Patch-DCA: improved protein interface prediction by utilizing structural information and clustering DCA scores

Abstract

Motivation Over the past decade, there have been impressive advances in determining the 3D structures of protein complexes. However, there are still many complexes with unknown structures, even when the structures of the individual proteins are known. The advent of protein sequence information provides an opportunity to leverage evolutionary information to enhance the accuracy of protein–protein interface prediction. To this end, several statistical and machine learning methods have been proposed. In particular, direct coupling analysis has recently emerged as a promising approach for identification of protein contact maps from sequential information. However, the ability of these methods to detect protein–protein inter-residue contacts remains relatively limited. Results In this work, we propose a method to integrate sequential and co-evolution information with structural and functional information to increase the performance of protein–protein interface prediction. Further, we present a post-processing clustering method that improves the average relative F1 score by 70% and 24% and the average relative precision by 80% and 36% in comparison with two state-of-the-art methods, PSICOV and GREMLIN. Availability and implementation https://github.com/BioMLBoston/PatchDCA Supplementary information Supplementary data are available at Bioinformatics online.

The extracellular milieu ofToxoplasma’s lytic cycle drives lab-adaptation and promotes changes in lipid metabolism primarily driven by transcriptional reprogramming

“…For selecting the right regularization parameters, Λ in estimating the inverse covariance matrix, a path of parameters (from 0.1 to 1 with 0.1 step size) was used and the Kullback–Leibler divergence was calculated. The best Λ was chosen as the value where the second derivative of the KL (Θ Λ j , Θ Λ j+1 ) function was smaller than a constant [77]. Once the optimal Λ was selected, the inverse covariance matrix was estimated accordingly.…”

“…, Θ * ! "# ) function was smaller than a constant [77]. Once the optimal Λ was selected, the inverse covariance matrix was estimated accordingly.…”

A computational framework for modeling functional protein‐protein interactions

“…Once the neighborhood residues are determined for each residue independently in each protein, a square symmetric matrix M of size (L/10) is constructed. Each element of the matrix M denotes the similarity between two neighborhood residue pairs a and b according to Vajdi et al 25 Each element of M is derived as follows: where a and b denote two residue pairs ( i 1 ,j 1 ) and ( i 2 ,j 2 ), and N i1 A and N j1 B denote the set of neighborhood residues of i 1 residue in structure A and set of neighborhood residues of j 1 residue in structure B .…”

Simplified geometric representations of protein structures identify complementary interaction interfaces

“…Many recent approaches have attempted to use multiple sequence alignments to predict residues that coevolve between proteins through direct coupling analysis, mutual information, or a combination of the two and show improved prediction capabilities. 8,28,29 One important challenge that remains for partner-specific, structure-based predictors is accounting for conformational changes that occur upon binding. The performance of these methods decreases with increasing conformational rearrangements and dynamics of the protein pairs upon binding.…”

“…Both approaches only use sequence information and do not incorporate spatial data. Many recent approaches have attempted to use multiple sequence alignments to predict residues that coevolve between proteins through direct coupling analysis, mutual information, or a combination of the two and show improved prediction capabilities 8,28,29 …”