Structure- and sequence-based function prediction for non-homologous proteins.

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Structural genomics projects generate numerous protein structures with unknown functions.
• Experimental methods alone are insufficient for rapid functional elucidation.
• Conventional computational methods rely on homology, failing for proteins lacking apparent similarity.

Purpose of the Study:

• To review and present novel computational methods for predicting protein function.
• To address the challenge of functionally characterizing proteins without obvious homology.
• To highlight the contribution of computational approaches alongside experimental methods.

Main Methods:

• Developed local structure-based methods (Pocket-Surfer, Patch-Surfer) to identify similar ligand-binding sites.
• Developed sequence-based methods (protein function prediction, extended similarity group) using weakly similar sequences.
• Focused on methods that do not rely on global protein fold similarity.

Main Results:

• Pocket-Surfer and Patch-Surfer identify functional sites based on local structural features.
• Sequence-based methods leverage weak sequence similarities typically ignored in annotation.
• These methods effectively extract function information from distantly related proteins.

Conclusions:

• Novel computational methods offer powerful tools for functional prediction of uncharacterized proteins.
• Local structure and weak sequence analysis are key to overcoming limitations of homology-based approaches.
• Integrating these computational strategies with experimental methods will accelerate protein function discovery.