FLORA: a novel method to predict protein function from structure in diverse superfamilies

Area of Science:

• Structural biology
• Bioinformatics
• Computational biology

Background:

• Predicting protein function from structure is crucial for structural genomics.
• Current methods often rely on global comparisons or local templates, with limitations in functionally diverse superfamilies.
• The relationship between protein structure and function is complex and not fully understood.

Purpose of the Study:

• To develop a novel method (FLORA) for automatically generating structural motifs associated with functional sub-families (FSGs) within protein domain superfamilies.
• To accurately discriminate between homologous domains with different functions.
• To provide novel predictions of enzymatic activity for uncharacterized protein structures.

Main Methods:

• FLORA generates structural templates based on specificity for functional sub-families, without prior assumptions about functional sites or residue properties.
• The method utilizes patterns of structural conservation across all residues to identify motifs.
• Benchmarking was performed on enzyme superfamilies across all major protein classes.

Main Results:

• FLORA accurately discriminates between homologous domains with different functions.
• The method substantially outperforms popular structure comparison methods and a leading function prediction method, showing a 2-3 fold increase in coverage at low error rates.
• The functional relevance of identified motifs was demonstrated, and novel enzymatic activity predictions were made for uncharacterized structures.

Conclusions:

• FLORA effectively detects functionally similar protein domain structures using conserved structural patterns.
• The method offers a significant improvement in predicting protein function from structure, particularly for functionally diverse superfamilies.
• FLORA has implications for annotating protein function in structural genomics and identifying novel enzymatic activities.