Prediction‐based fingerprints of protein–protein interactions

Abstract

The recognition of protein interaction sites is an important intermediate step toward identification of functionally relevant residues and understanding protein function, facilitating experimental efforts in that regard. Toward that goal, the authors propose a novel representation for the recognition of protein-protein interaction sites that integrates enhanced relative solvent accessibility (RSA) predictions with high resolution structural data. An observation that RSA predictions are biased toward the level of surface exposure consistent with protein complexes led the authors to investigate the difference between the predicted and actual (i.e., observed in an unbound structure) RSA of an amino acid residue as a fingerprint of interaction sites. The authors demonstrate that RSA prediction-based fingerprints of protein interactions significantly improve the discrimination between interacting and noninteracting sites, compared with evolutionary conservation, physicochemical characteristics, structure-derived and other features considered before. On the basis of these observations, the authors developed a new method for the prediction of protein-protein interaction sites, using machine learning approaches to combine the most informative features into the final predictor. For training and validation, the authors used several large sets of protein complexes and derived from them nonredundant representative chains, with interaction sites mapped from multiple complexes. Alternative machine learning techniques are used, including Support Vector Machines and Neural Networks, so as to evaluate the relative effects of the choice of a representation and a specific learning algorithm. The effects of induced fit and uncertainty of the negative (noninteracting) class assignment are also evaluated. Several representative methods from the literature are reimplemented to enable direct comparison of the results. Using rigorous validation protocols, the authors estimated that the new method yields the overall classification accuracy of about 74% and Matthews correlation coefficients of 0.42, as opposed to up to 70% classification accuracy and up to 0.3 Matthews correlation coefficient for methods that do not utilize RSA prediction-based fingerprints. The new method is available at http://sppider.cchmc.org.

Integrative computational modeling of protein interactions

“…Other methods analyze the structural properties of the monomers, such as surface-exposed residues, and/or employ propensity scales for pairs of residues in protein complexes (e.g. InterProSurf [90], SPPIDER [91], ProMate [92]). Others yet, combine both types of information or several different methods to offer a consensus or meta-prediction, which has been shown to be more accurate than the individual methods (e.g.…”

“…None SPPIDER [91] Integrates enhanced relative solvent accessibility predictions, evolutionary information, high-resolution structural data and physicochemical properties in a machine learning approach.…”

Computational Methods for Prediction of Protein-Protein Interaction Sites

“…(Caffrey et al, 2004;Porollo and Meller, 2007) On the other hand, detailed structural information can be used to characterize patches on the surface of a protein in terms of their geometric and other properties (see, e.g., (Bordner and Abagyan, 2005;Koike and Takagi, 2004;Neuvirth et al, 2004)). Structural conservation can also be taken into account when multiple structures within families are available.…”

“…Residues with a significant difference in ASA between the isolated chain and complex structures are then classified as "interacting". The following cutoffs for ASA were used: the loss of > 99% ASA for a given atom (Bradford and Westhead, 2005); a residue loses > 1Å 2 ASA in the complex (Chen and Jeong, 2009;Jones and Thornton, 1995;Liang et al, 2006); a residue ASA change by more than 20Å 2 (Kufareva et al, 2007); relative solvent accessibility (RSA) of a given residue decreased by more than 4% and its ASA decreased greater than 5Å 2 (Porollo and Meller, 2007). The latter definition uses relative ASA to address the considerable difference in size of amino acids, e.g.…”

“…In what follows, we will refer to protein interfaces derived using our own ASA-based definition, dRSA > 4% and dASA > 5Å 2 (Porollo and Meller, 2007), unless stated otherwise. This definition takes into account both relative and absolute change in ASA, and it attempts to filter out noise related to variation in RSA observed in structures resolved under different conditions, or for closely related homologs.…”