:
Shekhar Garde

Assistant Professor, Chemical and Biological Engineering, Rensselaer Polytechnic Institute

Angel Garcia

Professor of Physics & Senior Constellation Chaired Professor in Biocomputation and Bioinformatics, Rensselaer Polytechnic Institute

Background

For broader applications that involve studies of hydration of libraries of several hundreds (or even thousands) of molecules of bio, pharma, or health interest, we have developed an efficient potentials-of-mean-force expansion (PMF) based method, employing a library of lower-order correlation functions derived from explicit simulations to predict the average equilibrium density and the orientation profile of water in the space surrounding biomolecules or ligands. The method efficiently approximates the effective free energy of interaction of a biomolecule with surrounding water in terms of two-particle and multi-particle potentials-of-mean-force between constituent sites and water molecules. We have previously shown that a truncation of the expansion (Figure 2) at the three-particle level provides sufficient accuracy for a variety of applications of interest. Significant validation work has been performed by exhaustive comparisons of PMF results to those from detailed protein MD simulations.

PMF expansion uses pre-calculated libraries of hydration two- and three- particle correlation functions between protein constituent sites and water molecules. Only simplistic description of proteins (described by three sites - hydrophobic, polar and ionic) have been employed to date. In our new approach, we include 11 different constituent protein sites that represent varies sizes and charge densities, so as to describe the protein chemistry with much higher fidelity. Clustering algorithms were applied in the force-field parameter space to obtain these different sites. We are currently generating a large library of two and three particle hydration correlations for these sites (requiring over 2000 different MD simulations, and over 100 microseconds of simulation data). The database needs to be calculated only once and can be extended as necessary with inclusion of new canonical sites.