|Molecular Simulation-Based Descriptors
Co-Investigator: Shekhar Garde
Assistant Professor, Chemical and Biological Engineering, Rensselaer Polytechnic Institute
Simulation-derived Hydration-based descriptors:
Statistical analysis of the dynamic behavior of water distributions
solvating large and small molecules was used to create a set of regional
property descriptors found to generate better ligand binding and QSPR
models than field- or atom-based electrostatic potential, HBDA or
lipophilicity descriptors. Initial data shows that models of molecular
behavior created using these features required fewer descriptors and were
more interpretable by bench chemists than models built using traditional 3D descriptors.
Interactions of proteins and other biomolecules with water influence the
properties of those vicinal water molecules. This effect is especially
pronounced locally, especially in the first hydration shell. We quantify
the hydration shell properties by calculating spatially resolved (i)
average local water density, (ii) water density fluctuations, (iii) local
water orientations, (iv) electron density profile due to water packing and
orientations (i.e., polarization), (v) electrostatic potential on protein
surface induced by the vicinal water structuring, and (vi) dynamics of local water.
This local property information can be projected (Figure 1) on to a
triangulated surface of the molecule of interest and forms input to
descriptor generation algorithms based on surface properties.