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

Abstract

In studying the strength and specificity of interaction between members of two protein families, key questions center on which pairs of possible partners actually interact, how well they interact, and why they interact while others do not. The advent of large-scale experimental studies of interactions between members of a target family and a diverse set of possible interaction partners offers the opportunity to address these questions. We develop here a method, DgSpi (Data-driven Graphical models of Specificity in Protein:protein Interactions), for learning and using graphical models that explicitly represent the amino acid basis for interaction specificity (why) and extend earlier classification-oriented approaches (which) to predict the ΔG of binding (how well). We demonstrate the effectiveness of our approach in analyzing and predicting interactions between a set of 82 PDZ recognition modules, against a panel of 217 possible peptide partners, based on data from MacBeath and colleagues. Our predicted ΔG values are highly predictive of the experimentally measured ones, reaching correlation coefficients of 0.69 in 10-fold cross-validation and 0.63 in leave-one-PDZ-out cross-validation. Furthermore, the model serves as a compact representation of amino acid constraints underlying the interactions, enabling protein-level ΔG predictions to be naturally understood in terms of residue-level constraints. Finally, as a generative model, DgSpi readily enables the design of new interacting partners, and we demonstrate that designed ligands are novel and diverse.

Systematic Testing of Belief-Propagation Estimates for Absolute Free Energies in Atomistic Peptides and Proteins

“…Building on previous work, [35][36][37][38][39][40]66,67 the present study reinforces the potential of BP calculations for equilibrium molecular mechanics calculations. In particular, we have begun to clarify the requirements for constructing peptide and protein MRFs of sufficient accuracy, and our results suggest that BP calculations using the latest hardware and BP-specialized software could be extremely powerful.…”

Computational Design of Selective Peptides to Discriminate between Similar PDZ Domains in an Oncogenic Pathway

“…FPD takes as input a starting structural model of the domain-peptide complex and samples peptide conformations (rigid body orientation and backbone/side-chain dihedral angles) as well as domain rotamers in and around the binding site, seeking to minimize an empirical scoring function [37]. Six-mer peptides were used in all calculations throughout this study, as the last six residues have been shown to encode much of the binding specificity [11, 18, 25]. To reduce the possibility of getting trapped in local minima, we used 70 existing PDZ-peptide complex structures to generate a diverse set of starting domain-docked peptide conformations, each initiating an independent FPD simulation (Fig.…”

“…Stanefa and Wallin studied PDZ/peptide binding free energy landscapes using implicit-solvent Monte Carlo simulations, showing qualitative agreement with experiments [21]. Other studies have demonstrated reasonable ability to describe the space of peptide binders using structure-based calculations [22-24] or sequence-based models trained on high-throughput experimental data [11, 25]. Machine-learning methods can even enable sequence-based prediction of affinities, as demonstrated by Kamisetty et al .…”

“…Machine-learning methods can even enable sequence-based prediction of affinities, as demonstrated by Kamisetty et al . [25]. PDZ domains have also been subjects of structure-based computational design studies.…”