Prediction of protein functional residues from sequence by probability density estimation

Area of Science:

• Protein bioinformatics
• Computational biology
• Structural bioinformatics

Background:

• Predicting ligand-binding and catalytic residues aids genetic and biochemical studies.
• Existing sequence-based methods suffer from low precision in identifying these functional residues.
• Current methods rely heavily on evolutionary conservation from multiple sequence alignments.

Purpose of the Study:

• To develop a more precise method for predicting protein functional residues.
• To enhance the accuracy of identifying ligand-binding and catalytic sites.
• To overcome limitations of existing sequence-based prediction techniques.

Main Methods:

• Combined evolutionary conservation, amino acid distribution, predicted secondary structure (ss), and relative solvent accessibility (rsa).
• Measured conservation by comparing site amino acid distribution to expected distributions for predicted ss and rsa states.
• Incorporated neighboring residue conservation and used conditional probability density estimation for functional site probability.

Main Results:

• The FRcons method demonstrates higher precision in predicting both ligand-binding and catalytic residues compared to existing methods.
• Achieved 50% precision for ligand-binding residues and 40% precision for catalytic residues at 20% sensitivity.
• Validated performance using large datasets from the Catalytic Site Atlas and PDB SITE records.

Conclusions:

• FRcons offers a significant improvement in predicting protein functional residues.
• The integration of diverse features enhances prediction accuracy.
• This method provides a valuable tool for guiding protein research.