A comparison of three theoretical approaches to the study of side-chain interactions in proteins

Abstract

The continuing increase in the number of high-quality protein crystal structures means that a considerable amount of data is now available to those studying the interactions of side-chain atoms. These experimental data can be used as a benchmark for theoretical studies of the spatial distributions of side-chain–atom and side-chain–side-chain interactions. We use the program suite SIRIUS to calculate experimental side-chain–atom distributions, complementing the existing side-chain–side-chain distributions, for each of four systems: phenylalanine–carboxylate, phenylalanine–aromatic, arginine–carboxylate, and arginine–aromatic. Three theoretical methods are tested: first the drug design program GRID, which is suited to calculating side-chain–atom distributions, secondly the distributed multipole analysis (DMA) electrostatic approach, and thirdly the empirical potential CHARMm. We look at the predictions of each method for the four systems, and compare theory with experiment.

Our results show that the strongly hydrogen-bonded arginine–carboxylate interaction is relatively easy to model, but that proper description of aromatic systems requires an explicit representation of the π-electrons, as modelled by the atomic quadrupoles in DMA. For arginine–aromatic, we have to consider the effect of competing interactions before we can successfully reconcile experiment with theory.