Applications and developments in free energy calculation

Optimized structures and relative energies for conformers of succinic acid and its monoanion in the gas phase were obtained using ab initio calculations at the MP2/6-311+G**//HF/6-31G* and MP2/6-311+G**//HF/6-31+G* levels, respectively. The lowest-energy conformer for succinic acid has a gauche conformation about the central bond and Z-acids. The lowest-energy structure for the monoanion has the expected internal hydrogen bond. Monte Carlo/Free Energy Perturbation (MC/FEP) using OPLS-AA was then used to predict that both compounds are ca. 80% gauche in water at 25°C. Though these results are in excellent agreement with experiment and consistent with the gauche effect, their true origin requires a detailed understanding of the solvation effects.; MC/FEP was used to predict the relative binding free energies for a series of pp60c-src SH2 domain ligands. Comparison of two procedures for disappearing atoms during an FEP suggests that the coupling of the removal of angular constraints with the disappearance of an atom may significantly slow free energy convergence. This series of ligands cover a range of modifications at the 3-position of 4-({lcub}[4-(cyclohexylmethoxy)benzyl]amino{rcub}carbonyl)phenyl phosphate. The 3-amide analog binds best by virtue of two hydrogen bonds within the glutamic acid binding region, one to the protein backbone and one to a highly coordinated water molecule.; An extensive examination of the conformational energetics of gas-phase dipeptides as a model for protein side-chains has been performed. Minimizations using OPLS-AA reproduces the geometries and energies of 229 conformers optimized at the ab initio level LMP2/cc-pVTZ(f)//HF/6-31G** with remarkable accuracy; the average root-mean-square deviation (rmsd) on geometries is 0.28 A and the rmsd on relative energies is 1.3 kcal/mol. Much of the discrepancies can be ascribed to overestimated electrostatic interactions by OPLS, a feature that improves condensed-phase properties.; A probabilistic algorithm for solvent-accessible surface area determination has been optimized to improve resolution of compounds within a congeneric series. Average error in the small-molecule training set improved from 30--40 A2 to less than 10 A2. Difficulties in expanding this algorithm to reproduce geometrical results in conformationally flexible systems and in the dense environments are discussed.