Improved inference of intermolecular contacts through protein–protein interaction prediction using coevolutionary analysis

Abstract

Motivation Predicting residue–residue contacts between interacting proteins is an important problem in bioinformatics. The growing wealth of sequence data can be used to infer these contacts through correlated mutation analysis on multiple sequence alignments of interacting homologs of the proteins of interest. This requires correct identification of pairs of interacting proteins for many species, in order to avoid introducing noise (i.e. non-interacting sequences) in the analysis that will decrease predictive performance. Results We have designed Ouroboros, a novel algorithm to reduce such noise in intermolecular contact prediction. Our method iterates between weighting proteins according to how likely they are to interact based on the correlated mutations signal, and predicting correlated mutations based on the weighted sequence alignment. We show that this approach accurately discriminates between protein interaction versus non-interaction and simultaneously improves the prediction of intermolecular contact residues compared to a naive application of correlated mutation analysis. This requires no training labels concerning interactions or contacts. Furthermore, the method relaxes the assumption of one-to-one interaction of previous approaches, allowing for the study of many-to-many interactions. Availability and implementation Source code and test data are available at www.bif.wur.nl/. Supplementary information Supplementary data are available at Bioinformatics online.

Coevolution-based prediction of protein–protein interactions in polyketide biosynthetic assembly lines

“…We first set out to automatically classify docking domains into classes, and to assess for which of these classes sufficient data is available to predict their protein-protein interactions (PPIs). The performance of coevolution-analysis-based PPI prediction depends highly on the alignment quality and shifts in interaction site can lead to mispredictions 26,43 . While a previous method 23 used joint alignment of the docking domains from all three compatibility classes 22 , new structural studies have shown that these docking domain classes have a markedly different tertiary structure and therefore should not be aligned 19,21,25 .…”

“…Second, considerable amounts of new data have been collected into standardized repositories 17 , making it possible to substantially expand the training sets. Third, we have recently developed a new computational method that reduces noise in coevolution-based intermolecular contact predictions 26 , leading to considerable improvements in distinguishing between protein-protein interactions and non-interactions.…”

“…Here, we provide a new prediction pipeline, PKSpop, that exploits these new insights and developments, through automated recognition of docking domain classes, coevolutionary prediction of interaction probabilities between class I docking domains using expectationmaximization with the Ouroboros method 26 , and combining these results in a new algorithm that is usually able to predict a single most probable assembly line order. We show that PKSpop is able to accurately identify assembly line orders in non-collinear PKS systems in ~80% of the cases.…”

Coevolutive, Evolutive and Stochastic Information in Protein-Protein Interactions

“…Advances in PPI prediction (14)(15)(16)(17)(18) are highly welcome in the contexts of paralog matching, hostpathogen PPI network prediction and interacting protein families prediction. Recent studies suggest strategies like maximizing the interfamily coevolutionary signal (14), iterative paralog matching based on sequence "energies" (15) and expectation-maximization (18), which have been capable of accurately matching paralogs for some study cases. Despite these advances, the problem of PPI prediction remains unsolved for sequence ensembles in general, especially for proteins that coevolve in independent genomes though likely resulting from the same free-energy constraintsexamples are phage proteins and bacterial receptors, pathogen and host-cell proteins, neurotoxins and ion channels, to mention a few.…”

Coevolution-based prediction of protein-protein interactions in polyketide biosynthetic assembly lines

“…Second, considerable amounts of new data have been collected into standardized repositories 17 , making it possible to substantially expand the training sets. Third, we have recently developed a new computational method that reduces noise in coevolution-based intermolecular contact predictions 26 , leading to considerable improvements in distinguishing between protein-protein interactions and non-interactions. Here, we provide a new prediction pipeline, PKSpop, that exploits these new insights and developments, through automated recognition of docking domain classes, coevolutionary prediction of interaction probabilities between class I docking domains using expectationmaximization with the Ouroboros method 26 , and combining these results in a new algorithm that is usually able to predict a single most probable assembly line order.…”