An empirical comparison of distance matrix techniques for estimating codon usage divergence.

The quantitative description of codon usage divergence among taxa is a two-step process involving first the calculation of a standardized measure of intertaxon divergence for each possible pair of taxa followed by a summary and visual display of the patterns among the taxa inherent in these measures. Three different measures have been proposed in the literature, and three different summaries have been used. These different techniques are empirically compared using a data set consisting of gene sequences from seven species of Apicomplexa. The results suggest that the manhattan distance measure may be preferable to the use of the chi-square measure, although the separation of amino acid usage and codon usage by the genetic distance produces theoretical advantages. The multidimensional scaling ordination and unweighted pair-group clustering are both successful in displaying the patterns, while the eigenanalysis ordination is not.