DRUG DEVELOPMENT · Three companies reveal innovative uses of technology and data as key to their success.

By Malorye A. Branca

November 19, 2004 | It's what every biotech startup promises but few deliver — a turbo-charged discovery engine. Every now and then, however, a new company with fresh ideas starts hitting home runs.

In many respects, the role model is Genentech. With the launch of cancer drug Avastin early this year, the company's stock soared dramatically. As Alyce Lomax wrote for The Motley Fool, "Paying 72 times forward earnings for a stock that has tripled since this time last year seems a bit steep."

But it wasn't just excitement about a single new drug that was exciting investors. Genentech seems to have the Midas touch with a class of drugs called monoclonal antibodies. Avastin is just one of a stream of breakthrough monoclonals the company has developed, including Herceptin and Rituxan. The company doesn't do just monoclonals, however. Its next launch will be a small molecule, Tarceva.

The secret, says Andrew Chan, vice president of research-immunology at Genentech, involves three things: "We have to be innovative, use the most cutting-edge science, and be data-driven."

Genentech's scientists concentrate on biological areas where they have a deep understanding. "We have long stressed biology and pathways," Chan says. "We don't rely on huge high-throughput screens."

MODEL CYTOKINE: ZymoGenetics' approach is to zero in on certain proven proteins.

Rather than go it alone, Genentech seeks hot opportunities in other labs. Chan also credits a mixture of focus and creative freedom in the company's own discovery departments, allowing scientists to spend time on their pet projects. The hope is that some of these researchers will eventually "develop a whole new axis in biology," Chan says. That's pretty much what Avastin's discoverer, Napoleone Ferrara, helped do for the field of angiogenesis inhibition.


Buff Bioinformatics 
Like Genentech, ZymoGenetics is also gaining traction from the thoughtful use of technology. Spun out in 2000 from diabetes giant Novo Nordisk (a major partner), the company uses bioinformatics to identify important new protein drugs for conditions such as diabetes.

"People asked, 'Are you seriously telling us you can beat companies like Amgen and GSK to the patent office?'" says ZymoGenetics CEO Bruce Carter. Sometimes, he concedes, it seemed like a long shot, particularly during the peak of the genomics boom when bioinformatics and genome mining were all the rage.

ZymoGenetics distinguished itself by searching only for proteins in families already proven to be useful as drugs, such as certain cytokines, including erythropoietin, interferon, and GMCSF. The strategy has yielded two compounds in clinical trials and more than half a dozen intriguing pre-clinical candidates.

Carter credits "good bioinformatics" and a unique strategy for much of that success. "We use tools similar to what other companies have, but we have our own special way of using them," says Patrick O'Hara, vice president of bioinformatics. While other companies were using techniques that reaped highly expressed proteins, ZymoGenetics sought snippets of DNA related to specific structures. "Evolution expands on themes," O'Hara explains.

The work is not all high-throughput. "Bioinformatics only gets us so far," Carter says. "We still have to painstakingly work out the biological functions."

Lexicon Genetics also had doubters early on. This company's discovery engine is fueled by gene knockout mice, which are notoriously time-consuming to create. "People laughed and said, 'There is no way you can scale it,'" says Brian Zambrowicz, executive vice president of research. Today, Lexicon can house 300,000 mice in its facility and is systematically knocking out genes at a rate of about 1,000 per year.

Lexicon's Genome5000 program selects the 5,000 genes that encode the most pharmaceutically relevant proteins, such as G-protein-coupled receptors (GPCRs), kinases, and ion channels. "We're after the clean targets, with the maximum probability of efficacy and the minimum probability of mechanism-based side effects," Zambrowicz says.

More than a third of the way through that effort, Lexicon already has some 40 targets under investigation, including candidates for obesity and diabetes, atherosclerosis, cancer, inflammation, and cognitive disorders. Skeptics charge that knockout mice are not always good models of human disease. But Zambrowicz counters that publication of the mouse genome vindicates Lexicon's approach. The mouse has counterparts to 99 percent of human genes, and "in 85 percent of cases, the knockout phenotype correlates," he says. "No other method comes close."

The bottom line, Zambrowicz says, is that "genetics works. Done right, it will find you the genes that modulate every physiological trait of interest."

Both ZymoGenetics and Lexicon's lab prowess is impressive, but a surplus of drug candidates doesn't guarantee success. "Once we reach the clinic, our hurdles are the same as anyone else's," Zambrowicz admits.