Large-Scale Phylogenetic Analysis
The goal of phylogenetic analysis is to reconstruct the evolutionary history of different taxa (e.g., species, genera). Recent advances in molecular biology and genomics have provided biologists with molecular data at an unprecedented rate and scale. New approaches are necessary because the most accurate analyses are obtained through solving (or attempting to solve) NP-hard optimization problems. Furthermore, any such analysis can return hundreds or thousands of trees. Finally, some taxa evolve down networks rather than down trees. Our research uses discrete algorithms, graph theory, probability theory, and experimental performance analyses. Below we describe just three of several ongoing projects in the group.
1. Genome Rearrangement Phylogeny:
Whole genomes evolve via events, such as inversions, transpositions, deletions, and duplications, that change the order and content of genes within genomes. Such events are relatively rare, compared to nucleotide substitutions, and thus contain significantly stronger phylogenetic “signal”. Our group works on developing new methods for reconstructing phylogenies on whole genomes. The software suite GRAPPA, a result of collaboration with Bernard Moret and David Bader at the University of New Mexico, is several orders of magnitude faster than the previous best software for this kind of problem.
2. Absolute Fast Converging Methods:
Statisticians are interested in the sequence lengths needed by methods to reconstruct trees under Markov models of evolution. Our new “absolute fast converging methods” are methods that recover the true tree from polynomial lengths. Our experimental performance analyses show that our new methods greatly out perform other polynomial time methods with respect to topological accuracy, especially on big trees.
3. Visualization and Clustering of Sets of Phylogenetic Trees:
When a phylogenetic analysis of a dataset produces thousands of equally good trees, biologists summarize the information in the analysis with a consensus tree. We replace the traditional one-consensus approach with multi-consensus methods using clustering of trees, and develop tools for visualizing the distributions of trees in tree space. Large-Scale Phylogenetic Analysis
Bernard Moret (U. New Mexico)
David Bader (U. New Mexico)
Katherine St. John (CUNY)
Tamara Munzner (Compaq)
Linda Raubeson (Central Wash. U)