InterPepRank: Assessment of Docked Peptide Conformations by a Deep Graph Network

MotivationPeptide-protein interactions between a smaller or disordered peptide stretch and a folded receptor make up a large part of all protein-protein interactions. A common approach for modeling such interactions is to exhaustively sample the conformational space by fast-fourier transform based docking, and then refining a select top percentage of decoys. Commonly, the methods capable of ranking the decoys for selection in short enough time for larger scale studies rely only on first-principle energy terms derived directly from structure such as electrostatics, van der Waals forces, or on precalculated statistical pairwise potentials. ResultsIn this work, we present InterPepRank for peptide-protein complex scoring and ranking. InterPepRank is a machine-learning based method which encodes the structure of the complex as a graph; with physical pairwise interactions as edges and evolutionary features such as PSSM and sequence conservation as nodes. The graph representation is trained to predict the LRMSD of decoys by using edge-conditioned graph convolutions on a large set of peptide-protein complexes, and performance is maximised by an ensembled prediction of variying network architectures. InterPepRank is tested on an independent test set with no targets sharing CATH family nor 30% sequence identity with any target in training or validation data. On this set, InterPepRank has a mean AUC of 0.70 for finding peptide-protein complexes with LRMSD<4[A]. This is an improvement compared to other state-of-the-art ranking methods that have a mean AUC in the 0.59-0.66 range. When included as the selection-method for selecting decoys for refinement in a previously established peptide docking pipeline, InterPepRank improves the number of Medium and High quality models produced by 80 and 40% respectively, proving its usefulness in decoy selection. AvailabilityThe program is available from: http://wallnerlab.org/InterPepRank ContactBjorn Wallner bjorn.wallner@liu.se Supplementary informationSupplementary data are available at BioRxiv online.