Benchmarking orthology methods using phylogenetic patterns defined at the base of Eukaryotes.

Area of Science:

• Evolutionary biology
• Bioinformatics
• Genomics

Background:

• Manual analysis of protein phylogenetic profiles is time-intensive and limits scale for studying ancestral complexes and pathways.
• Automated orthology inference offers potential for large-scale analyses but exhibits significant variability between methods.

Purpose of the Study:

• To evaluate various orthology inference methods for their accuracy in recapitulating known eukaryotic genome evolution observations.
• To compare the performance of different orthology methods regarding phylogenetic profile similarity, ancestral gene content, gene loss, and overlap with manually curated groups.

Main Methods:

• Comparative analysis of multiple automated orthology inference methods.
• Evaluation based on phylogenetic profile similarity (complex co-occurrence).
• Assessment of inferred gene content of the Last Eukaryotic Common Ancestor (LECA) and gene loss patterns.
• Comparison of automated orthologous groups against manually curated datasets.

Main Results:

• Most orthology methods suggest a large LECA with significant gene loss and reasonably predict interacting proteins via phylogenetic co-occurrence.
• Orthologous groups derived from automated methods show imperfect overlap with manually curated groups.
• No single orthology method demonstrated superior performance across all evaluated aspects.

Conclusions:

• Automated orthology inference methods exhibit comparable large-scale performance in reconstructing eukaryotic genome evolution.
• Despite similar overall performance, different orthology methods generate substantially divergent orthologous groups.
• The variability highlights the need for careful method selection and potential integration with manual curation for robust evolutionary analyses.