Catalytic residues in hydrolases: analysis of methods designed for ligand-binding site prediction

Abstract

The comparison of eight tools applicable to ligand-binding site prediction is presented. The methods examined cover three types of approaches: the geometrical (CASTp, PASS, Pocket-Finder), the physicochemical (Q-SiteFinder, FOD) and the knowledge-based (ConSurf, SuMo, WebFEATURE). The accuracy of predictions was measured in reference to the catalytic residues documented in the Catalytic Site Atlas. The test was performed on a set comprising selected chains of hydrolases. The results were analysed with regard to size, polarity, secondary structure, accessible solvent area of predicted sites as well as parameters commonly used in machine learning (F-measure, MCC). The relative accuracies of predictions are presented in the ROC space, allowing determination of the optimal methods by means of the ROC convex hull. Additionally the minimum expected cost analysis was performed. Both advantages and disadvantages of the eight methods are presented. Characterization of protein chains in respect to the level of difficulty in the active site prediction is introduced. The main reasons for failures are discussed. Overall, the best performance offers SuMo followed by FOD, while Pocket-Finder is the best method among the geometrical approaches.

Fig. 1

Relative solvent accessibilities (RSAs) of the 20 amino acid residues for catalytic and non-catalytic residues separately

Fig. 2

Fractions of chains for which a method produced at least one true positive (A), F-measure (B) and MCC (C) parameters for both perspectives (AA and SO)

Fig. 3

Points in the ROC space representing following results: CASTp, ConSurf—CS, ConSurfDB—CSDB, FOD, PASS, Pocket-Finder—PF, Q-SiteFinder—QF, SuMo, WebFEATURE—WebFT. ROC convex hulls are denoted by solid lines, while points related to the same method but different perspectives are connected by dashed-lines

Fig. 4

Correspondence analysis presenting relation between chain difficulty and subclass of hydrolases (A, B), CATH class (C), quaternary structure (D) and presence of ligands (E, F). The two perspectives are analyzed: AA (A, E) and SO (B, C, D, F)

Fig. 5

Distribution of RSA for catalytic residues of easy, medium and hard chains

Fig. 6

Surface representations of structures found as the hardest chains. Colouring scheme applied to opaque surface: ConSurfDB9-yellow, CASTp-magenta, FOD-blue, Q-SiteFinder-orange, SuMo-green, WebFEATURE95-cyan. Catalytic residues are red spheres or pointed by an arrow. (A) Surface of type-2 restriction enzyme Cfr10I (1CFR:A) with depicted predictions of CASTp, FOD, Pocket-Finder and WebFEATURE95. (B) Ribbon model of quaternary structure of type-2 restriction enzyme Cfr10I with one chain as transparent surface. The four monomers are blue, purple, orange and yellow, two chains have residues indicated by CASTp as spheres. (C) Surface of ribonuclease H (1RDD:A) with depicted predictions of CASTp, ConSurfDB9, SuMo and magnesium ion as sphere. (D) Transparent surface of ribonuclease H with underlying ribbon model. RNA and DNA binding sites [99] are shown as yellow and purple spheres, respectively. (E) Surface of endo-alpha-sialidase (1V0E:A) with depicted predictions of CASTp, FOD, SuMo, Q-SiteFinder and WebFEATURE95. (F) Ribbon model of quaternary structure of endo-alpha-sialidase and transparent surface of one chain. The three monomers are blue, purple and yellow. Residues involved in sialic acid binding are shown as spheres. (G) Surface of Arg-gingipain (1CVR:A) with depicted predictions of CASTp, SuMo, WebFEATURE95 and ConSurfDB9. Ligand molecule in sticks, calcium ions in green spheres and zinc ions in grey spheres. (H) Transparent surface of Arg-gingipain with underlying ribbon model