|
Events CCMB Distinguished Lectures Series 2010-2011
In this talk, I will describe new algorithmic tools that my group has developed, and which make it possible, for the first time, to obtain highly accurate estimates of trees from very large datasets, even when the sequences have evolved under high rates of substitution and indels. In particular, I will describe SATe (Liu et al. 2009, Science Vol 324, no. 5934). SATe simultaneously estimates a tree and alignment; our study shows that SATe is shows that SATe is very fast, and produces dramatically more accurate trees and alignments than competing methods, even on datasets with 1000 taxa and high rates of indels and substitutions. I will also describe our new method, DACTAL (not yet submitted). DACTAL stands for "Divide-and-Conquer Trees without Alignments", and uses an iterative procedure combined with a novel divide-and-conquer strategy to estimate trees from unaligned sequences. Our study, using both real and simulated data, shows that DACTAL produces trees of higher accuracy than SATe, and does so without ever constructing an alignment on the entire set of sequences. Furthermore, DACTAL is extremely fast, producing highly accurate estimates of datasets in a few days that take many other methods years. Time permitting, I will show how DACTAL can be used to improve the speed and accuracy of other phylogeny reconstruction methods, and in particular in the context of phylogenetic analyses of whole genomes.
Wednesday December 1, 2010 Hosted by: Ben Raphael
In computational genomics today an important problem is to understand as well as determine the manifestations of recombination events in chromosome sequences, and this problem is the main motivation behind this work. In this talk, I will focus on two interrelated problems. First, we explore the general problem of reconstructability of pedigree history. How plausible is it to unravel the history of a chromosomal segment? And, to what extent? By modeling the evolutionary dynamics of evolving populations as random graphs, we obtain fundamental new insights into one of the most interesting mathematical object in population genetics -- the Ancestral Recombinations Graph (ARG). The discovery of a structure-preserving core of the ARG, which we call the minimal descriptor ARG, aids in answering important questions related to reconstructablity, redundancies, boundedness, among others, of a generic ARG. The second problem deals with reconstructing the recombinational history of a collection of chromosomal segments as an ARG. I will discuss our approach to constructing the ARG topology centered around recombination events, each of which has been estimated with high confidence. I will conclude with a discussion on our ongoing work in the Genographic Project, a joint study with National Geographic.
Wednesday September 29, 2010 Hosted by: Sorin Istrail
_______________________________________________________ Events |
|||||||||
