FAS: assessing the similarity between proteins using multi-layered feature architectures

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Bioinformatics

Background:

• Protein sequence comparison is crucial in bioinformatics.
• Annotated features enhance sequence comparisons but existing methods struggle with multiple or overlapping annotations.
• Accurate scoring of feature architectures is needed for reliable bioinformatics analyses.

Purpose of the Study:

• To introduce FAS, a new scoring method for protein sequence comparison.
• To address limitations in handling multiple and overlapping feature annotations.
• To improve the accuracy and reliability of protein architecture similarity assessments.

Main Methods:

• FAS integrates features from multiple annotation sources using a directed acyclic architecture graph.
• Redundancies are resolved by finding paths that maximize pair-wise architecture similarity.
• The method was evaluated on over 10,000 human-yeast ortholog pairs.

Main Results:

• FAS provides more plausible architecture similarities compared to existing methods.
• The method successfully resolves overlapping and redundant feature annotations.
• Case studies show FAS utility in orthology assignment, identifying diverged orthologs, and diagnosing gene prediction errors.

Conclusions:

• FAS offers a robust solution for integrating multiple feature annotations in protein sequence comparison.
• The method enhances the accuracy of protein architecture similarity assessments.
• FAS can be routinely integrated into various bioinformatics applications for improved analysis.