Tie trees generated by distance methods of phylogenetic reconstruction

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• The neighbor-joining (NJ) and unweighted pair group method with arithmetic mean (UPGMA) are common methods for constructing phylogenetic trees.
• Previous studies have identified instances where these methods produce multiple distinct trees (tie trees) from a single dataset.
• The frequency and specific conditions leading to tie tree generation remain incompletely understood.

Purpose of the Study:

• To investigate the frequency and conditions under which tie trees are generated using the NJ and UPGMA methods.
• To analyze the impact of sequence characteristics, such as divergence levels and number of sites, on tie tree occurrence.
• To evaluate bootstrap values associated with clusters in tie trees and explore methods to mitigate bias.

Main Methods:

• Computer simulations were employed to generate and analyze phylogenetic trees.
• Simulations examined scenarios with and without substitutions along interior branches, including parallel and backward substitutions.
• The study analyzed tie tree occurrence for sequences with varying divergence levels and numbers of sites.

Main Results:

• Tie trees can arise when no substitutions occur on interior branches or by chance due to parallel/backward substitutions.
• Tie trees were found to occur relatively frequently for sequences with low divergence or few sites.
• UPGMA generated tie trees more frequently than the NJ method for such data; bootstrap values for differing clusters were typically low (<60%).

Conclusions:

• Tie trees are a notable artifact in phylogenetic reconstruction, particularly for datasets with low sequence divergence or limited sites.
• The NJ method generally produces fewer tie trees compared to UPGMA under these conditions.
• Randomly selecting paths in bootstrap replications can mitigate input order bias in bootstrap values for NJ and UPGMA trees.