A Changepoint Approach for Locating Regions of Differential Variability in DNA Sequences

• Activity Number: 273
• Type: Invited
• Date/Time: Tuesday, August 5, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Hua Tang*+ and David O. Siegmund and Dmitri Petrov
• Companies: Fred Hutchinson Cancer Research Center and Stanford University and Stanford University
• Keywords: phylogenetic analysis ; differential variability ; retrotransposable element ; Poisson process ; DNA sequences ; changepoint

Abstract

When related DNA sequences are aligned and compared, we often observe that at some sites one or more sequences may exhibit a different nucleotide from the rest of the sequences. Do these variant sites occur randomly along the sequence? Are there any patterns according to which mutations have occurred and been preserved? Our goal is to develop a statistical method, which detects any region in which the probability of seeing a variant site is elevated or depressed. We combine phylogenetic analysis and a change-point approach to test the null hypothesis that the substitution rate is the same at each site in the sequence. The analytic approximation of the significance level is much faster, compared to hypothesis testing methods that rely on Monte Carlo simulation, and is very accurate in many situations. This framework can easily incorporate more complicated models. In particular, we test the null hypothesis in which the substitution rate is spatially homogeneous but depends on the type of substitution. We apply this changepoint approach to analyzing substitution patterns in two retrotransposable elements.