ResBoost: characterizing and predicting catalytic residues in enzymes

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Identifying enzyme catalytic residues is crucial for understanding enzyme function, drug design, and protein engineering.
• Experimental methods for determining catalytic sites are costly and time-consuming.
• There is a need for efficient computational approaches to identify catalytic residues.

Purpose of the Study:

• To develop and evaluate ResBoost, a novel computational method for identifying catalytic residues in enzymes.
• To integrate established heuristics (evolutionary conservation, 3D clustering, solvent accessibility, hydrophilicity) into a predictive model.
• To provide a method with tunable sensitivity and specificity for catalytic residue prediction.

Main Methods:

• ResBoost employs machine learning, specifically the AdaBoost algorithm and Alternating Decision Trees.
• The method formalizes and combines rules of thumb into a logical expression for residue prediction.
• Performance was assessed using cross-validation on 100 enzymes from the Catalytic Site Atlas (CSA).

Main Results:

• ResBoost achieved 85% sensitivity with a 9.8% false positive rate and 73% sensitivity with a 5.7% false positive rate.
• The method reduced false positives by up to 56% compared to using evolutionary conservation alone.
• ResBoost successfully identified recently validated catalytic residues not present in the CSA.

Conclusions:

• ResBoost offers an effective computational strategy for identifying enzyme catalytic residues.
• The method demonstrates improved accuracy and reduced false positives over traditional methods.
• ResBoost has potential applications in drug discovery and protein engineering.