The computational peptide screening method is a Monte Carlo-based procedure to systematically characterize the specificity of a peptide-binding site. The method is based on a generalized-ensemble algorithm in which the peptide sequence has become a dynamic variable, i.e., molecular simulations with ordinary conformational moves are enhanced with a type of "mutational" move such that proper statistics are achieved for multiple sequences in a single run. The peptide screening method has two main steps. In the first, reference simulations of the unbound state are performed and used to parametrize a linear model of the unbound state free energy, determined by requiring that the marginal distribution of peptide sequences is approximately flat. In the second step, simulations of the bound state are performed. By using the linear model as a free energy reference point, the marginal distribution of peptide sequences becomes skewed towards sequences with higher binding free energies. From analyses of the sequences generated in the second step and their conformational ensembles, information on peptide binding specificity, relative binding affinities, and the molecular basis of specificity can be achieved. Here we demonstrate how the algorithm can be implemented and applied to determine the peptide binding specificity of a PDZ domain from the protein GRIP1.
Keywords: Affinity; Binding free energy; Generalized-ensemble techniques; Monte Carlo simulations; PDZ domains; Protein–peptide interaction.