MEYER, Christopher P., Museum of Paleontology and Dept. of Integrative Biology, Univ. of California, Berkeley, CA 94720-4780
Phylogenies provide evolutionary biologists with a framework to test evolutionary hypotheses. However, it is well known through simulation studies that phylogenies based on molecular sequence data are susceptible to erroneous hypotheses of relationships and that taxon sampling has a major influence in accurately assessing relationships. This study provides an empirical example of how using exemplars (single taxon representatives from monophyletic lineages) leads to incongruent phylogenies. A data set of two mitochondrial gene sequences (COI and 16S) from 135 cypraeid species representing 40 monophyletic groups was used to investigate the role of taxon sampling. 300 taxonomic sets were constructed by randomly selecting one taxon from each of the 40 clades. Heuristic searches were completed for each of the 300 taxonomic sets under five weighting schemes to explore the effects of transition: transversion bias. The resulting topologies were collected and compared to each other by both the tree comparison method of symmetric distance and consensus methods.
The expectation that each taxonomic set should result in identical topologies was not met. Disturbingly, analyses that failed to take into account transition:transversion bias lacked resolution at 50% congruence. Accuracy, measured as a comparison to the topology generated when all taxa were included, was maximized at a Ti:Tv bias of 5:1, but fell off as weighting increased. These results strongly support the need for increased taxon sampling in molecular phylogenies. Effective sampling strategies should strive to break up long branches, decrease overall lambda (l), and increase the evenness of branch lengths in tree topologies. Additionally, the role that morphology can contribute to increasing the robustness of phylogenetic estimations cannot be neglected.