Department of Statistics Seminar
North Carolina State University

presents

Spencer V. Muse

University of Missouri - Columbia

"Evolutionary Analyses When Nucleotides Do Not Evolve Independently"

ABSTRACT

The genetic information for most organisms can be represented as linear strings of the four characters A, C, G, and T, which denote the four nucleotides that compose DNA sequences. Studies of DNA sequences offer the biologist detailed views of organisms, but they also offer a rich set of novel problems to the statistician. Evolutionary analyses of DNA sequences provide a wealth of such problems. DNA sequences are sampled from modern organisms. However, since the DNA sequences descend (with mutations) from a common ancestral sequence at some unknown time in the past, the sequences can not be considered independent. These types of correlations are modeled by incorporating the evolutionary tree connecting the various species. Aside from the historical correlation among characters in different species, there are also spatial correlations among nucleotides within a single species. These correlations are the result of the biological function of the sequences, and have largely been ignored in most analyses. In this talk I will describe stochastic models that account for both types of correlation and demonstrate the use of these models in likelihood based analyses of multi-species datasets. The use of these procedures to infer biological properties of DNA sequences will be emphasized.

Thursday, February 19, 1998

8:00 - 9:00 am

124 Dabney Hall

Refreshments will be served in 124 Dabney Hall at 7:45.