A Novel Method for Guiding Protein-Ligand Docking with QSAR-Derived Pharmacophore Maps

Currently, QSAR and computational ligand docking studies are valuable but independently used tools for drug design. Data from Pharmacophore maps produced by tools such as COMFA are typically compared to the results of docking simulations by hand in a qualitative manner. RosettaLigand has been previously successful at predicting binding poses with high resolution resolution (Kaufmann, et. al, Proteins, 2009). We are developing RosettaHTS, an extension to RosettaLigand which will integrate these two methods by using information from QSAR derived pharmacophore maps to guide the low resolution phase of ligand docking. Pharmacophore maps are generated using BCL::PharmMap (unpublished), and contain information about hydrogen bonding, steric bulk, and polarizability. Discrete cartesian grids describing the hydrogen bonding ability, steric bulk and polarizability of the ligand binding site are overlaid on the protein structure, and these grids are used to score the initial placement of the ligand prior to h fine grained docking. As the scoring grids are precomputed, ligand scoring is extremely fast, and thorough Monte Carlo sampling of the ligand binding site can be rapidly performed before to fine grained ligand docking.This rapid initial sampling makes it possible to predict accurate binding poses with a significantly smaller amount of fine grained sampling, decreasing the amount of CPU time necessary to predict a single binding interaction, and increasing the practicality of structure based virtual High Throughput Screening (vHTS). The integration of structure based and ligand based vHTS techniques allows the full range of pharmacological information surrounding a target and drug scaffold to be considered in a single approach. This technique can be used to rapidly develop small focused libraries for High Throughput Screening, increasing the hit rate and decreasing the number of compounds that need to be purchased for testing.