Performance assessment of phylogenetic inference tools using PhyloSmew

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Phylogenetic inference tools are typically evaluated using simulated or empirical sequence alignments.
• A key challenge is determining the representativeness of these alignments compared to those used by researchers.
• The RAxMLGrove database enables simulation of DNA and amino acid sequences based on extensive empirical tree inferences.

Purpose of the Study:

• To develop and implement PhyloSmew, a tool for automating the benchmarking of phylogenetic inference tools.
• To assess the accuracy of phylogenetic tree inference using realistic simulated alignments.
• To compare the performance of inference tools on both simulated and empirical datasets.

Main Methods:

• Implemented PhyloSmew to simulate approximately 20,000 multiple sequence alignments (MSAs) from RAxMLGrove empirical trees.
• Analyzed 5,000 empirical MSAs from the TreeBASE database.
• Evaluated the inference accuracy of FastTree2, IQ-TREE2, and RAxML-NG on these datasets.

Main Results:

• All tested tree inference tools exhibited poor performance on quantifiably difficult-to-analyze MSAs.
• FastTree2 emerged as a practical alternative for inferring trees from challenging MSAs.
• Significant discrepancies were observed between the accuracy results obtained from simulated versus empirical data.

Conclusions:

• PhyloSmew facilitates more realistic benchmarking of phylogenetic inference tools.
• The performance of phylogenetic inference tools varies considerably on difficult datasets.
• Empirical data evaluation is crucial, as simulated data may not fully capture real-world complexities in phylogenetic analysis.