Frela

Frela computes protein functional similarity between protein pairs using Gene Ontology (GO) annotations to quantify functional relationships.

Key Features:

• GO ontologies: Uses Gene Ontology (GO) annotations across Biological Process (BP), Molecular Function (MF), and Cellular Component (CC).
• Semantic similarity measures: Implements various established semantic similarity (SS) measures to quantify similarity between GO terms.
• Term-pair comparisons: Calculates semantic similarity between all GO term pairs associated with two proteins.
• Mixing strategies: Aggregates term-level SS values into a protein-level functional similarity (FS) using multiple mixing strategies.
• Similarity z-score: Computes a similarity z-score per protein that accounts for the background distribution of FS to mitigate annotation bias.
• Annotation corpus impact: Assesses the effect of different annotation corpora on FS calculations.
• Benchmarking: Benchmarked on the task of distinguishing orthologous gene pairs from random pairs, reporting moderate improvements in accuracy.
• High-throughput calculation: Enables high-throughput computation of protein functional similarity.

Scientific Applications:

• Protein-protein interaction prediction: Infers functional relationships that can support prediction of protein-protein interactions.
• Gene prioritization: Ranks candidate genes by functional similarity for downstream analysis.
• Disease gene discovery: Assists identification of disease-associated genes through functional similarity profiling.
• Orthology assessment: Differentiates orthologous gene pairs from random pairs for comparative genomics analyses.
• Genomics and proteomics analyses: Supports comparative functional analyses in genomics and proteomics studies.

Methodology:

Frela uses GO annotations (BP, MF, CC), computes semantic similarity (SS) between all GO term pairs for two proteins, aggregates SS values into protein-level functional similarity (FS) via mixing strategies, computes a per-protein similarity z-score from the background FS distribution, and evaluates effects of annotation corpus choice; it employs various established semantic similarity measures.