Dates/Time: Tuesdays (10:30-12:18pm) and Thursdays (10:30-11:18am), March 30 - June 3, 2010
(No 4/13 or 4/15)
Location: Jennings Hall Room 355
The field of phylogenetics seeks to study the evolutionary history for a collection of organisms. Phylogenetic techniques can be applied to a range of problems, including problems in epidemiology, forensic medicine, studies of parasite-host relationships, and the systematic classification of organisms. In this course, we will discuss the statistical models and algorithms that are used to infer phylogenetic trees using genetic data. Topics covered will include phylogenetic inference using parsimony, maximum likelihood, and Bayesian techniques, models for DNA sequence evolution, and models relating gene trees and species trees, with particular attention to the coalescent model.
The course is open for credit to graduate students in Statistics, Mathematics, or Biological Sciences. Students registered for credit as Stat 882 will be required to attend Thursday classes, which will include the discussion of background information to enrich the lecture as well as instruction in the use of software for carrying out phylogenetic inference. The Thursday sessions will be optional for other participants. Grading of Stat 882 will consist of homework and a final project.
These lectures will be available by live streaming video. If you would like to view MBI talks live, please email us at email@example.com and ask for a link.
In 2001, Rogers's gave a proof of identifiability for the popular general time reversible (GTR) Markov model with Gamma distributed rates mixed with invariable (I) sites for DNA evolution along a phylogeny. Recently, Allman, Ane, and Rhodes have pointed out an error in Rogers's proof and provided a proof using three-way species comparisons to show that the model without invariable sites is identifiable. We will discuss Rogers's approach and provide the proof of the missing link for the model with invariable sites using only pairwise species comparisons. There are a few exceptional cases that our method cannot handle, mainly the Jukes-Cantor model for DNA evolution. We will discuss what is known about that situation to date. Our proof involves no more than single variable calculus and should be highly accessible.