Title: Mathematical models for comparative genomics---successes and challenges
Speaker: Vladimir Pavlovic, Rutgers University
Date: Wednesday, May 12, 2004, 1:00 pm
Location: Hill Center, Room 260, Rutgers University, Busch Campus, Piscataway, NJ
Abstract:
The wealth of genomic data generated since the beginning of the genomic revolution in 1995 has given rise to the field of comparative genomics: a fast-growing branch of genomic research that exploits similarities and differences of functional elements in genomic sequences in order to elucidate genes, their regulation and function. A common wisdom inspired by the theory of evolution has lead to computational tools and models that can, for instance, significantly improve identification of genes in the Human genome by comparing it to mouse or zebra-fish genomes.
In this talk I will review our recent work on mathematical modeling for comparative genomics, from computational models for comparative gene identification to analysis of errors in these models. I will introduce two examples of our own comparative approaches: a product hidden Markov model and a comparative evidence-integration model. Both models attempt, from different perspectives, to model joint evolutionary statistics of two or more related genomic sequences. Product hidden Markov models model evolutionary conservation of amino-acids important for preservation of gene functions. However, they become computationally infeasible for pairs of long sequences and other than simple genomic structures. Approximate inference is one way in which we address this problem. As an alternative, we also considered the framework of evidence integration that considers phylogenetic homology as yet another source of integrable information.
Finally, I will show how an error analysis using our modeling framework for the first time answers a biologically interesting question of why and how an ``optimal'' pair of organisms should be selected for computational genomic comparison.
Bio:
Vladimir Pavlovic is an Assistant Professor in the Computer Science Department at Rutgers University and an adjunct assistant professor in the Bioinformatics Program at Boston University. He received his PhD in electrical engineering from the University of Illinois in Urbana-Champaign in 1999. From 1999 until 2001 he was a member of research staff at the Cambridge Research Laboratory, Cambridge, MA. Vladimir's research interests include time-series modeling, bioinformatics, and statistical computer vision.
Seminar sponsored by DIMACS/BIOMAPS Seminar Series on Quantitative Biology and Epidemiology.