ModelOMatic: fast and automated model selection between RY, nucleotide, amino acid, and codon substitution models

Area of Science:

• Evolutionary biology
• Bioinformatics
• Computational biology

Background:

• Molecular phylogenetics uses genomic data to study evolution.
• Statistical methods like maximum likelihood and Bayesian inference are standard.
• Current model selection methods are limited to single data types (nucleotides, amino acids, codons).

Purpose of the Study:

• To extend model selection methods to compare models across different data types.
• To introduce adapter functions for projecting aggregated models onto observed sequence data.
• To develop and implement the ModelOMatic software for enhanced phylogenetic analysis.

Main Methods:

• Developed adapter functions to project aggregated models onto observed sequence data.
• Implemented these projections in the ModelOMatic program.
• Applied ModelOMatic to analyze nucleotide, amino acid, and codon models across diverse phylogenomic datasets.

Main Results:

• Amino acid models were predominantly selected for PANDIT and arthropod datasets.
• Codon models were overwhelmingly selected for the vertebrate dataset.
• ModelOMatic demonstrated computational efficiency, processing most PANDIT families in under 150 seconds.

Conclusions:

• ModelOMatic successfully enables cross-data-type model selection in phylogenetics.
• The choice of evolutionary model is influenced by data characteristics like sequence divergence and selection pressures.
• ModelOMatic offers a fast and integrable solution for improving phylogenetic inference pipelines.