Andreas and Aris Persidis love to tackle big problems. As graduate students in the 1980s, the two brothers—a naval architect and a biochemist, respectively—would spend long nights knocking around ideas for how to engineer a better propeller or a sleeker keel, tangible challenges involving multiple factors and many data points.
A decade later, the brothers turned their sights toward the big problems of a new discipline, one where data were beginning to pile up. “We thought, ‘Why don’t we use engineering theories and techniques to attempt to address some of the unbelievably complicated questions in biomedicine?’” recalls Aris Persidis. “Someone had to do something like this, so we decided to give it a go.”
In 1996 the Persidis brothers cofounded Biovista, a biotech based in Charlottesville, Virginia. Over the next eleven years, using their engineering know-how, the duo developed a novel technology to answer a simple question: If we know how a drug works, can we look at data from laboratory studies and clinical trials to predict other diseases for which that drug might be therapeutic? “Drugs surprise us with new activities all the time,” says Persidis. “A drug that is well seasoned is an excellent starting point for innovation.”
Drug repositioning or repurposing—an effort to find new uses for existing drugs—is not a new idea. Viagra, for example, was first tested as a hypertension treatment before it became a blockbuster drug for erectile dysfunction. Arsenic, once used to treat syphilis, is now regularly used to combat leukemia. And thalidomide, a reviled antinausea medication for pregnant women that was pulled from international markets in the 1960s after it was found to cause terrible birth defects, got a second wind in 2006 with US Food and Drug Administration approval for its use in bone-marrow cancer.
“Nature is very parsimonious,” says J.K. Aronson, president emeritus of the British Pharmacological Society, a professional association for UK pharmacologists. “If you’ve got a system used in one place that can be used somewhere else, it is used somewhere else.”
By simply identifying a different disease that can be treated with an existing drug, companies can skip preclinical and early clinical trials, and thus leapfrog much of the estimated 10–15 years, and the more than $1 billion it takes to bring a new drug to market. Companies developing new drug candidates can also recoup major losses if a drug fails in Phase II or III trials by finding a new indication in which to move it forward. But in the past, drug repositioning has been an unpredictable process—an occasional happy accident when a doctor noticed a strange side effect or a researcher documented an off-label use for a drug.
—Aris Persidis, Biovista
Biovista already has two candidates for treating progressive multiple sclerosis, both of which proved successful in animal models of the disease, and the company has established partnerships with Pfizer and the FDA. It is one of a handful of data-savvy young biotechs working to transform drug repositioning from an occasional coincidence to a systematic pursuit of new markets.
“The value of drug repurposing is underappreciated,” says Pankaj Agarwal, director of computational biology and bioinformatics at GlaxoSmithKline. “If you can find a new use for something that’s been out there for 5, 20, or 50 years, that’s very powerful.”
NuMedii, a California-based biotech started in 2008, is one of the youngest companies to join the repositioning movement, but it has made some of the quickest strides toward proving the effort worthwhile. The company’s technology, developed by Stanford bioinformatician and pediatric endocrinologist Atul Butte, maps gene activity patterns from a database containing molecular profiles of over 300 diseases. If two diseases share a molecular profile—a similar set of activated genes—perhaps they could also share drugs. Drugs that work for heart attack patients, for instance, should perhaps be tested for effectiveness in people with muscular dystrophy, says Butte, as the two conditions share similar activated pathways.
Butte has been validating such predictions at Stanford, with unpublished but promising results in animal models, testing two existing medications that may be repurposed to treat Crohn’s disease and lung cancer, respectively. The patents for both drugs have long since expired, which turns out to be a double-edged sword. While the drugs are both cheap and available to repurpose as new treatments for two serious illnesses, it is challenging to find a pharmaceutical partner to fund the Phase II trials required to get the drugs approved for these new uses, says Butte. Though the FDA offers patent protection for repurposed drugs in their new indication, there’s no guarantee that a doctor won’t prescribe a generic version that came on the market after the original patent expired, even though it’s technically not approved for the new use.
“If there’s no one clear partner who’s going to benefit, the funding for the trial is questionable,” adds Butte. “It’s a catch-22.” Still, the experiments demonstrate the technology’s promise in predicting which diseases an existing drug might treat—a critical proof of concept that the technology could prove valuable when applied to patented drugs that aren’t selling well in their current indication, says Gini Deshpande, cofounder of NuMedii. “Our goal is to help our pharmaceutical partners realize the full potential of the drug that they’ve spent enormous amounts of money developing,” she says.
Biovista is also eager to partner with pharmaceutical companies to reposition marketed drugs, as well as those languishing on dusty shelves after failing at some step during preclinical or clinical development. The company’s technology collects publicly available data on diseases, drugs, targets, and adverse events, then organizes the data into 20 comparable categories—such as gene associations and comorbidities—and scans for similarities. It’s eHarmony for medicine, says Persidis with a laugh: “compatibility in many dimensions. That’s why we can navigate all 23,000 diseases and all 6,000 adverse events against all 20,000 human targets and 95,000 drugs and pharmacologically active compounds with reasonable data in the public domain.”
The technology’s predictions are then tested in vitro and in vivo, Persidis says—with an impressive 70 percent success rate. The platform’s prognostic value has caught the attention of Pfizer, who last November inked a deal with Biovista to identify novel indications for a number of existing Pfizer medications.
While Biovista and NuMedii take a bioinformatics-heavy approach, leveraging massive computing power to reposition drugs, Melior Discovery uses a more opportune modus operandi. Since 2005, the small, Pennsylvania-based biotech has relied on what CEO Andrew Reaume calls “systematic serendipity” to identify potential new uses for old drugs, a.k.a. the stumble-upon-them technique. Using a nonhypothesis-driven method, the company runs drugs through a series of 40 animal models representing a wide gamut of illnesses, from Alzheimer’s disease to asthma to overactive bladders, looking to see what works. “It’s a way to uncover potential therapeutic effects that otherwise would not have been predicted,” says Reaume.
As haphazard as the approach sounds, it has produced remarkable results: In 5 years, the company has run over 250 compounds through the 40 animal models and found potential therapeutic uses about 30 percent of the time, says Reaume. Their lead candidate, MLR-1023, a kinase activator discarded as a treatment for gastric ulcers after poor Phase II results, showed activity in a mouse model of Type II diabetes and will soon begin a Phase II trial for the disease. “Nobody in the field of diabetes, as well studied as it is, was predicting that [this] kinase would be an attractive target,” says Reaume. “Now we’ve shown it to be one.”
Though these biotechs are validating their platforms using existing drugs, their executives believe the true hotbed for the technology might be even earlier in drug development. During preclinical analysis, these approaches may be able to identify two to four indications for a compound that companies can pursue in parallel to see which is the most successful. “That’s where you’re going to start to see some gains in this approach,” says Deshpande. “You’ll see minimization of failed clinical trials and better drugs coming out for therapeutic indications.”
“It would make sense, if you’ve got a new drug, to try and see if that drug affects all the different targets in your pharmacology lab,” adds Aronson. “You might find there’s a drug that you’d synthesized for one purpose that has an action in another system entirely.”
Big Pharma is also making a play at developing systematic repositioning programs. In 2007 Pfizer established an Indications Discovery Unit in St. Louis, a group dedicated to repositioning the company’s failed compounds and finding new indications for the promising ones. GlaxoSmithKline has shown a similar interest, but is exploring bioinformatics methods to prove that drug repositioning is a profitable effort before committing significant resources. “The approaches are largely unproven. We still have a long way to go to show how much value it can generate,” says GSK’s Agarwal, who is leading the work. “But I think it’s promising.”
“There’s a good dose of healthy skepticism in the pharmaceutical industry about [bioinformatics] approaches,” adds Deshpande. “But if we see a couple of success stories come out of this approach, we’ll see this field open up really, really widely.”