Phylogenies -- evolutionary trees that represent the historical relationships among species -- provide the framework for comparative analysis in all fields of biology. Most phylogenies are now inferred from DNA or protein sequence data using methods that assume the evolutionary process is largely homogeneous. In reality, however, evolutionary dynamics often differ among sequence sites and among lineages. Our preliminary data indicate that several kinds of heterogeneity can cause current methods to infer the wrong phylogeny. The goal of this proposal is to develop, implement, and validate a new family of mixed-model phylogenetic methods that incorporate evolutionary heterogeneity in a maximum likelihood framework. The accuracy of our method versus current techniques will be evaluated with experiments using both simulated and empirical data sets. We will distribute user-friendly software to the scientific community that implements our method and provides a high-throughput platform for simulating and analyzing heterogeneous sequence data. This project will accelerate the scientific community's success in reconstructing the Tree of Life and improve our ability to interpret genomic, developmental, and physiological data in a comparative framework. More reliable phylogenies are beneficial to society because an understanding of evolutionary relationships is crucial for characterizing biodiversity and developing strategies to preserve it. Sound phylogenetic knowledge is also central to understanding the evolutionary processes that affect agriculture, ecosystem function, and infectious disease. Our software will also provide scientists a tool for high-throughput phylogenetic experimentation and data analysis, a key goal as whole-genome sequence data become available. This project will also provide graduate and undergraduate education and research training in computer science and biology, an important need as biology becomes increasingly information-driven in the 21st Century.
