Computational studies of protein-ligand interactions

Structural models have been produced for the oxygenase domain of the three human nitric oxide synthases (NOSs): endothelial (eNOS), inducible (iNOS) and neuronal (nNOS). The determinants of selectivity between the three NOSs were investigated using a variety of approaches: visual inspection, electrostatic calculations, and active site characterization using a variety of probes and systematically analysing for any significant difference. Analysis of the modelled active sites suggests that a single residue Asn 365 (eNOS), Asp 382 (iNOS) and Asp 602 (nNOS) could be useful in the design of selective inhibitors. However, the likely determinants of selectivity between human iNOS and nNOS (probably subtle) remain to be identified. Additionally, docking studies were able to accurately reproduce the crystallographically observed binding modes of ligands in iNOS and eNOS these became available subsequent to the modelling studies. Prediction of binding affinities was also investigated. A number of ligands were docked into the active site of the crystal structure of human iNOS and their binding affinities investigated using regression based scoring methods, producing predictive models. The activities of iNOS ligands were also predicted using Quantitative Structure-Activity Relationship (QSAR) methods, based on the ligands alone and in the absence of the receptor (i.e. iNOS). It was found that the best results were obtained by averaging the four individual predictions generating a 'Jury' model. This QSAR paradigm was successfully applied to predict the activities of a novel set of ligands. Docking has also been used to assist in the redesign of the active site of ascorbate peroxidase (APX). Enantiomeric product ratios following sulphoxidation of seven different aryl sulphide compounds in both wildtype and mutant (W41A) APX are predicted accurately using docking techniques. Movement of arginine 38 is predicted to be pivotal in determining the correct product ratio.