Prediction of binding energetics from structure using empirical parameterization.

Abstract

We have presented an empirical method that can be used to predict the binding energetics for protein-protein or protein-peptide interactions from three-dimensional structures. The approach differs from other empirical methods in yielding a thermodynamic description of the binding process, including delta Cp, delta H degree, and delta S degree, rather than predicting delta G degree alone. These thermodynamic terms can provide a wealth of detail about the nature of the interaction, and, if sufficient experimental data are available for comparison, a greater assessment of the accuracy of the calculations. A recurring theme throughout this article is the need for more complete thermodynamic and structural characterizations of protein-ligand interactions. This includes not only characterization of the binding delta H degree, delta S degree, and delta Cp, but a thorough investigation into equilibria linked to binding, such as protonation, ion binding, and conformational changes. Sufficient data will allow parameterization on binding data rather than protein unfolding data. Further inclusion of information obtained from unfolding studies is not likely to generate significant improvement in the accuracy of the calculations. As additional binding data become available, the parameterization can be further extended to include relationships derived from analyses of these data. Not only will this increase accuracy and thus confidence, but allow extension of the method of additional types of interactions.