Topological Bias in Distance-Based Phylogenetic Methods: Problems with Over- and Underestimated Genetic Distances

• Main Author: Xuhua Xia
• Format: Article
• Language: English
• Published: SAGE Publishing 2006-01-01
• Series: Evolutionary Bioinformatics
• Subjects: topological bias; minimum evolution; least-squares method; Fitch-Margoliash
• Online Access: http://la-press.com/article.php?article_id=124

I show several types of topological biases in distance-based methods that use the least-squares method to evaluate branch lengths and the minimum evolution (ME) or the Fitch-Margoliash (FM) criterion to choose the best tree. For a 6-species tree, there are two tree shapes, one with three cherries (a cherry is a pair of adjacent leaves descending from the most recent common ancestor), and the other with two. When genetic distances are underestimated, the 3-cherry tree shape is favored with either the ME or FM criterion. When the genetic distances are overestimated, the ME criterion favors the 2-cherry tree, but the direction of bias with the FM criterion depends on whether negative branches are allowed, i.e. allowing negative branches favors the 3-cherry tree shape but disallowing negative branches favors the 2-cherry tree shape. The extent of the bias is explored by computer simulation of sequence evolution.