HemeBIND: a novel method for heme binding residue prediction by combining structural and sequence information.

Area of Science:

• Structural bioinformatics
• Computational biology
• Biochemistry

Background:

• Predicting protein-ligand binding residues is a key challenge in structural bioinformatics.
• Heme is a vital ligand involved in numerous biological processes, necessitating accurate identification of its binding sites.
• Existing generic ligand binding prediction algorithms do not specifically address heme binding residue identification.

Purpose of the Study:

• To develop an efficient computational method for predicting heme binding residues.
• To integrate structural and sequence information for enhanced prediction accuracy.
• To provide a tool that complements experimental techniques for heme-protein interaction studies.

Main Methods:

• Systematic investigation of heme-protein complex interfaces to identify distinguishing sequence and structural attributes.
• Development of structure-based classifiers using Support Vector Machine (SVM) with features like evolutionary conservation and solvent accessibility.
• Creation of a sequence-based classifier using SVM and sequence profiles.
• Implementation of a post-processing procedure to reduce false positives.
• Combination of structure-based and sequence-based classifiers using a voting method.

Main Results:

• Several sequence and structural attributes effectively differentiate heme binding from non-binding residues.
• Combining these features in SVM classifiers significantly improved prediction performance.
• The developed post-processing step reduced false positives.
• The sequence-based classifier provided an alternative for predictions using only sequence data.
• A voting method combining structure-based and sequence-based predictions yielded superior performance compared to individual methods.

Conclusions:

• HemeBIND is the first specialized algorithm for predicting heme binding residues in protein structures.
• Both structure-based and sequence-based approaches effectively identify heme binding residues.
• Combining these complementary methods leads to significant improvements in prediction accuracy.
• The HemeBIND web server is available for public use at http://mleg.cse.sc.edu/hemeBIND/.