(FORTUNE Magazine) – Imagine, if you will, that you are an executive, lazily staring out the office window, when suddenly you spot your wife on the sidewalk in another guy's arms. But, staring more closely, you realize something is different. That isn't your wife, exactly, but a nearly identical clone. Weirder still, it's a clone that replicates only the best qualities of your wife. Gone are her crankiness, impatience, and distaste for World War II submarine movies. What remains are her kindness, her companionship, and that killer recipe for peach cobbler. To top it off, the guy on the sidewalk is gesturing madly in your direction, offering to give you this marvelous creature's companionship. For a price. Or maybe sell her companionship to someone else.

This approach, approximately, is what has made Sepracor, a small company in Marlborough, Mass., one of the most promising, best-capitalized pharmaceuticals startups in the country--despite having lost more than $200 million since it went public in 1992. Sepracor's business plan is equal parts ingenuity and brazenness: Take a blockbuster drug made by some other company, eliminate molecules that cause unpleasant or even fatal side effects, patent the molecules that remain, and sell 'em.

Four years ago Sepracor got a patent for an improved version of Seldane, an antihistamine by Hoechst Marion Roussel, that is so much safer that last year the FDA moved to ban the original Seldane. But before Seldane got yanked, Hoechst licensed Sepracor's patents and rushed the improved drug to market under the brand name Allegra. Its sales are likely to reach $600 million this year, three times what Seldane pulled in during 1997.

Sepracor now has promising patents on improved (though not entirely proven) molecules of two dozen blockbuster medicines, including Prozac, Claritin, and Propulsid, and of some lesser-known but potentially huge products including a pill for urinary incontinence and an asthma treatment. The company is merrily standing on the figurative sidewalk, dangling most of the compounds in front of drug company execs, looking for fat royalties or a fifty-fifty share of revenues.

"It's a simple, elegant strategy," says Douglas Lind, a securities analyst at Morgan Stanley. Others on Wall Street agree. Despite losing money at a burn rate of $20 million to $50 million a year for ages, Sepracor has lately seen its shares sell for more than $54 each, giving it a surprisingly big market cap of $1.5 billion. If the drugs now on or close to market succeed (see table), Lind figures Sepracor will earn about $1.20 a share in 2000 and $6 a share in 2001.

The science that gave Sepracor its start is intriguing but not particularly complicated, CEO Timothy Barberich freely admits. Most prescription drugs are made up of fairly elaborate molecules, arranged in a three-dimensional way. (Such a 3-D arrangement is called an isomer.) When the molecule finds a receptor in the body that it "fits" tightly, like a hand in a glove, it typically turns the receptor off or on, thus thwarting some unhealthy bodily process or microbe run amok, or triggering some beneficial process that cures the illness.

But there is a quirky--and sometimes dangerous--problem with many man-made drugs. Until a few years ago, virtually all the molecules produced in a pharmaceuticals company's factory came out of the spigot as a mix of two isomers: one whose elements are arranged with a rightward rotation, and a mirror-image twin whose elements rotate to the left. These jumbles of left-hand/right-hand isomers are called racemic mixtures. Most of the time, only one of the twins does any good, fitting into the desired receptor. The other version is generally useless. But sometimes it is a downright evil twin, wandering all over the body until it accidentally gloms on to some other receptor, triggering dangerous and even deadly responses. (One of the worst examples was Thalidomide, a drug first sold during the 1960s, whose right-twisting molecules alleviated morning sickness, but whose sinister twins caused horrific deformities in the arms and legs of fetuses.) For years, it was too difficult to remove the offending molecules in most racemic mixtures, so the FDA allowed both twins to remain if the drug was reasonably safe and effective.

Barberich, Sepracor's burly chief executive, figured years ago that those dangerous molecules could be eliminated. In the early 1980s he was a senior executive at Millipore Corp., a specialist in purifying chemicals. He unearthed some university data on using enzymes to create one-handed molecules and started Sepracor in 1984 with the aim of cleaning up drugs for pharmaceuticals companies.

But within a few years Barberich set his sights on much bigger targets. Rather than just tidying up molecules belonging to others, he wanted Sepracor to control some molecules of its own. "It occurred to us that we could extend the market life of blockbuster drugs that were about to lose patent protection, by isolating and patenting a better, safer version of the original molecule," he recalls. With the new strategy, Sepracor transformed itself from molecule janitor to molecule opportunist.

Before long the company expanded beyond twin-molecule problems. It studied the numerous molecules formed when the body breaks down, or metabolizes, a medicine. Often, only one of several metabolites produces the drug's beneficial effect, while unwanted metabolites may lead to unpleasant or deadly side effects. By zeroing in on and patenting just the good metabolite of a popular drug, the company found a second way to prosper. Sepracor used this approach to go after the popular new nonsedating antihistamines, such as Seldane, because some varieties were causing fatal heart arrhythmias.

The legal challenges the company faced were more daunting than the scientific ones. "It wasn't clear that part of an existing drug could even get a patent," says David Southwell, Sepracor's chief financial officer. It turned out, though, that Sepracor could get a certain, very useful, kind of patent. Most drug-industry patents are what are called composition-of-matter patents, which give the inventor exclusive rights to make and sell a chemical for a set purpose. However, if someone else discovers that the same chemical can be used for some completely different purpose--using Crest toothpaste to cure dandruff, say--the newcomer can get a "use patent" on Crest ingredients for dandruff.

A use patent is not a perfect victory; the bad news for Sepracor is that the newcomer who gets the use patent doesn't win the right to make and sell the Crest ingredients. For that, he has to get permission from Procter & Gamble. The good news: P&G cannot sell Dandruff Formula Crest without permission from the guy who patented the discovery that it works on scalps. And when P&G's composition-of-matter patents on Crest eventually expire, the newcomer can make all he wants, and has exclusive rights to sell it for dandruff for a decade or more, until his use patents expire.

All of which puts Sepracor in a strange negotiating position with the likes of Johnson & Johnson, Lilly, and other giants whose blockbusters Sepracor has quietly tweaked. Even if the new version of a drug is better, Sepracor can't run out and sell it, but rather must negotiate prices and terms with the company holding the original patents. At first, some in the industry disdained Sepracor as a kind of patent pirate; many companies, especially those with popular, profitable drugs, decided to ignore it for as long as possible. "They don't pay attention until their cash cow is about to die," says Paul D. Rubin, Sepracor's head of drug development.

But when a blockbuster faces the loss of patent protection, or if its side effects are causing financial and ethical problems, Sepracor finds the makers far more willing to negotiate. Seldane, the Hoechst drug, had some of both troubles in the early 1990s. It was going off patent in a few years, and Hoechst had been forced by the FDA to post prominent warnings--a so-called black box--on the label, telling of heart dangers. Hoechst was losing market share to the likes of Schering's Claritin, which had far fewer side effects.

So Hoechst licensed Sepracor's version in 1993. Because the Sepracor drug was "identical to but less than" the original, the FDA let Hoechst use its original research on carcinogenicity and best dose size, shaving a decade or more off the time it takes to get a new drug approved. The new version of Seldane, called Allegra, was approved for sale in 1996. Even better for Hoechst, after it stopped selling Seldane, the FDA prohibited anyone else from selling it too. That meant that all the generic drugmakers who had lined up to produce cheap, off-brand versions of Seldane couldn't compete.

Allegra's success made Sepracor far more respectable. Earlier this year, Johnson & Johnson agreed to work with the company to get FDA approval for a safer version of its antihistamine, Hismanal. J&J is cooperating, too, on developing a safer form of Propulsid, a $1.2-billion-a-year drug and J&J's second-biggest seller. Propulsid is used to relieve a serious form of heartburn, called GERD, but other drugs are usually prescribed first because a number of users have died from heart irregularities in the past five years. If, as expected, Sepracor's improved Propulsid gets FDA approval and becomes the first-line treatment for GERD, sales should climb sharply. J&J officials declined to comment on their work with Sepracor.

One of the elephantine prizes that Sepracor is currently pursuing is Prozac, Lilly's $2.6-billion-a-year depression treatment. Though still the bestseller in its class, Prozac has a minor drawback that has let rivals such as Zoloft and Paxil grow faster in recent years. If you take Prozac and decide it doesn't work well, you have to wait five or six weeks for all the Prozac to leave your body before trying another antidepressant. The others wash out in a week or less. As you might have guessed, Sepracor has a right-twin version of Prozac it thinks will wash out very quickly. It is so fast-acting and quick to depart that Sepracor thinks it could be used to counter premenstrual syndrome for some women. And preliminary trials suggest the left-twin version of Prozac is good too: It appears effective in preventing migraines. Not least, the new versions could keep Lilly's revenues from getting depressed when Prozac goes off patent around 2003. Sepracor is lobbying Lilly to license the molecules now. Lilly won't say much about Sepracor, only that it is looking at several ways to hold its position in the antidepressant market and that it has not ruled out licensing products from other companies.

You'll be pleased to hear that Sepracor is also working on a pill for loss of bladder control. A drug called Ditropan minimizes spasms that cause a form of incontinence for millions of women and men, but it has so many unpleasant side effects, such as dry mouth and dizziness, that it's not used much. Barberich cautions that Sepracor's improved version will take several years of testing to be proved effective. If it works, he figures it will capture much of the $2 billion spent annually on adult diapers and the like and be far more attractive than sending aging parents off to nursing homes. And because the original drug is off-patent, Sepracor won't have to share revenues with other companies.

If teasing out bad molecules from blockbusters is such a simple way to improve medicines, why, you ask, didn't the drugs' owners do it before Sepracor? Michael Tong, a securities analyst at Salomon Smith Barney, says Barberich pounced on the idea before just about anyone else and was shrewd in quietly locking up patents early. Barberich, though, says he climbed through a big cultural loophole that exists at nearly all big drug firms. "At any big research-based company, the emphasis is on finding new drugs, not going back to reexamine old products," he explains. "Most good research scientists would quit if they were told to tinker with a 15-year-old molecule." Drugmakers, too, were reluctant to unearth flaws in products, for fear they would be sued for not repairing or withdrawing a profitable drug fast enough.

Such easy pickings never last forever, of course. In 1992 the FDA began pushing all drugmakers to eliminate evil-twin molecules from new drugs. By now, most big companies know to study and patent their metabolites too. That means Sepracor probably can't patent any more single-isomer or metabolite versions of blockbusters, and the company could wither when its patents disappear in a decade or so.

How will Sepracor survive? It has made the brave decision to try to become a full-fledged drug company, complete with original research and a sales force. Already it has 65 salesmen out flogging an asthma drug that won FDA approval this year. It has 35 research scientists busily swirling their alembics and peeping into microscopes.

Whether Sepracor will become the next drug giant is anybody's guess. Far bigger firms have failed to come up with profitable new drugs and been forced to merge to survive. But Sepracor is in a better position than most small wannabes. It has money in the bank and reasonable prospects of a decent or better revenue stream for the next ten years, enough time to bring original research to market if all goes well. That would be like having World War II submarine movies and peach cobbler all rolled into one.