Graphical Models of Residue Coupling in Protein Families

Abstract

Identifying residue coupling relationships within a protein family can provide important insights into the family's evolutionary record, and has significant applications in analyzing and optimizing sequence-structure-function relationships. We present the first algorithm to infer an undirected graphical model representing residue coupling in protein families. Such a model, which we call a residue coupling network, serves as a compact description of the joint amino acid distribution, focused on the independences among residues. This stands in contrast to current methods, which manipulate dense representations of co-variation and are focused on assessing dependence, which can conflate direct and indirect relationships. Our probabilistic model provides a sound basis for predictive (will this newly designed protein be folded and functional?), diagnostic (why is this protein not stable or functional?), and abductive reasoning (what if I attempt to graft features of one protein family onto another?). Further, our algorithm can readily incorporate, as priors, hypotheses regarding possible underlying mechanistic/energetic explanations for coupling. The resulting approach constitutes a powerful and discriminatory mechanism to identify residue coupling from protein sequences and structures. Analysis results on the G-protein coupled receptor (GPCR) and PDZ domain families demonstrate the ability of our approach to effectively uncover and exploit models of residue coupling.

Identification of family-specific residue packing motifs and their use for structure-based protein function prediction: I. Method development

“…residue pairs that change together in an alignment [28]; neighborhood correlation in the sequence similarity network, which accommodates multidomain proteins and domain swapping [29]; and Protein Function Templates (PFT/LIMACS, [30]) which include quantitative information about functional sites in their multiple alignment and profile/PSSM to yield better annotation. ModFun [31] adds similarity of protein interaction partners to improve the specificity of sequence annotation with PSI-BLAST.…”

Learning Sequence Determinants of Protein:Protein Interaction Specificity with Sparse Graphical Models

“…We employ a graph-structured model (which we refer to simply as a graphical model) that explicitly models amino acid interactions and provides a probabilistic interpretation for them. Sequence-based graphical models of protein families have been used to capture amino acid interactions in order to predict protein structure (Morcos et al, 2011;Jones et al, 2012;Kamisetty et al, 2013) and function (Thomas et al, 2008;Balakrishnan et al, 2011) and design new proteins (Thomas et al, 2009b;Kamisetty et al, 2011b). We build here on our sequence-based models of interacting protein families for binary prediction FIG.…”

Learning Sequence Determinants of Protein: Protein Interaction Specificity with Sparse Graphical Models

“…To do so, we employ a graph-structured model (which we refer to simply as a graphical model) that explicitly models amino acid interactions and provides a probabilistic interpretation for them. Sequence-based graphical models of protein families have been used to capture amino acid interactions in order to predict protein structure [26, 10, 12], function [37, 1] and design new proteins [39, 15]. We build here on our earlier work on sequence-based models of interacting protein families for binary prediction of interaction [38], significantly extending that approach to incorporate quantitative data and to predict Δ G .…”