SCIENCE AT THE EDGE SEMINAR

Friday, 10 September 2010 at 11:30am

Room 1400 Biomedical and Physical Sciences Bldg.

Refreshments at 11:15


Speaker: Dr. Jun S Liu
         Department of Statistics, Harvard University

Title:   All about Interactions

Abstract:

In this talk we will report some of our recent efforts in developing methods
for detecting various interactions, such as protein-protein interactions,
gene-gene interactions, and amino-acid mutation interactions. We will start
with a simple Bayesian network method for mining protein-protein interactions
from unstructured text (such as pubmed abstracts). The method not only extracts
the names of the interacting proteins, but also the verb describing how they
interact. The method achieved an overall accuracy of 87% on a cross-validation
test, and can be used to complement human annotations to extract a large number
of new PPIs from literature. The webserver can be found at 
http://www.biotextminer.com/TextMining3/index.html.

Next, we consider the problem of detecting gene-gene interactions that may 
affect a certain trait, such as the risk of type-1 diabetes, or the 
expressions of a large set of genes. Because of the large number of genes 
and genetic markers in such studies, it is extremely challenging to 
discover how a small number of genetic markers interact with each other to 
affect certain human traits. We propose a Bayesian partition model to 
tackle the problem. Our extensive simulation studies mimicked such data 
structures and demonstrated that our Bayesian approach is much more 
powerful than the standard two-stage step-wise approach as practiced by 
statisticians and geneticists.