A phylogeny-based benchmarking test for orthology inference reveals the limitations of function-based validation

Area of Science:

• Bioinformatics
• Comparative Genomics
• Evolutionary Biology

Background:

• Accurate orthology prediction is vital for post-genomic research.
• Current benchmark methods often rely on functional annotation, which is not an evolutionary definition and may be limited by experimental data.

Purpose of the Study:

• To construct a high-quality, manually curated, phylogeny-based benchmark dataset for bacterial orthology inference.
• To demonstrate the superiority of this dataset over functional annotation-based proxies for benchmarking.
• To guide database design and parameterization for improved orthology inference.

Main Methods:

• Construction of a
• gold standard
• orthologous groups dataset for bacteria.
• Phylogeny-based curation of orthologous groups.
• Error quantification for orthology inference methods.

Main Results:

• Function-based tests often misjudge the performance of orthology inference methods by failing to detect false assignments.
• The curated dataset effectively guides database design and parameter optimization.
• Selection of a core species repertoire based on the dataset improves detection accuracy.

Conclusions:

• Manually curated, phylogeny-based datasets are more appropriate for benchmarking orthology than other practices.
• Including more genomes at appropriate evolutionary distances enhances orthology detection quality.
• The Reference Orthologous Groups dataset is publicly available for use.