Cancer: a complex material
|
|
|
|
|
|
By Steven Schultz
When Sal Torquato received a collaboration proposal from a cancer researcher at Harvard, he was skeptical.
What use would a physician who studies brain tumors have for a professor of civil and environmental engineering? As a faculty member of the Princeton Materials Institute, Torquato is a theoretician who studies the physical properties of complex materials such as blends of ceramics and metals, colloids, foams and geologic materials.
But that's exactly what Thomas Deisboeck of Massachusetts General Hospital's Brain Tumor Center wanted. Stymied by conventional methods for investigating tumor growth, Deisboeck needed a fresh approach. His idea was to build a virtual brain tumor, a computer model that describes how the tumor lives and grows--and how it could be eliminated--based on the theories that govern the behavior of other complex materials. If a computer could predict how cancer spreads, Deisboeck reasoned, it could save lives.
"I was skeptical in the beginning. Here was this guy contacting me out of the blue," recalls Torquato. "It didn't take me long to decide."
Eclectic work
That was in October 1997. Since then Torquato and Deisboeck have experienced both the benefits and the challenges of such a far-flung collaboration. In an important affirmation of their approach, they received in July a grant from the National Cancer Institute of National Institutes of Health. "It's really the first time the NCI is funding computational work of this kind," says Torquato.
The grant was awarded under a new program aimed at fostering such eclectic work. "It's one of those high-risk, high-payoff things," says Dr. Dan Gallahan, program director in the Division of Cancer Biology at NCI. "Though a lot of mathematical models of cancer have been generated in the past, this marriage of a materials scientist and an oncologist was very exciting for us."
IBM Corp. also awarded an equipment grant to the project.
Although their investigations are just beginning, Torquato and Deisboeck already have developed a preliminary simulation and have had one paper accepted for publication in Biosystems. They have submitted another and have two more in preparation.
But perhaps the biggest accomplishment has been to establish the collaboration in the first place.
"We speak different languages," says Torquato. "Part of the process has been to communicate our ideas to each other. We feed off each other in a very complimentary way."
For example, as Torquato asked Deisboeck about the properties of tumors so he could factor them into a mathematical model, he raised questions that physicians had not considered before. And his own ideas about modeling have been stretched by the deep complexity of human biology.
Fortunately for Torquato, he was in a position to take time for such reflection. Last year he was a member of the Institute for Advanced Study, on leave from Princeton with a Guggenheim Foundation fellowship.
New level of exploration
For Deisboeck, interdisciplinary research is becoming more and more crucial for finding the approaches that will push exploration to a new level. But he notes that establishing such a collaboration is difficult and risky. Funding is hard because most grant agencies are geared toward evaluating a particular area of science. There are relatively few scientists qualified to participate in peer review of grants and papers dealing with such large-scale multidisciplinary projects. Above all, he says, it's hard to be sure that the collaborators share the same goals for their research.
In the case of his work with Torquato, however, Deisboeck says that was never in question. "There was an immediate understanding. And the initial step has been so promising that I hope it will be the start of a long collaboration." Torquato also envisions that this research will comprise a significant part of his activity in the future.
The medical problem Deisboeck and Torquato are addressing is one of the most deadly and enigmatic human diseases. Highly malignant brain tumors are almost always lethal, and they kill in a short period of time. The average survival after diagnosis is about eight months; few patients survive two years, and those that do suffer from the consequences of surgery, chemotherapy and radiation. And the incidence of brain cancer--about eight cases per 100,000 people--has risen inexplicably for the past several years.
A lot of urgently needed information about brain tumors is hard to gather. For one thing, the tumors grow so fast that doctors see only widely spaced snapshots of their development. Vast amounts of research have resulted in detailed information about the genetics, the biochemistry and the clinical characteristics of the disease, but these data remain isolated and have yet to be fitted into a more complex picture of how and why the tumors grow.
Deisboeck and colleagues at Massachusetts General are developing "a new generation of experimental models in order to investigate such tumors as complex dynamic, selforganizing biosystems," he says. He and Torquato hope to incorporate these data sets into computational models, creating a virtual cancer patient that can be investigated at will.
"These are things they never could hope to do in vivo," says Torquato.
Voronoi tessellation
Torquato started by dividing a simulated tumor into randomly sized volumes, using a process called Voronoi tessellation. Working with information from Deisboeck, he assigned these volumes a series of rules about how they should react with their nearest neighbors. The rules are based on the local concentration of various biological chemicals and other factors. When the simulation gets started, these units of space act as "cellular automatons," growing, dividing and invading neighboring spaces. The resultant model has many characteristics of an actual tumor.
Torquato and Deisboeck intend to keep adding levels of complexity to make the simulation more and more lifelike. Eventually, Deisboeck hopes to plug in characteristics from a specific patient and use the computer model to make predictions about how that patient will respond to certain therapies, or where new tumor masses are likely to crop up after surgery. They also hope that the techniques they develop will prove applicable to other kinds of cancer.
"We're not trying to replace traditional scientific methods," says Deisboeck. "We're trying to broaden the focus." Ideally, the computer simulation will trigger new areas of clinical and laboratory investigations, which will, in turn, add accuracy to the computer model. "Complex biosystems science is sure to have a tremendous impact on biomedical research in the very near future," says Deisboeck, "and this is certainly one of the most visionary projects in this emerging scientific field."
Despite much progress on the research level, Deisboeck does not expect to see such novel computer simulations playing an immediate role in cancer clinics for at least three to five years. "The challenge is enormous," he says. "But the point is that somehow you have to start."
For more information, see project websites at http://cherrypit.princeton.edu/cancer.html and http://brain.mgh.harvard.edu/TumorModeling.
top