A difficult issue that directly affects hit discovery and hit-to-lead optimization in drug design projects involving peptide-based pharmaceuticals is the virtual screening of protein-protein and protein-peptide interactions. Although a number of screening algorithms to predict the binding affinity of protein-protein complexes have been presented, methods developed especially to predict the binding affinity of protein-peptides are rather uncommon. Despite the higher complexity and variability of interactions provided by peptide binders, predictors developed to score the affinity of small molecules are frequently utilised for peptides indistinctively.

We present PPI-Affinity, a programme that uses support vector machine (SVM) predictors of binding affinity to screen datasets of protein-protein and protein-peptide complexes as well as to build and rank mutants of a given structure, in order to address this issue. On four benchmark datasets, including data on protein-protein and protein-peptide binding affinity, the performance of the SVM models was evaluated. A series of mutations of EPI-X4, an endogenous peptide inhibitor of the chemokine receptor CXCR4, as well as complexes of the serine proteases HTRA1 and HTRA3 with peptides were used to test our model.

Optimized PPIs are essential for the robust antibody binding to protein antigens during treatment. Therefore, it is crucial to characterise PPIs in terms of their binding affinity (BA) while developing new biologics and medicinal substances. In particular, peptides are a promising class of bioactive molecules that frequently exhibit greater selectivity and fewer adverse effects than small-molecule medications. More than 60 peptide medications are currently on the market and hundreds of other peptidic substances are undergoing preclinical or clinical testing. Due to their flexible architectures and variety of binding sites, peptide medicines continue to be difficult to design.