COLLOQUIUM
Department of Computer Science and Engineering
University of South Carolina
 
Simple Math Is Enough: Two Examples of 
Inferring Functional Association from Genomic Data

Shoudan Liang

NASA Ames Research Center

Date: April 5, 2004 (Monday)
Time: 3:30-4:30PM
Place: Swearingen 1A03 (Faculty Lounge)

Abstract

High-throughput experiments, such as genome-wide monitoring of mRNA
expressions and protein-protein interactions, are expected to be
fertile ground for deriving protein functions. These data usually
contains a high rate of false positives. This talk discusses how to
extract reliable biological associations from genome-wide data. Using
two examples, we emphasize the importance of p-values, which are often
derivable using simple mathematics. The first example is based on our
recently published work (PNAS 100, 12579) on the reliable prediction
of protein functions from certain non-random features in large-scale
2-hybrid data. Our method assumes that if two proteins share
significantly larger number of common interaction partners than
random, they have close functional associations. Based on an analysis
on yeast protein-protein interaction data, we have derived more than
2800 functional associations. We derived tentative functions for 81
non-annotated proteins. Since the completion of the work, 23 of them
have been annotated in Saccharomyces Genome Database, and all but one
of our predictions proved to be correct. In the second example, we
will discuss our improvement to the REDUCER algorithm of Bussemaker et
al. (Nat. Genet. 27 167) that extracts protein-binding motifs from
microarray experiments. In this method, statistical significance is
derived from linear regression of the expression values to the copy
number of motifs in gene's promoters.

Shoudan Liang works at the NASA Advanced Supercomputing Division of the NASA
Ames Research Center. Trained in theoretical physics with a degree
from the University of Chicago, where he obtained a Ph.D. in 1986, 
Dr. Liang taught physics at Penn State
University before moving to NASA Ames six years ago. His main
interests include metabolic network modeling and cis-regulatory
elements analysis in development.