The University of North Carolina at Chapel Hill School of Pharmacy is dedicated to drug discovery and the assessment of existing drugs and treatments, but the group had a problem -- and no, it wasn't someone getting sick.
The school tests various methodologies and applications within the drug discovery market, and uses complex computer models to test processes and analyze molecules that protect people and save lives.
However, with a staff that fluctuates between 12 and 15 graduate students, technicians and professors, time is almost as valuable as the drugs waiting to be discovered.
About a year and a half ago, the school's progress had started to become affected by several error-prone manual processes that had slowed down turnaround time on valuable research, said UNC professor Alex Tropsha, the director for the Laboratory for Molecular Modeling.
The problem was compounded by the fact that over the years, the university had gained access to much larger data sets, Tropsha said. Each data set needed to be developed into a computer model that a chemist would then use to make predictions and develop new components for biological activity. The structures also had to be translated into a numerical format so that a matrix could be created for future analysis.
Until IBM WebSphere and Linux came along 16 months ago, those cumbersome tasks were conducted manually by one of the graduate students attending the school.
"We went with grid computing on Linux because a lot of the grid managers are all designed for the Linux system," said Scott Olosof, a graduate student at the UNC School of Pharmacy.
Olosof said he was also very impressed with the stronger security of Linux and the way in which it easily incorporated itself with the UNC file system.
"One of the major advantages we experienced was speed, but it also increased the system's reliability," Olosof said. "A lot of things depended on manual input from the user, so by automating the system we eliminated typos and improper tool implementation."
Tropsha said the entire implementation process took about 16 months to complete, from initial installation to the point when all the processes were being performed at optimum capacity. This period of time was considered to be an "exceptionally fast" implementation by Tropsha, considering the complexity of the information being automated.
"In one year of working with IBM, we developed an integrated functioning workflow … and we also realized that we could do hundreds of simultaneous calculations on many machines," Tropsha said. "We had built a workflow that would both integrate the problem and provide easy access through a Web interface."
Both Tropsha and Olosof agreed that what the department saved was more than monetary in value.
"The amount of time we have saved is 10 or 20 orders of magnitude," Olosof said. "The more samples we processed, the more accurate [the] models. We were given much more accurate results because we could run more calculations."
The fact that WebSphere was designed for mostly one directional queries presented a challenge for the team, Olosof said, because the often-used technology and processes often split or looped.
Olosof worked side-by-side with IBM engineers during the process to address and overcome these challenges.
"We had to push the middleware beyond what it was initially designed for," he said.