New Weighting Methods for Phylogenetic Tree Recons

Authors: Kazuharu Misawa Fumio Tajima

Publish Date: 2012/08/08

Volume: 75, Issue: 1-2, Pages: 1-10

PDF Link

Abstract

Efficient determination of evolutionary distances is important for the correct reconstruction of phylogenetic trees The performance of the pooled distance required for reconstructing a phylogenetic tree can be improved by applying large weights to appropriate distances for reconstructing phylogenetic trees and small weights to inappropriate distances We developed two weighting methods the modified Tajima–Takezaki method and the modified leastsquares method for reconstructing phylogenetic trees from multiple loci By computer simulations we found that both of the new methods were more efficient in reconstructing correct topologies than the noweight method Hence we reconstructed hominoid phylogenetic trees from mitochondrial DNA using our new methods and found that the levels of bootstrap support were significantly increased by the modified Tajima–Takezaki and by the modified leastsquares methodA phylogenetic tree is a powerful tool for investigating the evolutionary history of organisms and genes Nowadays molecular phylogenetic analysis has become one of the most important methods for not only comparative studies of organisms Harvey and Pagel 1991 but also for the study of the evolution of genes Nei et al 2008 Molecular phylogenetic trees can be used for clustering gene families Misawa and Tajima 2000 Molecular phylogenetic analysis has gained importance because of advances in DNA sequencing techniques and sequence databasesAt present a large number of DNA and amino acid sequences are available for molecular phylogenetic studies Kuma and Miyata 1994 Misawa and Janke 2003 Murphy et al 2001 Nozaki et al 2009 These sequences may have different amounts of information about the phylogenetic relationships of the organisms in the study and different amounts of noise obscuring those relationships Russo et al 1996 Phylogenetic information is encoded in the DNA or protein sequences of contemporary species in a manner that allows the information from data such as DNA sequences to be used to trace the history back to the most recent common ancestor of the species Liu et al 2009The method of phylogenetic inference currently used in molecular phylogenetics can be classified into four major groups distance methods maximum likelihood methods Bayesian methods and parsimony Nei and Kumar 2000 In distance methods an evolutionary distance is computed for all pairs of sequences and a phylogenetic tree is constructed from pairwise distances such as neighbour joining NJ method Saitou and Nei 1987 When the phylogenetic tree is reconstructed using the distance methods the error in phylogenetic tree reconstruction can be reduced by applying large weights to distances with large information and small noise and small weights to noisy distances with small information Bull et al 1993 To recover correct phylogenies many authors have developed methods to determine the weights for transitional and transversional substitutions in cases where the Kimura 2parameter model is used Tajima and Takezaki 1994 defined an accuracy index for evolutionary distance and determined the weights that maximize the accuracy Goldstein and Pollock 1994 used a leastsquares method to determine the weights that produce a minimumvariance estimator from transition and transversion substitutions Unfortunately there is no consensus on the method for pooling distance data obtained for multiple loci Dutilh et al 2004 Huelsenbeck et al 1996The purpose of this study is to develop improved methods to weight distances from different genes for accurate reconstruction of phylogenetic trees We have modified the Tajima–Takezaki method and the Goldstein and Pollock method for multiple genes Two new methods developed were a modified Tajima–Takezaki method and a modified leastsquares method We used computer simulations to compare these two new methods to the leastsquares method and a noweight method evaluating their abilities to recover the correct tree topology In this study “efficiency” means the ability to recover the correct tree topology We determined the weights required to pool the distances estimated for the mitochondrial genes and reconstructed a hominoid phylogenetic treeIn this paper we used four weighting methods the noweight method the leastsquares method the modified Tajima–Takezaki method and the modified least squares method In the leastsquares method each locus was weighted by the average reciprocal of the sampling variances for the estimates of evolutionary distances for that locus Lynch 1999 Goldstein and Pollock 1994 also followed this approach to obtain an efficient distance by pooling transitional and transversional distances to recover correct phylogenetic trees from DNA sequences The purpose of Goldstein and Pollock’s method 1994 is to bring transversional distance and transitional distance together Our purpose is to bring together distances from several loci The purpose of Lynch 1999 is to obtain the divergence time while our purpose is to reconstruct phylogenetic trees Therefore the leastsquares method used by us differs from those reported by Goldstein and Pollock 1994 and Lynch 1999 The modified Tajima–Takezaki method maximizes the accuracy index Tajima and Takezaki 1994 of the pooled distance whereas the leastsquares methods minimizes its variance Goldstein and Pollock 1994 The modified leastsquare method is similar to the leastsquare method but it puts a single weight for all OTU pairs for one gene insuring a suboptimal weight will be used with all but one of OTU pairs while the leastsquare method puts one weight for each OTU pair for one gene The details of these weighting methods were described in Supplemental Materials

Keywords: