Techniques for Assessing Phylogenetic Branch Support: A Performance Study

Abstract

The inference of evolutionary relationships is usually aid ed by a reconstruction method which is expected to produce a reasonably accurate estimation of the true evolutionary history. However, various factors are known to impede the reconstruction process and result in inaccurate estimates of the true evolutionary relationships. Detecting and removing errors (wrong branches) from tree estimates bear great significance on the results of phylogenetic analyses. Methods have been devised for assessing the support of (or confidence in) phylogenetic tree branches, wh ich is one way of quantifying inaccuracies in trees. In this paper, we study, via simulations, the perfo rmance of the most commonly used methods for assessing branch support: bootstrap of maximum likelihood and maximum parsimony trees, consensus of maximum parsimony trees, and consensus of Bayesian inference trees. Under the conditions of our experiments, our findings indicate that the actual amo unt of change along a branch does not have strong impact on the support of that branch. Further, we find t hat bootstrap and Bayesian estimates are generally comparable to each other, and superior to a consensus of maximum parsimony trees. In our opinion, the most significant finding of all is that there is no threshold value for any of the methods that would allow for the elimination of wrong branches while maintaining all correct ones—there are always weakly supported true positive branches.