"Teraflop" is not a word you're likely to hear at your local computer store. It's a measure of processing power that allows calculations at a rate of a trillion per second with nine floating decimal points. How about 450 teraflops? That's the computing speed of the newest generation of UAB's massive computer, named Cheaha after Alabama's highest geographic peak. The device is already helping the university's students, scientists, and researchers further their diagnosis and treatment efforts.
The processing capacity is four times that of the former computer, which is a boon to researchers such as Frank Skidmore, MD, UAB professor of neurology, who works in the field of Alzheimer's and Parkinson's. The supercomputer can process in two hours a high-resolution brain image that formerly took a day. UAB worked with Dell to add the 72 graphics processing units.
"Cheaha allows me to accelerate the entire work process," Skidmore says. "If you can see better, you can plan better and you can treat better."
David Crossman, bioinformatics director at UAB, sends a different type of data to the supercomputer. "We analyze big data that's being generated from these new sequencer machines," Crossman says. "What we find can help identify mutations or variants in a particular person's DNA that could possibly lead to what they're encountering.
"So I take gigabite upon gigabite of really large files and crunch them through various pipelines to return back to me the mutations that might be there. The format looks much like a large text file containing what we're sequencing.
"The DNA sequence code is very short snippets of the entire genome. And our genome is 3.2 billion bases long, so we take sequences like 75 to 100 bases at a time and then put those back together to find out what has changed compared to the reference genome that we're looking at. The supercomputer comes in very handy for all of my analysis needs, basically because all these files are so large.
"If you tried to take every snippet of DNA that was sequenced from those and align them to the genome using a desktop computer, it would take months, maybe even years, whereas the supercomputer can process that information in less than a day."
Another researcher making use of this amazing processing power is Kristina Visscher, PhD in Neurobiology, whose work involves determining how the parts of the brain involved in attention interact with parts of the brain involved with vision. "We use functional MRI to examine human brain function and activity," Visscher says. "We ask how an interaction changes with experience; like with the experience of using vision differently after diseases like macular degeneration.
"For example, in a person with macular degeneration, we're looking at people who allocate attention to visual space differently because they lost their vision and we think that this changes their brain. Some people who've lost central vision get really good at doing things like reading or recognizing faces with their peripheral vision. You wouldn't even necessarily know they had vision loss if you were to interact with them.
"But other people with similar deficits in the eye aren't able to cope as well. We don't understand that. We are studying a group of people who have lost central vision in an effort to understand how this loss and use of peripheral vision changes their brains. Does the change in their brain relate to how well they're compensating?
"If it does, then we can start to think about what kinds of training we need for people who have trouble compensating, to make them better able to use their spare vision."
Visscher's lab is currently recruiting volunteers for the research. "We are looking for all kinds of people to participate in our various studies," she says. "Right now, we are looking for older adults, especially people who are cognitively healthy and over age 85." The website for information is https://labs.uab.edu/visscher/participate
The newest improvements make the supercomputer not only the state's fastest by far, but one of the five fastest at Southeastern universities. The device is said to be capable of comparing, mapping, and processing brain images by the hundred and even of sequencing a whole human genome. In fact, it contains the equivalent of the world's total gene-sequencing machines of only three years ago.
"Biomedical research now is big data," Crossman says.
"Our commitment to expanding this infrastructure at UAB gives us an advantage as we attract the top talent and compete for grants," says UAB president Ray Watts, MD.
Visscher said: "The speed with which Cheaha processes data moves some aspects of our work from 'not do-able in my lifetime' to 'very do-able.'"
