On the Relevance of Sophisticated Structural Annotations for Disulfide Connectivity Pattern Prediction

Abstract

Disulfide bridges strongly constrain the native structure of many proteins and predicting their formation is therefore a key sub-problem of protein structure and function inference. Most recently proposed approaches for this prediction problem adopt the following pipeline: first they enrich the primary sequence with structural annotations, second they apply a binary classifier to each candidate pair of cysteines to predict disulfide bonding probabilities and finally, they use a maximum weight graph matching algorithm to derive the predicted disulfide connectivity pattern of a protein. In this paper, we adopt this three step pipeline and propose an extensive study of the relevance of various structural annotations and feature encodings. In particular, we consider five kinds of structural annotations, among which three are novel in the context of disulfide bridge prediction. So as to be usable by machine learning algorithms, these annotations must be encoded into features. For this purpose, we propose four different feature encodings based on local windows and on different kinds of histograms. The combination of structural annotations with these possible encodings leads to a large number of possible feature functions. In order to identify a minimal subset of relevant feature functions among those, we propose an efficient and interpretable feature function selection scheme, designed so as to avoid any form of overfitting. We apply this scheme on top of three supervised learning algorithms: k-nearest neighbors, support vector machines and extremely randomized trees. Our results indicate that the use of only the PSSM (position-specific scoring matrix) together with the CSP (cysteine separation profile) are sufficient to construct a high performance disulfide pattern predictor and that extremely randomized trees reach a disulfide pattern prediction accuracy of on the benchmark dataset SPX, which corresponds to improvement over the state of the art. A web-application is available at http://m24.giga.ulg.ac.be:81/x3CysBridges.

On the Encoding of Proteins for Disordered Regions Prediction

“…Indeed, contrarily to the observation made in [1] that suggested a very small number of relevant feature functions in the context of disulfide bridge prediction, the selection algorithm identified here a larger set of interesting feature functions.…”

“…For this purpose, we consider various feature encodings and, in addition to the primary structure, three in-sillico annotations: position-specific scoring matrices (PSSM), predicted secondary structures and predicted solvent accessibilities. We apply the feature function selection pipeline in combination with Extremely randomized Trees (ETs), a model which gave excellent results in previous work [1]. In order to avoid any risk of overfitting or over-estimation of our models, we use three distinct datasets: Disorder723 [19], Casp10 (http://www.predictioncenter.org/casp10/) and Pdb30.…”

Soft Computing Methods for Disulfide Connectivity Prediction

“…The method proposed by Becker et al 6 employs three different classification algorithms for the prediction of disulfide bonding probabilities: k-nearest neighbors, SVMs, and extremely randomized trees. Therefore, they propose a feature function selection, which determines a subset of feature functions and the best setting for associated window sizes.…”

“…Sequence alignment is a standard technique in bioinformatics for visualizing the relationships between residues in a collection of evolutionary or structurally related proteins. Existing DCP algorithms in the literature have used multiple sequence alignment, position-specific scoring matrices (PSSMs) 6 , 7 and correlated mutations 4 as input encoding.…”

Disulfide Connectivity Prediction Based on Modelled Protein 3D Structural Information and Random Forest Regression

“…As the position specific scoring matrix encodes the evolutionary information of a protein, the feature derived from PSSM has been widely and successfully applied to disulfide connectivity predictions [23], [34], [43]. In this study, we extract the PSSM feature as follows: the original PSSM of a given protein sequence is obtained by executing PSI-BLAST [44] to search the Swiss-Prot database through three iterations with a default E-value cutoff; then, we transform the original PSSM to a normalized one by applying the logistic function fðxÞ ¼ 1= 1 þ e Àx ð Þto each element x contained in the original PSSM; finally, each cysteine residue is encoded into a 13 Â 20 ¼ 260-D feature vector that consists of the normalized PSSM elements corresponding to a sequence segment of length 13 centered on the cysteine residue [23], [34].…”