Exploring Protein-Peptide Binding Specificity through Computational Peptide Screening

The binding of short disordered peptide stretches to globular protein domains is important for a wide range of cellular processes, including signal transduction, protein transport, and immune response. The often promiscuous nature of these interactions and the conformational flexibility of the peptide chain, sometimes even when bound, make the binding specificity of this type of protein interaction a challenge to understand. Here we develop and test a Monte Carlo-based procedure for calculating protein-peptide binding thermodynamics for many sequences in a single run. The method explores both peptide sequence and conformational space simultaneously by simulating a joint probability distribution which, in particular, makes searching through peptide sequence space computationally efficient. To test our method, we apply it to 3 different peptide-binding protein domains and test its ability to capture the experimentally determined specificity profiles. Insight into the molecular underpinnings of the observed specificities is obtained by analyzing the peptide conformational ensembles of a large number of binding-competent sequences. We also explore the possibility of using our method to discover new peptide-binding pockets on protein structures.Author Summary: The interactions between proteins play a crucial role for almost every undertaking of a cell. Many of these interactions are mediated by the binding of relatively short unstructured polypeptide segments, or peptides, in one protein to well-folded domains in other proteins. Such protein-peptide interactions have some interesting and special properties, e.g., promiscuity, which means many different peptide sequences are able to bind the same protein domain. Peptides also often exhibit structural flexibility even after binding a protein. These special properties make it desirable, but also challenging, to simulate protein-peptide binding in atomistic detail for many different peptide sequences. To this end, we have developed a computational algorithm that simultaneously explores the structure of protein-peptide complexes and the amino acid sequences of the peptide. In particular, our algorithm allows binding-competent peptide sequences to be generated in direct relation to their binding strengths. We also explored the possibility of using our method to locate new peptide-binding pockets on protein structures. Computational algorithms such as the one developed here may pave the way to reveal the full complexity of protein-protein interaction networks used in cells.