Protein structure based prediction of catalytic residues

Area of Science:

• Structural biology
• Bioinformatics
• Computational chemistry

Background:

• Structural genomics projects are rapidly increasing protein structure data.
• A significant portion of newly determined protein structures lack functional annotations.
• Accurate functional annotation is vital for understanding protein roles.

Purpose of the Study:

• To identify features for predicting functional residues from 3D protein structures.
• To develop a computational method for identifying catalytic residues.
• To evaluate the effectiveness of various structural and sequence-based features.

Main Methods:

• Explored graph-based centrality measures (closeness, betweenness, PageRank) for interacting residues.
• Analyzed distance to the general center of mass (GCM) and relative solvent accessibility (RSA).
• Utilized relative entropy for sequence conservation and trained neural networks to predict catalytic residues.

Main Results:

• Distance to GCM and amino acid type, combined with sequence conservation, effectively predict functional residues.
• The method identified 411 potential functional residues from 9262, including 70 of 111 annotated catalytic residues.
• Achieved a 14-fold enrichment of catalytic residues, with 63% sensitivity and 17% precision, competitive with existing methods.

Conclusions:

• Distance to GCM offers a simple yet effective alternative to complex graph-based measures for capturing structural features.
• Sequence conservation remains the most influential feature for functional residue prediction.
• Conservation calculations require recalibration based on specific sequence databases due to their dynamic nature.