(FORTUNE Magazine) – At first glance, genomics startup Perlegen Sciences seems a world apart from Google, the celebrated Internet search-engine company. But a closer look shows striking parallels: Both are Silicon Valley companies based on technologies for plucking gems of information from confused jumbles. Both chose to stay private while peers rushed out initial public offerings. Both have thrived post-bubble and are expected to engender monster IPOs.

But here's the coolest karma they share: Perlegen, like Google, has what it takes to rekindle excitement about the Promethean power of high-tech tools.

The company was spun off in 2000 from Affymetrix, the Santa Clara, Calif., biotech that pioneered "DNA chips," a breakthrough technology for probing genes. Venture capitalists pumped $100 million into the startup, jumping its valuation to over $250 million. The big numbers reflected a bold vision: Perlegen's aim is to apply Affymetrix biochip technology to scan thousands of human genomes and thereby illuminate DNA variations that render people genetically different. That could ultimately make biochips as important as computer chips.

Last fall Perlegen disclosed stunning progress: It had decoded the complete DNA of 25 people with custom-made biochips that enable it to speed-read a person's genome in about ten days. (Decoding a genome means determining the sequence of chemical letters in the DNA strands, or chromosomes, coiled in the nucleus of every one of our cells; recall that the international Human Genome Project took $2.7 billion and more than a decade to sequence the genome for the first time.) The news of Perlegen's rapid-sequencing feat brought Big Pharma flocking. The company quickly landed deals with Bristol-Myers Squibb, Pfizer, and other giants to study genes underlying patients' differing responses to drugs.

Among other things, Perlegen's collaborators hope to use the startup's technology to select patients for drug trials whose gene variants make them likely to respond well. "If we're starting with patients we know have a good chance of responding, clinical trials can be smaller, faster, and cheaper," says Allen Roses, senior vice president of genetic research at GlaxoSmithKline, a London drug company that was Perlegen's first client. "That will dramatically lower the cost of drug development." Patients could also benefit when the new drugs reach the market, for their use could be guided by the same tests developed to select good responders for clinical trials.

True, genetically targeted medicines are likely to have smaller markets than conventional "one size fits all" drugs. But the promise of preordained efficacy would afford a sales hook to offset that disadvantage--most medicines available today have only a 30% to 40% chance of working for a particular patient. "Despite all the talk about the need for new billion-dollar drugs," says Perlegen CEO Brad Margus, "behind closed doors [drug industry] people are telling me that if the choice is getting five drugs to market with $500 million each in annual sales instead of one with a $2 billion market, they'll take the former."

Gene-guided drug development also promises to help avert severe side effects. The value of that can hardly be overstated--getting blindsided by adverse reactions is the bane of the pharma industry. Rare side effects, typically caused by uncommon gene variants, often can't be clearly linked to medicines that trigger them until hundreds or thousands of people have been treated--that is, after a fortune has been invested in clinical trials.

One example is Propulsid, a Johnson & Johnson medicine approved in 1993 for people with severe nighttime heartburn due to "gastroesophageal reflux" disease, which causes stomach acid to erupt painfully into the esophagus and throat. Cardiac problems were linked to Propulsid after its approval, and in 1998 its label was revised to reflect the concern. The following year a stronger boxed warning was added. But scattered reports of severe cardiac effects, including deaths, continued to flow in even as Propulsid's sales rose to about $1 billion a year. In July 2000, after tens of millions of Americans had taken Propulsid, J&J stopped marketing it--it's now available only through a limited-access program. When J&J announced it was withdrawing the blockbuster, the company's market value plummeted nearly $10 billion, about 10%.

Drug flops due to such belated surprises are a major reason that pharma companies wind up shelling out about $800 million, on average, for each new medicine they get to market. But imperiled drugs could be rescued if tests were available to identify the patients whose uncommon "genotypes" put them at serious risk. Perlegen doesn't plan to sell such tests itself. Instead it will provide the information needed to develop them. For instance, it might scan the genomes of patients in early clinical tests of a new drug in order to identify DNA variants linked to abnormal heartbeats that occur in a small fraction of those who take it. That information could then be used to develop a fast, cheap genetic test that analyzes only the tiny fraction of the genome relevant to the cardiac side effect. The test would be used to screen patients before they're enrolled in the drug's large, final Phase 3 tests, enabling those at risk to be excluded.

Alternatively, if Perlegen's research showed that the DNA variants linked to the heartbeat problem occur in more than a small percentage of the population, the drug company involved might opt to cut its losses and drop the risky medicine. "We're going to save far more money using the new technology to kill drugs off early than we are by changing Phase 3 plans" to exclude poor responders, opines Rick Hockett, a senior clinical researcher at Eli Lilly & Co., one of the companies working with Perlegen. "We call it quick kill, quick win."

Perlegen's flurry of deals--Lilly has even taken an equity stake in the firm--suggests that the age of genetically guided drug research, or pharmacogenetics, isn't far off. At Glaxo "not only is pharmacogenetics coming down the pike, it's down the pike and has already changed the way we develop drugs," says Roses. And top FDA officials gave the trend a boost in May by publicly endorsing the idea of testing new medicines in genetically screened patients.

Whether this rising tide will lift Perlegen to profitability remains to be seen. A number of its predecessors in pharmaco-genetics have wound up as tech-bust rubble, much like pre-Google search-engine firms. DNA Sciences in Fremont, Calif., for instance, was launched with fanfare in 1998 to offer pharmacogenetics services. Boasting a board that included Nobel laureate James Watson, co-discoverer of DNA's structure, and Silicon Valley entrepreneur Jim Clark, it raised $110 million in venture capital. But gene-scanning technology was still too slow and costly to handle the large number of complex scans needed for in-depth studies, and expected revenues never materialized. DNA Sciences recently agreed to sell its assets to Genaissance Pharmaceuticals of New Haven for a miserable $1.3 million.

What makes Perlegen different?

One thing is its unequaled capacity for "whole-genome scanning"--the ability to rapidly scrutinize key sections of DNA across the entire genome. Several technologies can do that, but Perlegen's allows by far the most detailed scans. The core of its Affymetrix-based technology consists of five-inch-square glass "wafers" whose surfaces are studded with 60 million DNA probes, molecules designed to register the sequence of chemical letters that make up the genome. Perlegen's largely automated decoding process involves extracting genetic molecules from human cells, precisely replicating the molecules 275 billion--fold, and squirting the resulting samples onto the wafers. The probes then light up under a laser in patterns that reveal the DNA letters they have detected.

Only about 0.1% of the genome's letters matter in the quest for important genetic quirks. This key fraction consists mostly of single letters, called SNPs and pronounced "snips," that tend to vary from one person to another. (About 99.9% of any two genomes are identical.) By decoding the entire genomes of 25 anonymous volunteers last year, Perlegen was able to determine precisely where such variable letters occur along DNA's long, chainlike molecules. That, in turn, enabled the company's technical partner, Affymetrix, to churn out wafers for Perlegen that decode only SNPs. (Decoding all 3.2 billion DNA letters in the human genome is still far too costly for studies involving lots of people.) By scanning about 1.7 million SNPs scattered across a person's genome, Perlegen can capture most of what makes him or her genetically unique. The job requires just two of the $35,000 wafers, whereas decoding a person's entire genome requires 220 wafers, enough to cover a queen-sized bed.

For all its tools' Google-esque coolness, Perlegen's most important edge may be its people. Chairman and co-founder Stephen Fodor co-invented Affymetrix's seminal technology and, as CEO, built Affymetrix into the leader in the still-nascent biochip market. (Last year the company posted revenues of $290 million and a net loss of $1.6 million.) Alejandro Zaffaroni, a Perlegen investor and director, has founded a raft of major biotechs, including Alza, which Johnson & Johnson bought in 2001 for $10.5 billion. Advisor Paul Berg, a Nobel laureate, helped pioneer gene splicing. Chief scientist David Cox formerly co-directed the world-class genome center at Stanford University. Chief information officer Greg Brandeau hails from Pixar Animation Studios, where he led the massive computer operation behind Toy Story.

The background of Perlegen's CEO stands out most--he formerly ran a seafood wholesaler in Florida. Brad Margus was no ordinary fish guy, though. In 1993 two of his sons, then 2 and 4, were diagnosed with a rare inherited disease called ataxia-telangiectasia, or A-T, that causes loss of motor control, immune deficiency, and other ills. (Margus and his wife, Vicki, both carry the recessive gene that causes the disorder but don't have it themselves.) After learning that the disease's rarity made it a low priority in medical research, the Marguses started a nonprofit foundation, the A-T Children's Project, to sponsor studies on it.

Margus, a Harvard MBA, soon scouted out an ideal candidate to lead the foundation's scientific advisory board: Cox, who is both a pediatrician and a geneticist and who was then at Stanford. "I was ready to use every trick in the book to get him," says Margus. "I called him and said I was going to be in Palo Alto the next week. 'Could I stop by for a few minutes?'" When Cox acquiesced, Margus dropped everything and booked a flight to California. (He'd actually had no plan to go.) Having initially declined to advise the charity, the pediatrician found it impossible to refuse the determined father's appeal.

In 2000, Affymetrix CEO Fodor lured Cox from Stanford to help form Perlegen. Asked about possible CEOs, Cox knew just which never-say-die, street-smart, gene-obsessed entrepreneur he wanted. Margus's years of pushing gene researchers to get a move on--he even testified in Congress and was interviewed on ABC by Barbara Walters about the A-T campaign--made him a natural for leading an expedition into the great beyond of the genome.

Perlegen's dispatches from the frontier show it is indeed ranging fast and far. Among other things, it has identified gene variants that make some alcoholics unusually prone to liver damage--someday gene tests may allow doctors to warn patients for whom drinking poses a high risk. Perlegen is also spearheading a drug industry consortium on the genetics of "syndrome X," a widespread condition that leads to diabetes, heart disease, and other illnesses in middle age. The project could lead to a whole new generation of treatments and drugs.

Best of all, says Margus, "we've executed on plan and accomplished everything we said we would over the last two years." That doesn't mean Perlegen will soon be profitable. But given that Promethean innovators are known for having their livers torn out--nowadays by bears, not eagles--simply making steady progress over multiple years seems demigod-like. Maybe even Google-like.

FEEDBACK dstipp@fortunemail.com