Computers capable of mimicking life have long been the stuff of sci-fi nightmares-think The Terminator or 2001’s HAL 9000. But for researchers struggling to make sense of vast amounts of new biological data, and for drug companies anxious to cut costs and speed development, having accurate computer simulations of living systems is still a dream. To make that dream come true, they are turning to “in silico biology,” building computer models of the intricate processes that take place inside cells, organs, and even people. The ultimate goal: an entire organism modeled in silicon, allowing researchers to test new therapies much as engineers “fly” new airplane designs on supercomputers.

For more than a decade, medicinal chemists have tried to make drug discovery more “rational,” using computers to simulate how, for example, a new drug molecule binds to a receptor. But today’s computer models go far beyond that, taking advantage of data from areas ranging from genome sequencing to clinical trials to look at how a potential drug affects entire biological systems. Creating a “virtual cell” or, better yet, a “virtual heart patient” is still a work in progress, but even early models could begin to put a dent in the massive cost of developing new drugs.

According to industry figures, using traditional methods takes an average of $500 million and 15 years to develop and test a drug; in silico technologies could save at least $200 million and two to three years per drug, according to a recent PricewaterhouseCoopers report. One reason is that the drug-testing process-during which a compound is studied in animals, and then in humans-is far from efficient. According to statistics from the U.S. Food and Drug Administration, human trials fail for 70 to 75 percent of drugs that enter them. Some trials fail just because the dose is wrong.

To drive home just how inefficient such a trial-and-error approach can be, Thomas Paterson, chief scientific officer of Menlo Park, CA-based Entelos, makes this comparison: “If Boeing developed aircraft the way the pharmaceutical industry develops drugs, they would develop 10 very different aircraft, fly them, and the one that didn’t crash would be the one they sell to United Airlines.” So companies like Entelos and Princeton, NJ’s Physiome Sciences are developing computer models that can be used both to identify molecular targets for new drugs and also to simulate clinical trials. For example, the Leverkusen, Germany-based pharmaceutical giant Bayer is using one of Entelos’s models to evaluate a potential drug for asthmatics, testing a variety of patient types and treatment regimens on the computer.

The Internet could become a critical tool in developing such models, allowing researchers to collaborate around the world. So Physiome has partnered with the Bioengineering Research Group at New Zealand’s University of Auckland to develop an open-standard computer language for biological modeling. That language, called cellML, is available at www.cellml.org. The idea, says Physiome executive vice president Thomas Colatsky, is that researchers will be able to build models in a common format and share those models via the Web.

Still, many believe it’s premature for drug researchers to begin setting their lab rats free. Leslie Loew, a member of the cellML advisory board and the director of the Center for Biomedical Imaging Technology at the University of Connecticut Health Center, has made his own modeling tool kit accessible on the Web: the Virtual Cell, at www.nrcam.uchc.edu. Within five years, Loew predicts, modeling software will be a routine, perhaps indispensable, tool for anybody who seeks to understand how cells work. But, Loew cautions, it will still take many years to build complete, highly accurate models of whole cells, let alone organs or entire organisms. And bioinformatics professor Masaru Tomita-whose group at Keio University in Fujisawa, Japan, has put its E-Cell simulation software on the Web at www.e-cell.org-agrees. While E-Cell does aim to model whole cells and, eventually, interactions among a dozen or fewer cells, Tomita says modeling anything more complex “would be a whole different ball game.”