Thesis Defense: Models and Algorithms for Phylogenetic Reconstruction
, MIT EECS, CSAIL
Date: Friday, December 06, 2013
Time: 10:00 AM to 11:00 AM Note: all times are in the Eastern Time Zone
Location: 32-D463 (Star)
Host: Manolis Kellis, MIT EECS, CSAIL, Broad Institute
Contact: Yi-Chieh Wu, email@example.com
Speaker URL: None
TALK: Thesis Defense: Models and Algorithms for Phylogenetic Reconstruction
Computational techniques have long been applied to biological data to address a wide range of evolutionary questions. In phylogenetics, methods for reconstructing gene histories from sequence data have enabled researchers to better understand how evolution shapes gene content; for example, the identification of evolutionarily-related genes has allowed for the mapping of functions across species and the discovery of novel functions. Such predictions have become increasingly important over the last ten to fifteen years, as technology has reduced the cost of sequencing and increased processing power, leading to vast genomic datasets with little, if any, functional information. In turn, the growing availability of genomes has enabled comparative genomics, leading to increased power for biological signal discovery and revealing insight into the core evolutionary forces that govern our existence.
However, to realize the full potential in genomic and evolutionary studies, we require accurate, efficient, and scalable methods that are widely applicable. In this thesis, I address this need by developing novel computational approaches for reconstructing gene evolutionary histories. In particular, I consider models for gene family evolution that take into account (1) duplication, (2) loss, (3) horizontal gene transfer, (4) domain rearrangement, (5) genetic drift (leading to deep coalescence), and (6) nucleotide or amino acid substitution, and I present new phylogenetic algorithms for (1) eukaryotic gene tree reconstruction, (2) prokaryotic gene tree reconstruction, (3) gene tree-species tree reconciliation, and (4) sub-gene-level reconstruction. Through extensive benchmarking, I show that these methods dramatically improve reconstructions compared to state-of-the-art programs; in addition, they are efficient and require few modeling assumptions or parameters, making them applicable to a broad range of species and large datasets. As evidence, I apply these methods to clades of 12 Drosophila, 16 fungi, 15 primates, and 11 cyanobacteria, as well as to simulated phylogenies with up to 200 taxa, and demonstrate the large impact of accurate phylogenetic inference on downstream evolutionary analyses.
These results demonstrate the power of computational phylogenetics, and I believe that with the continued development and adoption of such methods, we can address fundamental biological questions with many important implications for future investigations of gene and genome evolution.
Created by Yi-Chieh Wu at Thursday, December 05, 2013 at 7:04 PM.