29 May 10:15 Robert Cowell - Modelling DNA mixtures using Bayesian networks

HIIT seminars in spring 2009 will be held in hall **C222** of Exactum, on Fridays starting at 10:15 a.m. Coffee available from 10.

Fri May 29
Dr. Robert Cowell, Senior Lecturer
Cass Business School, City University, London.

Modelling DNA mixtures using Bayesian networks

Abstract:
In a DNA sample collected from a crime scene it is not uncommon to find that two or more peoples' DNA is mixed together . Two common scenarios present themselves to forensic scientists trying to analyse such samples. In one scenario one has a suspected perpetrator whose DNA profile is known, and it is required to establish whether or not the suspect has indeed contributed some of their DNA to the sample. In another scenario, one does not have a suspect at hand, and instead the desire is to separate out the genetic profiles of the contributors to the DNA sample, for use, perhaps, in a DNA database search of previous offenders to look for a match.
In this talk I shall present some Bayesian network models developed to try solve these problems, and some of the computational tricks that have had to be employed to handle some larger mixture problems in which junction trees can have state spaces in excess of a billion states. Prior background knowledge of genetics will not be required.

Please note also that Dr. Cowell will be acting as the opponent at the public examination of Tomi Silander's doctoral dissertation titled "The Most Probable Bayesian Network and Beyond". The examination will take place at 10 o'clock on Saturday, May 30th, in Auditorium XII of the Main Building.