University at Buffalo - The State University of New York

Published: February 22, 2007

By ELLEN GOLDBAUM
Contributing Editor

Over the past year, UB's Center for Computational Research has quadrupled computing power, upgraded its high-performance storage system and installed a new state-of-the-art visualization room.

If that wasn't enough, it also moved its entire infrastructure, including a 2,000-processor supercomputer, from the university's North Campus into UB's New York State Center of Excellence in Bioinformatics and Life Sciences on the Buffalo Niagara Medical Campus in downtown Buffalo.

And CCR users noticed barely a hiccup in service.

Made possible by major investments in CCR by UB and the Center of Excellence during the past year, CCR's computing power has jumped from three teraflops (one teraflop equals a trillion operations per second) to 13 teraflops. Storage has been upgraded to nearly 30 terabytes.

While CCR staff and users enjoy the increases in power, as well as the center's new home, Thomas R. Furlani, CCR's director, says the real dividend from the move has come from new synergies it is generating with researchers in the Center of Excellence and its partners, Roswell Park Cancer Institute and the Hauptman-Woodward Medical Research Institute.

"The CCR staff has increased dramatically its interactions with medical researchers since the recent move and this has been highly beneficial to Buffalo's life-sciences projects," said Bruce A. Holm, senior vice provost and executive director of the Center of Excellence. "Dr. Furlani has done an outstanding job of educating our researchers about the possibilities open to them via CCR resources, and we are now seeing an increase in NIH (National Institutes of Health) grant applications that include CCR staff and services as part of their budgets."

At the same time, CCR is making sure that the needs of its existing users, many of whom work on the North Campus, remain a priority. Taking advantage of the boost in power are some of CCR's longest-standing users, who conduct research in fields ranging from computational chemistry and environmental modeling and simulation to earthquake engineering and anthropology.

CCR staffs full-time satellite offices on the North Campus in 107 Bell Hall and 331 Natural Sciences Complex.

Some of the newest North Campus users will be putting CCR machines to a major test later this year. UB high-energy particle physicists Avto Kharchilava and Karl Ecklund will use the center's supercomputers to help analyze the massive amounts of data that will be produced by the Compact Muon Spectrometer experiment at the CERN accelerator in Geneva later this year.

Since the move downtown, and partly as a result of energetic outreach efforts by CCR staff, existing partnerships with local medical institutions have intensified and become more productive, especially with the increase in computing power. CCR recently added a dozen or so new users, primarily in the life sciences.

Ironically, in a field so driven by virtual connections, sheer physical proximity to one's collaborators has turned out to be a terrific asset for researchers in specific disciplines.

Daniel Gaile, UB assistant professor of biostatistics whose office in the Center of Excellence is steps away from CCR, has noticed the change.

"I think we're really on an upswing now; there's some energy that has come just from CCR being located down here," he said. "There's been a definite increase in the number of life-sciences collaborations involving the CCR."

Gaile and Furlani agree that a key factor driving collaboration is CCR's proximity to faculty members in the Department of Biostatistics who have offices in the Center of Excellence and proximity to the adjacent Roswell Park Center for Genetics and Pharmacology, where RPCI's microarray facilities soon will be housed.

"This provides for one-stop shopping for researchers interested in conducting experiments and studies that require collaborations with all three groups," said Gaile. "They will not have to worry about coordinating their efforts across three disconnected groups; we are now very much connected."

These connections are critical because the large amounts of data generated by microarray researchers in the fields of genomics and proteomics—jointly referred to as "omics"—must be analyzed and managed by biostatisticians, operations that are enhanced greatly by access to CCR's hardware and expertise.

"The sheer amount of data generated by many of today's experimental techinques, such as microarray, flow cytometry and mass spectrometry, can be staggering and the need to store and analyze these data in a timely manner requires both high-performance computing and high-throughput storage," said Furlani. "Fortunately, the increase in computing power and storage at CCR over the past year has allowed us to provide better service to the Center of Excellence and UB researchers."

But fast machines and large storage arrays are only part of the story, Furlani pointed out.

CCR staff members also provide a broad range of support for users, including software engineering, graphical user interface development, advanced database engineering, scientific programming and modeling, algorithm optimization, bioinformatics expertise, scientific and medical visualization, and advanced computing administration.

And while CCR has seen a definite uptick in usage by the life-sciences researchers on the Buffalo Niagara Medical Campus, it continues to develop its cutting-edge visualization expertise by partnering with companies and government agencies on projects ranging from visualizing new toll plaza designs on the New York State Thruway to new traffic patterns on Main Street in downtown Buffalo to animations for MTV2 and the National Hockey League.

"We provide faculty with the hardware, software and human resources necessary to help enable their research, including custom software development," Furlani said.

Marc Halfon, UB assistant professor of biochemistry and biological sciences and a researcher in the Center of Excellence, provides a case in point.

Using the fruit fly as a model system, Halfon studies the gene regulatory elements in DNA sequences, the mechanisms that govern when genes are turned on and off. Information on what regulates genes is critical to understanding diseases, including birth defects, and evolutionary processes.

But little information had been gathered on regulatory elements, and what was known pertained only to single elements.

"We wanted to know, 'are there general principles involved in regulatory elements en masse that could be discerned from what we do know, rather than having to study them one at a time?" Halfon said.

That question was the impetus behind REDfly, a database of Drosophila gene regulatory elements that Halfon established with initial funding from the NIH. He recently published a paper on it in Bioinformatics.

Before establishing the database, fewer than 60 regulatory elements had been annotated, or described in detail as to their function. Halfon and his colleagues now have collected well over 600 and the database is not finished.

"Our resource is important not just for collecting the information, but it allows us to start doing computational studies on regulatory elements as a class and that was impossible before," he said.

Halfon said that while his research did not require supercomputers, it did require expertise in databases, so he contacted CCR to see if staff could recommend a graduate student who could assist him.

"Instead, they said 'We can do that for you, let's set up a meeting,'" he said. "They ended up doing the entire computational end of it."

Halfon said that Steve Gallo helped design a database schema, developed a Web-based interface and handled the back-end programming. CCR maintains REDfly on its computers; see http://redfly.ccr.buffalo. edu.

The level of service that CCR provides, Halfon said, would not have been possible if he had had to hire a part-time computer technician.

"This was not a hardware issue at all, but rather a service issue," he said. "I think there is some uniqueness to CCR, as supercomputing centers go, because they don't just provide access; they provide expertise."

Ping Liang, assistant professor of oncology in the Department of Cancer Genetics at RPCI, had a similarly positive experience with CCR when he received an urgent request from a collaborator at another institution.

Liang is working on a project that aims to identify the critical genetic factors responsible for the biological differences—including susceptibility to diseases—between humans and primates. The goal is to help prevent and treat human diseases, including AIDS and cancer.

Liang needed to conduct comparative genomics studies of more than 3 billion nucleotide sequences base by base—a job that could only be done by high-performance computers.

Even with the short time frame, CCR enthusiastically accepted the job. CCR staff member Cynthia Cornelius worked nights and weekends, providing Liang with access to U2, a high-performance computer cluster with 800 nodes.

"Not including set up and testing, it took U2 only one night to complete the job," said Liang. "It would have taken my small computer cluster months."

The individual attention that comes from CCR computational scientists and staff, such as Matt Jones, Zihua Hu, Steve Gallo and Martins Innus, all of whom work one-on-one with faculty members, has been received enthusiastically.

Over the next few months, CCR staff plans to meet with departments throughout UB and the Center of Excellence to understand better how the center can respond to their needs.

Also planned is a series of workshops describing CCR's capabilities, infrastructure and ways that CCR resources can be used.

Liang, who is director of Roswell Park's newly established Bioinformatics Core Facility, had used CCR when it was housed on the North Campus.

"Having CCR move into the Center of Excellence definitely has improved our interactions with their staff," he said. "Considering the nature of the research we do at Roswell, I envision that we and other biomedical researchers will become major users and beneficiaries of CCR in the near future."