University of Newcastle upon Tyne
School of Mathematics and Statistics
Statistics Seminars 2005-2006
9 December 2005 M502 2:00pm
Ian Wilson
Approximate Likelihood Methods for modelling human genomic variation.
Abstract
The variation within and between human populations is
being studied by the International HAPMAP project (www.hapmap.org), which has
produced an extremely large database of human genetic variation. Shared ancestry
causes complex correlations between chromosomes and between sites on the same
chromosome making the analysis of such datasets extremely challenging. One of
the more promising methods is the Product of Approximate Conditionals (PAC) (Li
\& Stephens). This method models the shared ancestry by coalescence and its
break-up by recombination as a sequence of hidden Markov models (HMMs). This
method was developed for the estimation of recombination rates but can be used
to make inferences about the block structure of data by sampling from the
underlying HMMs using the forward-backwards algorithm, and can be used to make
inferences for case-control association studies by looking at the properties of
unseen genetic variants within these blocks. An extension of this method is
developed to deal with data from multiple populations, or for Case-control data
from disease studies.
Return to timetable for 2005-06 seminars
School of Mathematics and Statistics Homepage