Analysis, Synthesis, and Applications of Biological Networks

Recent developments in molecular biology have resulted in a new
generation of experimental datasets that bear relationships and
interactions between biomolecules, commonly referred to as
biological networks. Comparative analysis of molecular interaction
data plays an important role in understanding biological processes
by providing a framework for understanding cellular organization,
functional hierarchy, and evolutionary conservation. However,
while vast amounts of high-quality data is becoming available,
efficient analysis counterparts to sequence analysis tools such
as BLAST and CLUSTAL are not readily available for molecular
interaction data.

It is possible to model biological networks using various graph
theoretic formalisms. As is the case with sequences, two key 
problems on graphs derived from biomolecular interactions correspond 
to aligning graphs and finding conserved subgraphs in a collection
of graphs. However, graph theoretic formalisms often lead to
computationally hard problems due to their relation to subgraph 
isomorphism. Therefore, we analyze these networks using biologically
relevant graph simplification techniques coupled with efficient
algorithms to effectively solve these problems in terms of solution
quality and computational performance. We develop analysis algorithms
based on simplification of graphs to find conserved patterns in a
collection of metabolic pathways. In order to compare protein interaction
networks (PPIs) that belong to two different species, we develop a
framework based on duplication/divergence models that focuses on
understanding the evolution of PPIs, formulates the alignment problem
as a graph optimization problem, and develop heuristics to effectively
solve this problem. Our experimental results show that the proposed 
algorithms are able to discover conserved molecular interaction 
patterns effectively.

Finally, experimental approaches to gathering interaction data can
be slow and expensive. Using a number of biologically sound hypotheses,
we show how we can infer interactions using available data on
sequences and phylogeny with a high degree of accuracy. I shall
conclude the talk by outlining some of our ongoing work that examines
evolution of interaction maps (as opposed to sequences), and what
clues they might hold about functional characterization.





Biography:

Ananth received his B.Engg. degree from the University of Roorkee
(now IIT Roorkee) in 1989, his M.S. from the Wayne State University
in 1990, and his Ph.D. from the University of Minnesota in 1996. He
joined Purdue in 1996 as an Assistant Professor of Computer Sciences
and was promoted to Associate Professor in 2001. In recognition of
his work, he was selected a University Faculty Scholar at Purdue
in 2002. He is a recipient of the NSF CAREER award and was elected
the Outstanding Teacher in the School of Science by students in
2002.

Ananth's research interests span the areas of large-scale computing
applications, algorithms, and infrastructure. On these topics,
Ananth has authored several papers and co-authored a text book
"Introduction to Parallel Computing" (Second Ed., Addison Wesley,
2003).