Selecting optimal partitioning schemes for phylogenomic datasets.

Area of Science:

• Phylogenetics
• Computational Biology
• Molecular Evolution

Background:

• Partitioning estimates independent molecular evolution models for sequence alignment subsets, improving phylogenetic inference.
• Current partitioning scheme estimation methods are computationally limited to small datasets (<100 loci).
• Large phylogenomic datasets with thousands of loci are becoming common, necessitating scalable methods.

Purpose of the Study:

• To develop novel, computationally efficient methods for estimating optimal partitioning schemes in large phylogenomic datasets.
• To address the limitations of existing methods in handling the scale of modern sequence alignments.

Main Methods:

• Developed two novel methods: strict and relaxed hierarchical clustering for partitioning scheme estimation.
• These methods cluster similar site subsets within alignments based on data characteristics.
• Builds upon existing clustering approaches for enhanced performance.

Main Results:

• Compared strict and relaxed hierarchical clustering against existing partitioning selection methods.
• Strict hierarchical clustering demonstrates superior computational efficiency for extremely large datasets.
• Relaxed hierarchical clustering offers scalable efficiency and significantly improved partitioning schemes, validated by AICc and BIC scores.

Conclusions:

• The developed strict and relaxed hierarchical clustering methods represent the current state-of-the-art for inferring partitioning schemes on large datasets.
• Open-source implementations are available in PartitionFinder software.
• These methods are expected to enhance the accuracy of inferences from large phylogenomic datasets.