Gaining an understanding of the evolutionary history among a group of species is a fundamental problem in biology. The ease with which biologists obtain and disseminate genetic information and the resultant desire to analyze this data necessitates a valid statistical methodology which is computationally feasible for reconstructing large evolutionary trees and assessing the associated uncertainty. Other existing methods for reconstructing trees with may leaves depend on the computationally-intensive bootstrap method of resampling, whose validity in this context has been criticized by several authors. The investigators are developing Bayesian methodology and software implementing a novel Markov chain Monte Carlo alorithm for searching a tree- indexed parametric space, providing evolutionary biologists with a computationally feasible and statistically valid method of assessing uncertainty in reconstructed trees. This methodology is substantially superior on computational and theoretical bases to existing bootstrap methods. This project provides a new sophisticated computational tool for understanding evolutionary relationships of species on the basis of genetic information via a collaboration of the fields of statistics, computer science, and biology. The methodology the investigators are developing uses high performance computing in a novel manner to greatly improve the analysis of genetic data to elucidate evolutionary relationships. The methodology is general and may prove to be useful in wide-spread applications quite unrelated to evolutionary biology. This work is funded by Computational Biology Activities.