The Ludwig Institute for Cancer Research (LICR) founded in 1971 is made up of a worldwide network of nine branches in seven countries dedicated to basic and clinical research. Each branch of LICR has its own area of specialization. The Melbourne Branch in Victoria, Australia, concentrates on tumor biology; specifically, for colon cancer.
“LICR researchers are working to understand the biology of tumors within the gut and how protein markers secreted by tumors can be used as an early detection mechanism for colon cancer,” says Dr Robert Moritz, Manager of the Proteomics Facility at LICR and Director of the Australian Proteomics Computational Facility. “Being able to detect tumors before they become life threatening would significantly increase survival rates for those with this type of cancer.”
Currently, common tests for colon cancer, such as the colonoscopy, are invasive and expensive and have their own inherent health risks, making them unsuitable for population-based screening. In contrast, if LICR researchers develop a test for the early detection of colon cancer using protein markers, future tests for this type of cancer could be as simple as having a blood test.
The Melbourne Branch has more than 20 proteomics researchers in various laboratories, each working on different aspects of proteomics. They use mass spectrometers to generate the data used to identify protein markers. These instruments can generate between 10,000 and 15,000 mass spectra per hour, which then needs to be converted into protein identifications. Smaller lab-based computers found it hard to keep up with the large amounts of dataflow.
In 2005, LICR asked Microsoft Global Alliance Partner, Dell, to propose a solution that would deliver greater computing power to researchers working on multiple projects.
The solution needed to meet the following criteria:
Increased processing power of all computers in the cluster.
Easy-to-use interface for researchers.
Simple, centralized management of a large cluster system.
Compatibility with algorithms used for proteomics research.
Scalability for future growth
Ability to integrate with existing technology infrastructure.
Ability for researchers to share results throughout LICR’s global network.
Dell suggested building a high-performance computing cluster (HPCC) that would pool each laboratory’s resources to create a system with far greater processing power that could be accessed remotely by all researchers. Following which, Dell joined forces with Microsoft and Global Alliance Partner, Intel, to carry out the project.
LICR joined the Microsoft Windows Compute Cluster Server 2003 Rapid Deployment Program, and Microsoft gave them access to the Beta 2 version of the software, which was released in November 2005. LICR’s proteomics staff visited Microsoft headquarters in Redmond, USA, to contribute to discussions about the product.
“It’s quite common for this sort of computing facility to be isolated from the rest of the technology environment… but with Microsoft, the computing cluster can be incorporated into the existing environment.”
Chris Green, Technical Specialist, Microsoft
The project began as a proof of concept, funded jointly by Dell and Intel. The proof of concept was completed in late 2005 and comprised a 16-node computing cluster.
“Combining Dell and Intel’s deep pool of technical resources and expertise to build the proof of concept helped the Ludwig Institute quickly identify the benefits and challenges of this project, ensuring hardware and software work harmoniously to deliver the best result for the customer.” Ivan Chan, Intel’s Dell Regional Account Manager for Asia Pacific.
The ease of use of Microsoft Windows Compute Cluster Server 2003 made it the obvious choice for the Ludwig Institute. LICR was also impressed with the power of Dell’s hardware. The powerful dual-core technology in Intel Xeon processors enabled LICR to process more data within a shorter period.
LICR soon discovered that researchers in other laboratories at different institutions in Australia were faced with the same problems as the LICR team. Thus, in 2005, it formed a consortium with 21 other Australian proteomics research centers to discuss the creation of the Australian Proteomics Computational Facility (APCF). The idea was to take the knowledge gained from the LICR project and expand this on a national scale.
Acting as a national test bed, LICR rolled out the solution, using a configuration with one head node to provide a cluster for production and the other head node for testing and development. The cluster connects researchers from at least 20 different locations across Australia, from Perth to Brisbane and beyond, through a single administrative interface.
In January 2006, the APCF was awarded an enabling research grant of Aus$ 2 million by the National Health and Medical Research Council to create a high-performance computing platform to assist other research centers with high-throughput proteomic analysis.
The APCF will process the raw mass spectrometry data using algorithms such as Mascot from Matrix Science, Phenyx from GeneBio, Sequest from Thermo and many other open source programs. These algorithms transform raw molecular data into a form that biologists can understand.
LICR is also developing its own algorithms using Microsoft Visual Studio .NET.
“As my experience is in biology, not computers, a good compiler and an intuitive development environment are important to me.” says Eugene Kapp, a computational biologist at the Proteomics Laboratory, LICR.
Faster Processing Times: The processing power of the high-performance computing cluster is at least 20 times faster than that available when researchers worked on isolated systems. This has increased the productivity and output of researchers. The servers in the HPCC can work on several individual tasks or combine to complete one large job.
Integrates with Existing Infrastructure: Microsoft Windows Compute Cluster Server 2003 integrates with the Melbourne Branch’s existing IT environment, including Microsoft Active Directory and Terminal Services for remote access.
Cost-Effective Computing: The solution is based on a pay-as-you-grow pricing structure, enabling LICR to expand the solution as it needs and as funds become available.
“The solution had a fairly small learning curve, so we didn’t need to increase staff numbers or conduct extensive training to manage the system,” says Moritz. “We have also reduced our support and maintenance costs.”
Centralized Management: Microsoft Windows Compute Cluster Server 2003 features a centralized management interface that allows the entire HPCC to be managed from one location, increasing ease of implementation and simplifying ongoing management, communication and monitoring of the progress of the research.
LICR is now working towards expanding the solution to give proteomics researchers across Australia access to the HPCC at the APCF. This would involve growing the computing cluster to 128 nodes (with 256 dual-core processors), making a total of 135 servers including the head nodes.