| This dissertation is focused on development of novel scoring functions for study of ligand-receptor interactions. It describes two types of scoring functions, and one novel Generalized Born solvation model. The scoring function based on four-body statistical potential targets lead discovery, and the scoring function based on simulation targets lead optimization.; Our laboratory recently embarked on the development of novel four-body statistical potential for ligand receptor interaction using Delaunay tessellation as applied to protein structures. Delaunay tessellation is applied for the first time in the analysis of protein-ligand interaction. Given the full atomic description of protein ligand interface, Delaunay tessellation generates a set of tetrahedra or simplices, which strictly define nearest neighbor atoms in sets of four (tetrahedral quadruplets). Thus, for every combination of quadruplets incorporating ligand and receptor atoms, a likelihood factor (or potential) can be generated. We applied this potential to two datasets used previously published. The results show strong correlation with actual binding affinities.; It is generally difficult to account for the effects of solvation in protein-ligand docking calculations. Still and co-workers developed a Generalized-Born/Surface Area (GB/SA) model (Qiu, D., Shenkin, P. S., Hollinger, F. P.; Still, W. C. J. Phys. Chem. A 1997, 101, 3005--3014) that gives a good balance of computational speed and accuracy in calculations on small molecules. Here, we examine the accuracy of the generalized-Born approach in protein-ligand docking calculations by making a comparison with electrostatic energies from finite difference Poisson-Boltzmann calculations, and then propose three specific refinements to improve the results.; The advance of computer technology makes the time-consuming force field-based scoring functions more practical. Among all force field methods, Linear Interaction Energy (LIE) method maintains a good balance between accuracy and efficiency. The application of continuum solvation model further improved the computational efficiency. We applied Generalized Born solvation model in the molecular dynamic (MD) simulation of HEPT (1-[(2-Hydroxyethoxy)methyl]-6-(phenylthio)thymine) analogs within the active site of HIV reverse transcriptase. We also compared the performance between different sampling methods: minimization, MD simulation with distance dependant dielectric, and MD simulation with GB model. Our result showed better correlation (0.82) than published model (0.77), and confirmed the performance of GB model in the simulation. |
