Development of Computational Methods for Modeling of Solvation Effects and Protein Flexibility in Protein-Ligand Docking
Protein-ligand docking programs are widely used tools for computer-aided drug design. Many docking programs have been developed. However two major challenges greatly limit the accuracy of the docking results: 1) inclusion of solvation effect caused by the active site water molecules; and 2) consideration of the protein flexibility. The overall goal of this thesis work is to develop a docking program to address these two challenges. We developed a new pharmacophore-based docking program, named PharmDock. It samples the ligand binding poses by enumerating all possible multiple-points matches between the pre-generated protein and ligand pharmacophores. The sampled binding poses are then ranked using a pharmacophore-based ranking scheme. A set of top ranked binding poses are locally optimized within the protein binding site to obtain the final ligand binding pose and binding energy. We tested PharmDock's performance on predicting the ligand binding poses and binding energies with a large set of protein-ligand complexes structures. The results showed that PharmDock is among the best compared to several widely used docking programs in both pose prediction and binding energy estimation. To account for the solvation effect caused by the active site water molecules, we developed a hydration site analysis program that predicts the potential hydration sites positions and their desolvation energy if being replaced upon ligand binding. We introduced the Hydration-Site-Restricted Pharmacophore (HSRP) models into PharmDock to select protein pharmacophores important for ligand binding. We tested this model on five different protein structures and revealed that the use of HSRP models is an efficient way of including solvation effects into protein-ligand docking and virtual screening. To include the protein flexibility, we introduced the Ligand Model Concept (Limoc) into PharmDock. Limoc represents a large ensemble of different chemical species binding to the same target protein. Molecular dynamics simulation with Limoc is able to sample the protein conformations that are most relevant for binding of structurally diverse ligands. We developed an efficient method, named LCS-MC, to combine Monte Carlo sampling in PharmDock with the Limoc sampled ensemble for ligand pose optimization and scoring. We applied LCS-MC in estimating the protein and ligand entropy upon binding for a set of protein-ligand complexes. We showed that the entropy estimations of LCS-MC correlate quite well with the entropies derived from extensive MD simulations.
Lill, Purdue University.
Pharmacy sciences|Biophysics|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our