SCIENCE AT THE EDGE SEMINAR

Friday, 10 September 2010 at 11:30am

Room 1400 Biomedical and Physical Sciences Bldg.

Refreshments at 11:15


Speaker: Jason Mezey
         Department of Biological Statistics and Computational Biology,
           Cornell University, Ithaca, NY
         Department of Genetic Medicine,
            Weill Cornell Medical College, New York, NY


Title:   Scalable Algorithms for Genome-wide Disease Locus Mapping and 
         Gene Expression Network Analysis

Abstract:

High throughput genomics and next-generation sequencing technology are changing
how we study the genetics of disease and complex aspects of physiology. In this
talk, I will describe our research on the computational algorithms used to
analyze these data, with the objectives of answering the following questions:

1. Which genes contribute to increased risk for disease?
2. Can we discover gene networks important for complex phenotypes?
3. Are there novel genome-wide factors contributing to the etiology of disease?

For the first question, we have developed scalable penalized regression
algorithms for mapping the location of disease genes when simultaneously
analyzing millions of genetic markers in genome-wide association studies (GWAS).
By applying these algorithms, we have been able to confirm associations that
have not been replicated when analyzing each genetic marker individually.

For the second question, we employ probabilistic graphical models, which are
applied extensively in machine learning, to identify novel network and pathway
connections from genome-wide gene expression profiles of specific cell and
tissue types.  Graphical models provide a balance between model fitting and the
simplest representation of a network connection that is relevant to biology,
i.e. two genes interact or they do not.  We have developed scalable algorithms
for both undirected and directed graphical network models, where the directed
approaches take advantage of genetic perturbations for novel network discovery.

For the third question, we have developed specialized algorithms for analyzing
next-generation sequence data to discover new biological factors that may
contribute to the initiation and production of complex disease.  As an example
of the application of our methods, we have analyzed RNA-Seq profiles of human
lung tissue and we have identified genome-wide splicing alterations that are
produced by smoking.