Use of amino acid sequence data in phylogeny and evaluation of methods using computer simulation

Abstract

The advantages and disadvantages of the use of amino acid sequence data for phylogenetic studies are critically examined. The accuracy of two of the main methods currently used to construct phylogenetic relationships from amino acid sequences is evaluated, using a computer program that produces model sequences by simulating the process of protein evolution. The results of this analysis show that the ancestral sequence method of Dayhoff & Eck (1966) is capable of greater accuracy than the matrix method of Moore et al. (1973), but only when the sequence data contain a small number of amino acid differences. It is estimated that that real data sets, such as plant cytochrome c, contain a high proportion (33%) of convergent substitutions, but that in spite of this, the ancestral sequence method can calculate the phylogeny of a protein such as cytochrome c with an error of only 2 to 8%. A phylogeny of a more variable protein, such as plastocyanin, is likely to be in error by 8 to 13%. A careful choice of proteins for a particular study is therefore necessary to ensure that the error in the calculated phylogeny is as low as possible.