Biogen Speed-Reads a Haystack

By Mark D. Uehling

February 18, 2004 | If cells had groundskeepers, the type I TGFβ receptor kinase would be world-class, creating and dismantling fibrotic proteins in the lung, liver, and kidney. The protein is an enticing target in cancer research, as well as in diseases such as scleroderma, ocular scarring, and idiopathic pulmonary fibrosis. Inhibiting the protein could theoretically derail a variety of disease processes.

Last year, two companies working on blocking the TGFβ kinase published finding exactly the same small molecule. One group was led by chemist Scott Sawyer, senior research scientist at Eli Lilly, using conventional enzyme- and cell-based assays. The other effort was more computationally intensive, headed by Juswinder Singh, associate director of structural informatics at Biogen-Idec. Lilly's results were published in September 2003 in the Journal of Medicinal Chemistry; Biogen's paper appeared in December's issue of Bioorganic & Medicinal Chemistry Letters.

At the outset of the Biogen project, Singh insists, the TGFb kinase was terra incognita-a genuinely novel target. The absence of a detailed literature about the protein usually delays the search for a drug-like molecule to bind with it.

Biogen didn't want to use high-throughput screening (HTS), combining each of a library of real molecules against the TGFβ kinase. Singh has no philosophical objection to HTS, but says it is expensive, starting at $1 million per project, and seldom produces tight matches between a lead and a target: "There is no guarantee with HTS that you're going to find the right answer."

Digital Synthesis
So Biogen chose virtual screening to look for a potential TGFβ kinase inhibitor. Using its own insights and the Accelrys Catalyst suite, which runs on IBM, Silicon Graphics and Linux machines, Biogen constructed a hypothesis, built a list of 200,000 compounds to test it, and filtered the results.

The Accelrys software is part of a smorgasbord of databases and applications that Biogen dubs its integrated virtual chemistry platform. That includes a methodology for drug discovery that has been encapsulated into a proprietary application called SIFt Biogen's computer came up with 87 hits, including "a small molecular weight compound with phenomenal potency," Singh says. Much traditional chemistry followed.

Lilly, however, is skeptical about the above account. Sawyer, speaking via spokeswoman Asia Martin,  notes that in January 2003, Biogen scientists saw two posters on Lilly's TGFb receptor kinase project at an American Association of Cancer Research conference in San Diego. The posters depicted the active site of the TGFβ receptor kinase and the crystal structure of the small molecule that Lilly calls LY364947. (Biogen's nom d'pharm of its identical compound is HTS466284.)

For now, no one at Lilly is planning to commercialize LY364947. But Martin concedes there is a remote possibility of litigation in the future if either company should somehow leapfrog the other and submit a compound for regulatory approval. Egos on both sides sound bruised. "We were surprised that this [Biogen] paper did not reference our earlier work," says Sawyer.

"It is very unfortunate of them to say that," says Biogen's Singh, who confirms that Biogen did see Lilly's posters. But Singh insists Biogen's scientists came up with their molecule entirely on their own: "What we had was two groups, independently, discovering a very important inhibitor against a novel drug target and basically using two very different methods. One was a computational method. One was an experimental method."

The origins of the Biogen molecule may never be sorted out. But if the company's researchers started with a blank slate, their in silico approach appears to be a powerful shortcut to an eternal chemical quest-and an endorsement of using software to expedite years of painstaking work at the bench.