HOPS: A Distributed Hybrid Optimization Technique for Protein
Structure Prediction

Alberto Maria Segre
Department of Management Sciences
Department of Computer Science
Program in Applied Mathematical and Computational Sciences
The University of Iowa
Iowa City, IA  (USA)

The key to understanding the mechanism of life lies in understanding
how proteins work. Nearly all functional aspects of an organism rely
on proteins; enzymes, brain chemicals like dopamine, hormones, and
hundreds of thousands of others. Surprisingly, a properly working
protein works because it has just the right three dimensional shape, a
shape determined by its molecular composition, which is described in
the genomic code.

Given that we now have access to extensive genomic information, the
next challenge for computational biologists is to determine a
protein's three dimensional shape -- and, consequently, its biological
function -- from its primary structure, expressed as the sequence of
constituent amino acids.  We have been working on a new hybrid
optimization approach to this problem that builds on previous work in
distributed search techniques for mechanized reasoning and
combinatorial optimization, various continuous optimization methods,
and protein energetics. In this talk, I will describe the general
architecture of our system, give an update on our recent progress, and
demonstrate some preliminary folding results.

Joint work with Yinyu Ye (Management Sciences/Applied Mathematics),
Kenneth Murphy (Biochemistry), Mauro Leoncini (CNR, Pisa, Italy), and
Giovanni Resta (CNR, Pisa, Italy).