RAxML and FastTree: comparing two methods for large-scale maximum likelihood phylogeny estimation.

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Maximum Likelihood (ML) methods provide accurate phylogeny estimation but can be computationally intensive.
• RAxML is a leading ML-based tool for large-scale phylogenetic analysis, often requiring extensive computation time.
• Faster alternatives like FastTree exist, but their performance relative to RAxML requires further investigation.

Purpose of the Study:

• To compare the performance of FastTree and RAxML for large-scale phylogenetic tree estimation.
• To evaluate accuracy (ML score, topological accuracy), and computational efficiency (running time).

Main Methods:

• Evaluated FastTree and RAxML on thousands of simulated and biological nucleotide alignments.
• Analyzed datasets with up to 27,634 sequences.
• Compared performance under identical running time constraints and when run to completion.

Main Results:

• When run for the same duration, FastTree produced topologically more accurate trees than RAxML in most cases.
• RAxML achieved a higher ML score when run to completion but offered no substantial improvement in topological accuracy.
• FastTree's relative accuracy increased with dataset size and decreased alignment accuracy, outperforming RAxML on large, less accurate alignments.

Conclusions:

• FastTree enables rapid estimation of very large phylogenies with comparable or improved topological accuracy to RAxML.
• RAxML's computational time can be orders of magnitude longer than FastTree's.
• FastTree is a highly efficient tool for large-scale phylogenetic analysis, offering a practical alternative to RAxML.