Exagen Diagnostics
Accelerating the Discovery of Genomic Marker Sets
It fell to Dr. Peter Hraber, Exagen’s Director of Computational Biology, to find a computing solution that met the company’s need for power, scalability, ease of administration and compact size.
“Before Peter came to work for us, I was running our discovery software on a single-processor Linux computer,” Harris recalls. “Even using an efficient search algorithm, it was taking me a week to make one pass through all the possible combinations. I was constantly frustrated because there was never enough time to look at all the combinations I needed to analyze.”
A faster computer — faster by an order of magnitude or more — was right at the top of Harris’s requirements. The proliferation of Apple Xserve G5 clusters on the TOP500 list of the world’s fastest supercomputers piqued Hraber’s interest in Apple’s high-performance computing technology. But raw processing speed was just one of many considerations.
“Since the people responsible for this system would be scientists rather than professional system administrators, ease of administration was central in our purchasing decision,” says Hraber. “Mac OS X Server is something we could administer ourselves. For anything less user-friendly than a Mac, we’d have needed to hire a specialist just to run our cluster.
“Mac OS X Server is something we could administer ourselves. For anything less user-friendly than a Mac, we’d have needed to hire a specialist just to run our cluster.”
“I had no trouble setting up the Xserve cluster and all its basic features myself,” he says. “And when I had questions about configuring the advanced security options offered by Mac OS X Server, AppleCare provided answers quickly, right over the phone.
“I had worked with other cluster architectures that took up a whole room, required special power, and generated lots of heat and noise,” Hraber adds. “I was delighted to learn we could fit a 16-processor cluster under a desk. The G5 processors consume a lot less power than comparable high-performance systems so the Xserves run cool, and our Xtreme Xrack enclosure makes the system very quiet.”
The Exagen system consists of nine dual-processor Xserve G5 systems and two Power Mac G5s, with a gigabit Ethernet switch for internode connectivity. Each node except the head node has a total of 2GB of RAM — 1GB per processor. The head node has twice as much memory to support system services.
Because the Exagen discovery process is more compute-intensive than it is data-intensive, the three drive bays on each Xserve G5 unit provide ample storage with plenty of growth options.
“I like the flexibility of disk configurations on the Xserve,” Hraber says. “With three drive bays, the head node runs the OS from one 80GB drive, using the other two drives to mirror the scientific data (RAID 1). This minimizes the risk of downtime in the face of drive failure.”
Increasing a Competitive Edge with the Right Technology
Exagen began with a technology-driven innovation in discovery methodology made possible by faster processors and refinements in computational analysis and data mining. The business process they developed around this technology gave them an important competitive advantage over companies that rely solely on traditional discovery methods.
By moving to the Apple cluster architecture and optimizing their software on the G5 processor for parallel processing, the researchers at Exagen get results much faster than before. The shorter turnaround lets them explore more scenarios than would otherwise be possible, increasing the odds of discovering the best possible set of genomic markers for any given application and of being first to market.
High-performance computers are helping researchers unlock the secrets of the human genome more quickly than ever before. The scientists at Exagen Diagnostics view their new Apple cluster — and the proprietary analytical software it lets them run so much faster than before — as a central factor in their quest for market-leading diagnostic testing products.
“Success, for us, is the ability to rapidly discover, validate and productize high-quality prognostic or predictive tests,”Mattingly concludes. “Our discovery engine — proprietary discovery algorithms enabled by fast, easy-to-use Xserve G5 technology — makes it possible for us to discover marker sets much faster than before. In the race for leadership in the market for genomic marker-based tests, we’re going to be hard to beat.”
The Computational Scientist’s System of Choice
“What really distinguishes equivalently powerful computers from each other is the user experience,” explains Peter Hraber, Exagen’s director of computational biology. “This includes the desktop, friendly productivity applications, intuitive widgets and controls, and tools to work on serious computational tasks. In my opinion, Apple’s combination of high-performance hardware and Mac OS X makes it the computational scientist’s system of choice.”
Promising New Areas for Research
“There’s a potentially important aspect to our business activities that’s easy to overlook,” muses Exagen co-founder Cole Harris. “Our focus at Exagen is on discovering genomic markers that we can develop into commercial tests. While it’s not necessary to completely understand what a gene or a protein is doing in order to use it in a test, we shouldn’t overlook the importance of our discovered markers for future research into the underlying mechanisms and associated pathways. No doubt there is substantial biological information in our data that will provide value, for example, in understanding the biology of cancer, hepatitis and other medical conditions.”
Cole Harris, VP of computational science at Exagen Diagnostics.
