Winship's doctors are gathering those elusive patterns into a computerized database, a kind of vast molecular fingerprint bank of cancer. What the doctors at Winship and other institutions are hoping is that the marriage of genetics and powerful computers -- lots of them -- will strip away one of the most frustrating problems about treating the disease: the overwhelming computational difficulty of figuring out which of the countless combinations of treatments to apply to a particular patient. Doctors know that cancers differ, that leukemia, for example, is a disease with a maddening mix of subtypes that respond differently to different drugs in different people. And yet physicians still use what Sagar Lonial, a 34-year-old Winship oncologist, calls "the nonspecific sledgehammer approach" to treating the disease. Essentially, says Lonial, you walk in the door with acute leukemia and you are going to walk out with roughly the same initial therapy that 90 percent of the other acute leukemia patients are getting. Then your oncologist will try to home in on what works best against your cancer through trial and error. It is an excruciatingly slow, hit-or-miss approach. Which may explain why the chances are about 3 in 4 that if you are diagnosed with leukemia today you won't be around in five years.

Lonial and his colleagues think they can improve those chances with what they are calling "tailored therapy" -- treatments customized for individual patients based on the specific genetic characteristics of their cancer. Within a year, a patient will walk into Winship's first-floor outpatient clinic and an oncologist will take a biopsy -- a tiny snip of solid-tumor tissue or a sample of diseased bone marrow or blood. The sample will be fed into a machine called a microarray, which resembles a souped-up ink-jet printer. Instead of spraying ink on paper, the device spits thousands of dots of tumor DNA onto special glass slides. Lasers and computer software churn through the slides, looking for the exact genes in the patient's tumor that are running amok.

The rogue genes then are assigned a digital code and ride the Internet downtown to the offices of NuTec Sciences, a company that made its name in computer-based oil exploration. NuTec, based in Atlanta, is assembling what it says is the world's largest privately owned computer cluster: half a floor of IBM pSeries 640 servers, 1,250 in all, grinding away in parallel, with the capability of doing 7.5 trillion calculations per second -- enough processing muscle to model a nuclear explosion. The computers will analyze the Winship patient's tumor for its genetic blueprint, scanning what eventually will be a vast collection of databases of gene combinations and information about which drugs have worked best against cancers with similar genetic makeups. The end result will be a swift determination of the best genetic match of disease to drug. The whole process can be wrapped up in less than a day, compared with months for the hit-or-miss approach. When the system is fully operational, Lonial says, "someone will come to me, I'll run them on a slide, and in 24 hours, they're done -- we'll be able to recommend treatments with the most probability of success."

For years, technologists have dreamed that information technology and biotechnology would someday converge into one seamless superscience that could crack the molecular code of disease and yield a gold mine of new treatments and cures. It always seemed so logical, even if it never quite seemed to happen. Some very big names in tech -- Bill Gates (MSFT), Paul Allen, and Jim Clark, among others -- for years have been placing bets on so-called convergence companies that promised to exploit the merging of computing and biotech. Allen alone has investments in more than 50 of them, mostly obscure companies that use words like "genomics," "bioinformatics," and "proteomics" to describe what they do. This industry is so new it hasn't settled on a single name yet.

Now, like a middle-age actor who has just been discovered, convergence has hit the big time. Corporate giants such as IBM (IBM) and Compaq (CPQ) are pouring $100 million dollops of cash into "life science" projects that mesh computers and biotech. The market for DNA microarrays grew to $527 million last year from $145 million in 1999, and Front Line Strategic Management Consulting, based in Foster City, Calif., estimates it will surpass $4 billion by 2005. The number of new companies peddling software tools to guide pharmaceutical and biotech companies through the genome is exploding. "There are hundreds of companies in a space that didn't exist five years ago," says G. Steven Burrill of Burrill & Co., a merchant bank in San Francisco with about $150 million invested in 40 life science companies.

All this enthusiasm raises an inevitable question: Why now? Investors have lost huge amounts of money over the years betting on the unfulfilled promise of miracle cures from convergence. The big difference today, of course, is the decoding of the human genome, announced in June 2000 by government and private researchers after years of work. In one sense, mapping the genome has turned out to be something of an ego-buster for the human species; we now know that we have as few as 30,000 genes, barely twice the genetic inventory of a fruit fly.

But the real message in the relatively modest number of human genes is that, in the micro world of molecular biology, genes are the couch potatoes. Basically, human genes are made of tiny pieces of DNA -- there may be as many as 3 billion DNA bits in our 30,000 genes -- that sit around ordering cells to make this or that protein. By some estimates, the human genome may be capable of producing, or "expressing," as many as 300,000 proteins. Those proteins are the body's real heavy lifters. They help digest breakfast. They provide the brain's circuitry. They also turn on -- and off -- cancer cells. Stanley Fields, a geneticist at the University of Washington and the Howard Hughes Medical Institute, says decoding and studying the genome has given us the protein parts list a cell might use. The next task, he says, is making sense of how genes and proteins interact at the cellular and molecular levels. That is the path to new treatments, but it is a massively computation-intensive job.

Convergence companies are working to create the technologies that will meet researchers' demands. One of the most interesting of these companies is Rosetta Inpharmatics (RSTA), which uses proprietary software and hardware to decipher gene functions and the way they influence disease. (In current industry parlance, Rosetta is known as a bioinformatics firm.) Stephen Friend, Rosetta's president, says the company grew out of a conversation he had five years ago with Lee Hartwell, head of Seattle's renowned Fred Hutchinson Cancer Research Center. The two were driving to a medical conference in eastern Washington, Friend recalls, when they started wondering aloud whether there wasn't a better way to look at how genes interact. Up to then, biologists had been laboriously pairing one gene against another, studying the resulting interactions, and then moving on to the next pair. Not only was the process slow, Friend says, but it also produced a kind of tunnel vision in researchers.

Friend, a clinical oncologist who headed the Hutchinson's molecular pharmacology department, says biologists have this way of focusing on a subject as tightly as possible. They will learn everything about, say, a single protein within a single cell -- and leave the cell's 50,000 other proteins for someone else to worry about. For example, according to Friend, over the years some 11,000 papers have been produced by biologists on a single protein, P53; he says that at one point more than 5,000 biologists were studying just that one molecule.

Mathematicians, on the other hand, tend to look through the other end of the glass, searching for broad patterns in masses of data. Friend and Hartwell hit on the idea that computer-driven microarrays might combine the best aspects of both sciences. Researchers could use the computer's brute power to simultaneously study thousands of genes and the proteins that drive them, looking for patterns in the way they interact. In 1997, Friend and Hartwell patented a process by which microarrays can analyze genetic activity. But microarrays in those days were relatively clunky machines that etched DNA onto specialized computer chips. The chips cost more than $2,000 each. Friend calculated they'd need at least 10,000 of the chips to start the genomics business they had in mind.

Hartwell asked his friend Lee Hood, then at the University of Washington, for help. Hood is a biotech superstar who had been involved in designing microarrays for the better part of a decade. One of his postdoctoral researchers had come up with a way to modify an ordinary ink-jet printer and turn it into a microarray. Instead of spreading ink on paper, the device sprayed nucleotides -- the building blocks of DNA -- on inexpensive glass slides in the form of tiny dots. These DNA nucleotides, as many as 50,000 on a slide, act as docking stations for any complementary genes. Spray a snippet of DNA from a tumor tissue over the nucleotides and they begin pairing off like college students at a beach party. They mix with fluorescent dye on the slides, enabling a laser scanner to pick out patterns of the most active genetic pairings by their glow. Friend describes the process as "having little reporters in the cell sending back bulletins about what happened with this gene and that gene." A computer analyzing the glowing patterns of the tumor sample can create a genetic fingerprint that suggests which therapies to apply to which cancers.

Friend still chuckles at the clumsiness of his pitch to venture capitalists. He waved around a half-dozen glass slides and outlined ideas on napkins. But it was enough to eventually persuade Allen to buy 12 percent of the company for a few million dollars to get it off the ground. It was a good investment: This July, in a landmark deal for convergence, pharmaceutical heavyweight Merck & Co. (MRK) bought Rosetta for about $540 million in stock.

Some of the hottest new convergence companies cluster under the banner of proteomics, the study of proteins and their cellular roles. Here researchers confront the genome at its most daunting: Sorting out which of our 300,000 or so proteins does what to which of the 3 billion bits of DNA that make up the genome -- and how those interactions trigger or inhibit disease -- will keep researchers and their heavy-duty computers busy for years to come. But scientists already see progress. For instance, Sangamo BioSciences, based in Richmond, Calif., is using proprietary software and computer models to explore ways to put to work a class of proteins called Zinc-Finger Proteins. ZFPs, as they're known, regulate gene expression by finding a specific DNA site among the billions in the genome and turning it on or off. Edward Lanphier, Sangamo's president and chief executive officer, likens the process by which a ZFP zeroes in on its DNA target to directing a cruise missile down a smokestack.

Sangamo and Edwards Lifesciences (EW), a medical-device company in Irvine, Calif., are working with researchers at Yale to use ZFPs to trigger angiogenesis, the process by which the body grows new blood vessels. Consider the possibilities: Designer proteins to speed up the healing of wounds. Or to cure angina -- chest pain that results from reduced blood flow to the heart -- without bypass surgery. Angiogenesis is also critical to feeding, or starving, tumors. Frank Giordano, the Yale assistant professor of medicine in charge of the ZFP project, has already cloned ZFPs into viruses and inserted them successfully into the heart and muscles of mice, activating vascular endothelial growth factor, or VEGF, a protein that stimulates angiogenesis. Computers, he says, are the critical element, both in analyzing DNA and understanding VEGF.

It's still a long jump from growing blood vessels in mice to doing it in people. Indeed, for all the excitement surrounding convergence, it pays to remember that, to date, almost all genetics-based therapies that looked great in the lab fizzled in the real world.

Still, the speed with which convergence is opening up new vistas is dazzling to some researchers -- not least Stephen Friend. He says he never dreamed that some idle pondering during a car ride would lead so swiftly to Rosetta's breakthrough technology -- and create a company worth hundreds of millions of dollars. Nor did he ever expect to leave his plum position at Hutchinson, one of the nation's premier cancer research institutions. But, he says, in academia he couldn't have had the impact that he believes Rosetta will have on real-world problems. To explain why, he tells the story of the hunt for an alternative to cyclosporine, an immune-system suppressant used to treat organ transplant patients. Cyclosporine was significantly toxic. In the late 1980s, several large pharmaceutical firms, including Japan's Fujisawa Pharmaceutical, got interested in a compound called FK506, which seemed to promise immunosuppression with fewer side effects. After years of clinical trials, costing several hundred million dollars, the verdict was handed down: FK506 wasn't dramatically superior to cyclosporine.

Friend says that after watching this costly exercise, he had Rosetta's mathematicians run their own FK506 tests on their computers. It took Rosetta only six weeks to come to the same conclusion the pharmaceutical firms did. "Using the genome as a sensor pad, we could have saved those guys hundreds of millions of dollars," Friend says, shaking his head.