Drug companies typically spend millions of dollars on clinical trials to find new drugs or other treatments.
It's an expensive process because they often have to enroll hundreds or thousands of patients in the trials, split them into groups that receive the new treatment or a placebo, and then monitor them for months or years to see if the therapy works.
But what if those trials could be done more accurately and quickly on a computer, using a set of virtual patients?
That's the potential that exists with so-called "in silico" biological research, in which both healthy and diseased cells are modeled on a computer.
One of the scientists pioneering that approach is Dr. Yoram Vodovotz of the University of Pittsburgh's McGowan Institute for Regenerative Medicine. This week, he received an IBM Shared University Research award, which will give him an IBM water-cooled supercomputer which he can use to model the inner workings of the human body.
The award, with a dollar value of about $600,000, will give Dr. Vodovotz and his team access to an IBM Power 575 supercomputer, which has 32 processing nodes that eventually could be expanded to 448 nodes. Each of the nodes is more powerful than the dual-core processors found in today's speediest PCs.
Ultimately, these supercomputers could be used to model "all the different types of cells and tissues in the body and be able to assemble that into a whole human being," said Jim Spohrer, director of IBM's global university programs.
Dr. Vodovotz's team already has developed computer algorithms that model what happens to liver cells infected by hepatitis, lung cells infected by viruses and skins cells that develop pressure ulcers.
His current work focuses on sepsis, one of the leading killers of people in intensive care units. Sepsis occurs when the body's own healing mechanism runs out of control and shuts down the function of vital organs.
When the body experiences a serious infection or trauma, Dr. Vodovotz said recently, it responds with inflammation, which is helpful and necessary, up to a point.
But sometimes, the inflammation process goes haywire. In those cases, he said, "inflammation ends up causing some damage to bystander tissues, and these scream out saying we've also got a problem, and then you sometimes get into this self-fulfilling amplifying loop" in which the inflammation keeps spreading.
That cascade can be lethal. If the sepsis escalates to septic shock, where more than one organ shuts down and blood pressure plummets, nearly half of such patients will die, according to a Web site for Xigris, an Eli Lilly anti-sepsis drug.
Xigris is the only medication approved by the Food and Drug Administration for severe, life-threatening sepsis right now, Dr. Vodovotz said.
To reach that point, drug companies tested more 250 other candidate medications. Many of those substances didn't pass muster with the FDA because, while they helped some sepsis patients, they harmed others, so that there wasn't a clear benefit, Dr. Vodovotz said.
That's one area where computer modeling could be a big boon, he said. By simulating what would happen to individuals with certain biological profiles, the modeling could predict which ones would benefit from certain medications and which ones wouldn't.
That could help drug companies identify which people might profit most from a potential new drug before enrolling them in a clinical trial, he said, and make drug discovery less of a hit-or-miss proposition.
"Simulating [patients'] biology in an increasingly sophisticated fashion potentially lets you say, 'Let's make this trial as likely as possible to succeed.' "
There might eventually be situations in which scientists could conduct "virtual clinical trials" using only simulated patients.
For instance, he said, in trying to develop treatments for anthrax infections caused by bioterrorism, it would be unethical to infect volunteers with anthrax, but if computer models were reliable enough, the infection and treatment could be tested in that way.
The great advantage of computer modeling, Dr. Vodovotz said, is that it can keep track of several variables at once. "Once you get past three or four variables," he said, "your human brain has a very hard time dealing with that, but the computer doesn't care -- it just may operate a little more slowly" in hashing out the answers.
The technique someday might be used in the opposite direction, he said -- rather than estimating which patients would benefit from an existing substance, it might simulate what goes wrong in a certain disease, and then indicate what kind of drug or treatment would most likely effect a cure.
IBM's Dr. Spohrer said his company believes in the potential of computer simulation, and hopes it will one day be as widely accepted as computerized modeling of aircraft design or robotic surgery is today.
"Fifty years ago, we designed airplanes in wind tunnels, but now it's done on supercomputers," he said. "In some sense, those big wind tunnels are like dinosaurs."
In the same way, computers hold the promise of enhancing or replacing much of the work now done in standard laboratories.
In coming years, he said, many scientists may "shift to more in silico experiments, where computers will replace the wet labs."
Mark Roth can be reached at firstname.lastname@example.org or at 412-263-1130.