ENGINEERING

Breast cancer affects more than two million women in the United States -- a startling 12.5 percent in California alone. Suhrit K. Dey, professor of mathematics at Eastern Illinois University, is teaming with NAS Division computer scientists to create three-dimensional mathematical models for predicting how cancer spreads -- and how it may be contained or even cured. Dey plans to integrate a programming technique called Multi-Level Parallelism(MLP) into his 3-D modeling code to enhance its computational performance on the 512-processor SGI Origin 3800 supercomputer at the NAS Facility. Dey’s mathematical models are based, in part, on his theory that there is a direct correlation between stress levels and the development of breast cancer. Dey is computing one-dimensional cancer models using a set of eight mathematical equations. These models include variables such as the number of lymphocytes (cells in the human body that attack cancer), number of cancer cells, types of medical treatments, angiogenesis (the development of new blood vessels), and glucose levels in the body. “I am building these models based on information I’m gathering in scientific journals, which is qualitatively accurate. Once the models are complete, I will input an individual’s data to provide us with some quantitative results.” According to Dey, this approach will help doctors predict whether an individual will get cancer, or if their body’s immune system can revert the process. Dey has received tremendous support from the medical community. “I feel that Dr. Dey’s approach is both novel and intriguing, and could lead to important developments in this field,” says Dr. Steven Oppenheimer, director of the Center for Cancer and Developmental Biology at California State University, Northridge. What he offers is an overview approach that will give mathematical weight to many specific factors involved in causing, preventing, and treating cancer.” Three-dimensional Approach Dey is currently seeking a grant that will allow him to begin generating 3-D cancer growth models. These models will solve more than 16 million mathematical equations used to describe cancer growth. NAS computer scientists Jim Taft and Rupak Biswas are slated to enhance Dey’s cancer modeling code to run on the Origin 3800, and will assist him with integrating the MLP technique, developed by Taft. Using these enhanced models, Dey will be able to predict the rate of breast cancer growth and/or decay in individuals based on several factors — levels of stress, amount of exercise, and food intake, for example. In order to create a fairly accurate cancer prediction model, all factors or conditions involved, such as the size of the tumor and level of carcinogens in the body, must be considered together, Dey explains. Since MLP is a shared-memory technique (meaning all blocks of work have access to the same area of memory on the computer at the same time) it is ideal for the cancer modeling code, according to Taft. “The kinds of problems that Dr. Dey is doing require very rapid communication between different parts of the calculation, in parallel. MLP will be able to do this very efficiently,” Taft says. “The fundamental construction of Dey’s code structure is very similar to that of NASA’s computational fluid dynamics code OVERFLOW-MLP. Because of these fundamental similarities, we expect the code to scale well with MLP. Collaborative Efforts Are Vital Dey has just returned to Eastern Illinois University after a five-month stint at Ames, and will continue teaming with Taft and Biswas to carry out his 3-D modeling work. Once these models are complete, Dey plans to collaborate with mathematics professor Glenn Webb, Vanderbilt University, to expand on the cancer prediction models. “Dr. Dey is developing elaborate models of tumor growth, incorporating multiple features of tumor cell and immune cell interactions. The most interesting feature of these models is their exploration of the importance of the immune system in controlling tumor growth,” says Webb, who has worked on mathematical modeling of cancer for 30 years. Although these mathematical models for cancer prediction are qualitatively accurate, Dey emphasizes that they are designed only to accompany clinical studies on breast cancer — not to replace them. “We need a combination of mathematical models, statistical models, and clinical studies, so we can see breast cancer from every possible angle,” he says.” "I can see the light at the end of the tunnel. The solution is there, but I need help from others so breast cancer can be contained,” says Dey. “If 10 percent of the people take an interest in this and protect themselves, that’s a large number of women being saved." Visit Suhrit Dey's website or contact him via e-mail at: cfskd@ux1.cts.eiu.edu By Holly A. Amundson NASA Advanced Supercomputing (NAS) Division NASA Ames Research Center ---------- Supercomputing Online wishes to thank both Holly A. Amundson and NASA’s NAS division for allowing us to bring this story to our readers. It would also like to thank NAS’s Jill Dunbar for her assistance. ----------