Heterotachy and tree building: a case study with plastids and eubacteria.

Area of Science:

• Evolutionary biology
• Phylogenetics
• Molecular evolution

Background:

• Heterotachy, the variation in evolutionary rates across sites, is a key factor in phylogenetic analysis.
• Previous studies suggested that the proportion of variable sites (p(var)) differs among orthologous and paralogous sequences.
• The inference of heterotachy has been weak, leading to alternative models for sequence evolution and long-branch attraction (LBA) issues.

Purpose of the Study:

• To investigate the empirical evidence for differing p(var) among orthologues in anciently diverged lineages.
• To assess the impact of this form of heterotachy on common tree-building methods.
• To evaluate the robustness of different phylogenetic inference methods to model misspecification.

Main Methods:

• Analysis of plastid and eubacterial RNA polymerase sequences.
• Application of capture-recapture estimates and simulation studies.
• Comparison of phylogenetic performance for parsimony, least squares distance, and maximum likelihood methods.

Main Results:

• Empirical data and simulations confirmed that p(var) can differ among orthologues in anciently diverged lineages.
• Parsimony and a simple least squares distance method were susceptible to LBA caused by this heterotachy.
• Homogeneous maximum likelihood inference showed robustness to model misspecification in this specific case.

Conclusions:

• The form of heterotachy involving differing p(var) among orthologues is empirically supported and can significantly impact phylogenetic inference.
• Certain tree-building methods are vulnerable to LBA when this type of heterotachy is present.
• While maximum likelihood demonstrated robustness, caution is advised against assuming its universal invulnerability to model misspecification.