May 01, 2008 –
Researchers examined millions of genes, analyzing multiple data sets simultaneously.
Most of us can remember a six-digit computer password or even a 10-digit phone number. But give us a string of 100 numbers and the ole memory blows a gasket.
A similar – but hugely magnified – dilemma faced UC Irvine researchers Steven G. Potkin and James H. Fallon. The CALIT2-affiliated professors were in the throes of solving a medical mystery: identifying the genes associated with schizophrenia, a debilitating mental disorder.
They had to examine 25,000 genes, one at a time, factoring in 200-300 variations on each that can present one million variations across the genome. (They were able to whittle the variations down to 4,000 by utilizing principles of probability.)
In addition, they analyzed FMRI (functional magnetic resonance imaging) data from the participants – schizophrenics and a control group – that documented the brain’s activation patterns and circuitry while the subjects performed various memory-related tasks.
The researchers were looking for the proverbial needle in a haystack of data: emerging patterns that could help them connect millions of dots into a genetic picture.
First, they had to get a grasp on the vast quantity of data.
Data Sets Come Alive
They turned to HIPerWall, CALIT2’s Highly Interactive Parallelized Display Wall. The 200-million-pixel, tiled, grid-based wall gives researchers a larger-than-life view of their data sets at extremely high resolutions.
There, they could scrutinize multiple data sets simultaneously, comparing and contrasting images while they rotated, dissected, spliced and superimposed them.
The wall also facilitated their collaboration with experts in a variety of fields – cognitive science, physics, informatics, computer science, neuroanatomy, statistics and genetics – who contributed crucial expertise.
“If you’re going to do transdisciplinary research, you have to develop a shared conceptual framework,” says Potkin, director of clinical research in the Department of Psychiatry & Human Behavior, and director of UCI’s Brain Imaging Center.
“HIPerWall is a way to literally see the essence of other disciplines’ findings; instead of having to try to remember so much detail, you can just look over and there it is,” he says. “And the information is dynamic, so you can explore it, manipulate it and immediately see the results.”
Organizing Overload
The schizophrenia research, funded by the National Institutes of Health through BIRN (the bio-imaging informatics research network), involved an intricate labyrinth of information.
Data from schizophrenic subjects’ brain-activation images, brain circuitry and genetic makeup were analyzed and compared, both to others with the disease and to the control group. Simultaneously, millions of gene variations were scrutinized to see if any relationships to the brain data were evident.
HIPerWall’s adaptability allowed the researchers to group data in several ways: for example, by gender, severity of illness and genotype information. They could overlay additional data simultaneously, as well as display relevant information from international databases. “We started to discern patterns that would not be obvious from the statistics alone,” Potkin says, adding that those patterns could be validated by checking corresponding statistical data.
Before high-tech visualization tools were available, researchers had no choice but to wade through millions of mind-numbing, data-generated numbers. “But with HIPerWall, people didn’t want to leave our meetings,” says Potkin. “Everybody was really high with creative excitement.”
Fallon, professor of psychiatry and human behavior, concurs, noting that the wall “really brings a bit of the art into science.
“HIPerWall is not replacing our statistical or mathematical tools, it’s enhancing everything,” he adds. “It’s really value-added. You get an extra boost to your entire perceptual and intuitive process.”
Mission Accomplished
The wall ultimately engendered success; researchers identified two genes associated with schizophrenia and have filed a patent application to protect use of the genes for diagnosis or as targets for therapy.
Although it took about a year to gather and adapt the required information, they say the wall reduced their research time significantly. Fallon says the gene identification process occurred in an afternoon.
“We had Steve (Potkin), a psychiatrist who knew the clinical aspects of schizophrenia; we had a geneticist who knew the ins and outs of the gene patterns; we had Padhraic Smyth, who knew statistics, understood databases and developed text-mining tools; and I know the neuroanatomy. When you wrap all that together on HIPerWall, all of a sudden patterns came out very quickly. Hidden layers were revealed that allowed us to say, ‘look at this!’”
The researchers are poised to launch their next project, investigating genetic risk factors for Alzheimer’s disease. This time, they know that a specific gene increases the risk of contracting the disease, but there’s more to the story; their job is to determine what other genetic factors are responsible.
The volume of data dwarfs their prior research. This study will include structural, metabolic and amyloid plaque brain imaging from 850 subjects. Each subject has 650,000 possible genetic variations.
“We’re looking forward to having HIPerWall help us digest this,” Potkin says. “It will make the work more interesting and more fun. And now that our procedures are in place, we should be able to finish in about one-tenth the usual time.”
— Anna Lynn Spitzer
