ACADEMIA

Researchers in MIT's Computational and Systems Biology Initiative (CSBi) are using Interactive Supercomputing Inc.'s (ISC) Star-P software to create new biological models that may someday yield new drug discoveries. CSBi is taking a new cultural approach to research by creating multi- disciplinary teams of biologists, engineers and computer scientists to systematically analyze complex biological phenomena. This unique multi- disciplinary approach places equal emphasis on computer simulations and lab experiments, and on molecular and systems views of cellular function. With Star-P, CSBi researchers from all disciplines can now tap powerful high performance computers (HPCs) to dramatically accelerate these analyses, without having to become parallel programming experts. In one key application, the team can create visualizations of microscopic cell images under different drug treatment conditions and cluster the cells based on attributes. Previously, the enormous datasets required to do analysis and generate visualizations of cell cluster models had overwhelmed the capabilities of researchers' desktop computers. The computational calculations utilized K- means, a powerful algorithm used to cluster objects. With Star-P, researchers can create parallel K-means models on their desktop PCs using MATLAB, and run them automatically on SGI Altix and Opteron-based parallel servers. CSBi's strategy gathers experts from a variety of experimental and numerical methods and partners them with commercial companies in information technology, proteomics, genomics and bioinformatics. Making supercomputing capabilities accessible to these non-computer programming experts was important to the initiative's success. "Accelerating computational processes associated with data mining and modeling enables us to find problems, optimize and ultimately refine these models more quickly. Creating the best possible computational models in the shortest time gives us significant predictive power and improves the effectiveness of experimental work," said James G. Evans, CSBi research scientist for Imaging and Informatics. CSBi researchers are also considering using Star-P for applications such as doing statistical analysis in preparation of cellular visualizations during high content imaging. Star-P is an interactive parallel computing platform that eliminates the need to re-program the biological models in C, FORTRAN or MPI languages to run on parallel servers. Star-P lets CSBi users continue to work with their preferred tools, languages and desktop environments, shielding them from the programming complexities of parallel systems.