Touting the cost efficiency of accessing a shared supercomputing resource, Glaucus Proteomics said last week that it had linked up with SARA, a supercomputing facility in the Netherlands, via a dedicated broadband connection to perform protein microarray-related bioinformatics calculations.
Although Bunnik, Netherlands-based Glaucus formally signed an agreement last December with SARA and GigaPort, a Dutch government broadband internet initiative, in the past few weeks the company has completed the initial infrastructure development for accessing the SARA supercomputer, said Ian Humphery-Smith, chief operating officer for Glaucus.
Glaucus, which is developing arrays of human recombinant proteins expressed in E. coli, and arrays of antibodies against those proteins, needs the computing resources for performing pattern recognition and data visualization analysis, Humphery-Smith said. But unlike other companies involved in proteomics, such as Celera and GeneProt, which have built in-house computing facilities, Glaucus decided to take advantage of a Dutch government program that offers number-crunching capabilities on a fee-for-service basis.
“The advantage is that it’s just done at an incredible cost efficiency,” Humphery-Smith said. “We’ve made it operational after about six months’ work, [and] I did not spend $90 million to do this. [Other proteomics companies] have spent that much and at some point their investors want their $90 million back. That’s the efficiency in the business model.
“I have the same infrastructure, if not superior, to both Celera and GeneProt,” Humphery-Smith said. “It’s certainly equitable infrastructure.”
Humphery-Smith declined to disclose how much Glaucus has spent to hook up to the SARA supercomputer, but said, “I can assure you we put infrastructure inside the company that’s worth on the order of $3 million.” That figure includes the cost of digging a 10 km (6.2 mile) ditch to connect Glaucus via a dedicated dual optic dark fiber to the nearest GigaPort node to SARA, and acquiring on-site storage capacity, but not the usage fees Glaucus must pay for computing time, Humphery-Smith said.
Even though other companies in the Netherlands such as KPN Royal Dutch Telecom, the Dutch National Bank, and Philips also access the SARA supercomputer, Humphery-Smith said he was not concerned about running into problems accessing the computing resources “for the foreseeable future.”
Currently, Glaucus is connected with SARA via a one gigabit per second connection, and has the ability to back up several terabytes of data at any time, he added. By the third quarter of 2003 Glaucus will have infrastructure in place to back up 18 terabytes per month, and the company has plans to upgrade its connection to 10 gigabits per second, Humphery-Smith said.
At the supercomputing facility at SARA, based in Amsterdam, Glaucus will be able to utilize the capacity of TERAS, a 1,024-CPU system consisting of two SGI Origin 3800 512-CPU systems with a maximum performance of 1 teraflop per second.
Glaucus hopes to use these data handling capabilities to tease out patterns relevant to disease from studies using the antibody and protein microarrays it is developing. Rather than just compare two experiments, Humphery-Smith said he’s aiming to use protein microarrays to study, for example, how a drug affects patients of different ages, in different organs, and in different doses every day over 10 days. “I can get the data out on my chips but how do I process that without high-end computing?” he asked.
Humphery-Smith said the company’s next stage in development is to find ways to serve the American market, perhaps by establishing a field office in the Northeastern US. “The next objective is to get across the pond — how do we do that efficiently?”