Active site prediction using evolutionary and structural information

Abstract

Motivation: The identification of catalytic residues is a key step in understanding the function of enzymes. While a variety of computational methods have been developed for this task, accuracies have remained fairly low. The best existing method exploits information from sequence and structure to achieve a precision (the fraction of predicted catalytic residues that are catalytic) of 18.5% at a corresponding recall (the fraction of catalytic residues identified) of 57% on a standard benchmark. Here we present a new method, Discern, which provides a significant improvement over the state-of-the-art through the use of statistical techniques to derive a model with a small set of features that are jointly predictive of enzyme active sites.Results: In cross-validation experiments on two benchmark datasets from the Catalytic Site Atlas and CATRES resources containing a total of 437 manually curated enzymes spanning 487 SCOP families, Discern increases catalytic site recall between 12% and 20% over methods that combine information from both sequence and structure, and by ≥50% over methods that make use of sequence conservation signal only. Controlled experiments show that Discern's improvement in catalytic residue prediction is derived from the combination of three ingredients: the use of the INTREPID phylogenomic method to extract conservation information; the use of 3D structure data, including features computed for residues that are proximal in the structure; and a statistical regularization procedure to prevent overfitting.Contact: kimmen@berkeley.eduSupplementary information: Supplementary data are available at Bioinformatics online.

Functional site prediction by exploiting correlations between labels of interacting residues

“…Supervised machine learning techniques like naïve Bayes classifier and support vector machines have been applied to predict functional residues from their features mentioned earlier. FEATURE [37], WebFEA-TURE [20], S-Blast [22], Xin et al [39], Bhardwaj et al [4], Discern [30] are a few examples of methods which apply machine learning techniques for addressing the problem. Amitai et al [2] use network centrality features derived from a network of residue interactions in a given protein to identify functional sites.…”

“…We evaluated our method on CATRES-FAM benchmark dataset [30]. The dataset contains functional site annotations for 140 proteins.…”

“…These methods classify residues into functional and non-functional classes based on their structural, physicochemical, evolutionary and electrostatic features. The following are some of the examples of ML techniques used for addressing the problem: Wei et al [37] used naïve Bayes classifier, Gutteridge et al [11] used neural network, Youn et al [40] used support vector machines (SVMs) and Sankararaman et al [30] used L1 logistic regression classifier. Among all the existing methods, the L1 logistic regression classifier achieves the best performance on standard benchmark datasets.…”

“…The L1 regularization enabled them to overcome the problem of overfitting, which adversely affects the performance of ML techniques due to lack of availability of large number of training examples. On CATRES-FAM benchmark [30], it achieves 18% precision at 69% recall.…”

“…We propose a cautious collective inference scheme that exploits the inherent structure in the problem to provide a better performance than the existing techniques. Of the existing techniques the closest to that of ours is the usage of CRFs [30] to predict the functional residues. CRF is one of the popular collective classification models.…”