PhyloBench

PhyloBench evaluates phylogenetic inference programs by comparing methods against established programs using natural protein sequences and reference species trees.

Key Features:

• Natural sequence benchmark: Uses natural, not simulated, protein sequences of orthologous evolutionary domains rather than simulated data.
• Species tree reference: Measures accuracy by the distance between an inferred tree and the corresponding reference species tree.
• Multiple distance metrics: Implements several tree-to-tree distance measures, with Robinson-Foulds distance identified as the most reliable.
• Program comparison suite: Enables comparison of user methods against popular phylogenetic programs including MrBayes, FastME, and other distance, maximum likelihood, and maximum parsimony methods.

Scientific Applications:

• Phylogenetic program evaluation: Assesses the quality and accuracy of phylogenetic inference methods on real protein sequence data.
• Method selection guidance: Provides empirical evidence for choosing among phylogenetic algorithms, parameters, and options.
• Benchmarking Bayesian methods: Enables testing of Bayesian phylogenetic programs such as MrBayes on natural protein sequences.

Methodology:

PhyloBench computes tree-to-tree distances between inferred phylogenetic trees and corresponding reference species trees, using Robinson-Foulds distance as the most reliable measure among tested distance metrics.