FAS

FAS computes similarity scores between protein feature architectures to improve protein sequence comparison by integrating annotations from multiple sources and resolving overlapping or redundant features.

Key Features:

• Integration of multiple annotation sources: Incorporates annotated features such as functional domains, transmembrane domains, low complexity regions, and secondary structure elements into a unified representation.
• Directed acyclic architecture graphs (DAG): Represents feature architectures as directed acyclic graphs to systematically integrate diverse annotations.
• Resolution of redundancies: Identifies paths through the architecture graph that maximize pairwise similarity scores to resolve overlapping or redundant annotations.
• Scalability and evaluation: Demonstrated performance in a comprehensive evaluation of over 10,000 human–yeast ortholog pairs, outperforming traditional methods based on e-values or unresolved overlaps.

Scientific Applications:

• Orthology assignment benchmarking: Evaluates the accuracy of orthology assignment software by providing architecture-based similarity assessments.
• Identification of functionally diverged orthologs: Pinpoints ortholog pairs that have diverged functionally for evolutionary and functional genomics analyses.
• Diagnosis of protein architecture changes: Detects protein architecture changes resulting from erroneous gene predictions to inform genomic annotation correction.

Methodology:

Constructs directed acyclic graphs to represent feature architectures derived from multiple annotation sources and computes similarity scores by finding optimal paths through these graphs that maximize pairwise similarities, thereby handling overlapping and redundant features.