Predicting protein function and binding profile via matching of local evolutionary and geometric surface patterns

Area of Science:

• Structural biology
• Bioinformatics
• Computational biology

Background:

• Determining protein function from solved structures is difficult, especially for orphan proteins.
• Existing methods struggle to separate functional and folding-related evolutionary constraints on protein sequences.
• Global sequence and structure similarity are often unreliable for inferring protein function.

Purpose of the Study:

• To develop a novel method for predicting protein functions from structures.
• To differentiate evolutionary patterns driven by protein function from those driven by protein folding.
• To accurately predict enzyme functions, binding specificity, and promiscuity.

Main Methods:

• Developed pevoSOAR (pocket-based evolutionary search of amino acid residues).
• Incorporates evolutionary information specific to individual binding regions.
• Matches local surfaces against a large database of precomputed protein surfaces.
• Generates a computed binding profile model for protein-binding activities.

Main Results:

• pevoSOAR accurately predicts enzyme functions.
• Achieved 80.55% correct function prediction at 99.98% specificity across 100 enzyme families.
• Demonstrated high performance with an Area Under the Curve of 0.955 and a Matthews coefficient of 86.6%.
• Successfully predicted biochemical functions for orphan proteins.

Conclusions:

• pevoSORO effectively predicts protein functions by analyzing binding surface evolutionary patterns.
• The method distinguishes functional selection pressures from folding-related pressures.
• pevoSOAR shows high accuracy and sensitivity in predicting enzyme function and binding characteristics.
• It offers a valuable tool for characterizing orphan proteins from structural genomics projects.