University of Pittsburgh
Reducing the Complexity in Complex Genetics
Think of Dr. Barmada and his colleagues as gene hunters and method developers. All of their research projects attempt to identify genes that lead to variations in human traits — from differences in height and bone density to susceptibility to disease.
The Xserve cluster supports dozens of highly specialized statistical genetics applications used in the ongoing research projects. Primary tools include open-source applications such as Allegro, Merlin, Simwalk2 and Transmit. Dr. Barmada is working with BioTeam to optimize Merlin and Allegro for the G5 processor. “A lot of the algorithms should be amenable to vectorization so we anticipate a major advance in performance speed,” he says. “It’s just a matter of going into the code and doing the work.”
Using different algorithms, these applications analyze unique markers that characterize a segment of an individual’s genetic material. They then correlate the markers with patterns of transmission through a family or population.
“Our studies — even those that analyze just 400 markers in a population of, say, 20 families with 25 people each — can be extremely compute intensive,” says Dr. Barmada.
“Then we plan to add to our current studies new MCMC-based methods and new laboratory techniques involving microarrays, genome-wide association studies with 10,000 to 500,000 markers,” he says.
“They will greatly increase the number of computations required for a genetics project, but the Xserve has both the horsepower and the storage capacity to handle whatever we throw at it.”
