A major focus after the completion of many genome sequencing projects in
how the genes encoded in the genome are expressed at the messenger RNA
(transcriptomics) and protein (proteomics) levels. Expression Proteomics
is concerned with correlating cellular protein changes with cell
behavioral changes. A typical question is: what protein changes correlate
with the transformation of a normal cell into a cancerous cell? 

We have developed an approach that greatly accelerates the discovery of
these protein changes. There are several computational challenges that
need to be met to reach the full potential of this methodology: (1)
automated detection of protein changes, (2) expanding the sensitivity
range of protein detection and (3) data handling and management.

The current protein detection technology outstrips our ability to identify
the genes that encode low-abundance proteins of interest. We have embarked
on developing a novel technology to overcome this barrier. Our first step
was to formalize an idealized version of the problem in a well-known
setting of set-covering problems, and to exploit the special structure to
make our solution strategies computationally feasible. Our second step has
been to perform computer simulations to model a more robust version of the
problem including errors that is also somewhat more universal in its
design.

Jon will describe the biological context and framework of our new approach
while Ravi will elaborate on the two models and solution strategies that
we are developing. We will be describing joint work with Bjarni
Halldorsson, a doctoral student in the ACO program in Mathematics, who is
currently supported by a Merck Computational Biology Graduate Fellowship.