Title: "Molecular Divergence Time Estimation"
Speaker: Dr. Peter Waddell, Department of Biology, Purdue University
Place: Mechanical Engineering (ME) 161; October 28, 2008, Tuesday, 4:30pm

Abstract

Evolutionary trees may be constructed between species, where edge (branch) lengths are measured in the expected number of substitutions. The process of substitution as DNA sequences evolve is stochastic, but given enough sites and an constant rate of substitution, the relative pathlength to internal nodes in the tree may be estimated and equated to relative time (a so called "molecular clock"). This is useful, as we often have fossil evidence that only directly illuminates the age of a few of the nodes. Often, the rate of evolution is not constant but evolves itself (for example, due to the change of the DNA repair mechanisms). In that case, we need to assume a model of the rate of evolution (for example, Brownian motion), then make a maximum likelihood estimate of the uncertain times in the tree by maximizing the likelihood of the resultant rates on edges. Time estimates of sets of different models may be made, but which is best? I show how models may be compared and their fit tested, using the linearity of their ranked residuals. I reality, there are also issues of errors on fossil-inferred ages of nodes, on edge length estimates, and ultimately even on the tree. These will be illustrated with worked examples, including the relative ages of ape lineages, including humans.



Click here for a full schedule of BIOINFORMATICS SEMINARS, past and present.