Binding affinity plays an important role in drug design. Accurate and fast prediction of binding free energies remains a major challenge for... Show moreBinding affinity plays an important role in drug design. Accurate and fast prediction of binding free energies remains a major challenge for structure-based calculation. We have developed a fast free energy calculation program AlGDock and applied it to different systems. In this thesis, I will first demonstrate the feasibility of estimating protein-ligand binding free energies using multiple rigid receptor configurations on T4 lysozyme. Based on 576 snapshots extracted from six alchemical binding free energy calculations with a flexible receptor, binding free energies were estimated for a total of 141 ligands. For 24 ligands, the calculations reproduced flexible-receptor estimates with a correlation coefficient of 0.90 and a root mean square error of 1.59 kcal/mol. The accuracy of calculations based on Poisson-Boltzmann/Surface Area implicit solvent was comparable to previously reported free energy calculations. Then we evaluate a number of common snapshot selection strategies using a quality metric from stratified sampling, the efficiency of stratification, which compares the variance of a selection strategy to simple random sampling. For docking sets of over five hundred ligands to four different proteins of varying flexibility, we observe that for estimating ensemble averages and exponential averages, many clustering algorithms have similar performance trends: for few snapshots (less than 25), medoids are the most efficient while for a larger number, optimal (the allocation that minimizes the variance) and proportional(to the size of each cluster) allocation become more efficient. Proportional allocation appears to be the most consistently efficient for estimating minima. Finally, we attempted a blinded prediction challenge D3R and applied AlGDock on several systems. I will describe the performance of our calculation. Overall, the study shows that AlGDock can work well for predicting the binding affinities and it demonstrates a strategy for developing an understanding of protein-ligand interactions. Show less
