Bayer: The Drug Development Clock is Ticking

In a computer lab behind locked doors in Cambridge, Mass., Anthony Caruso oversees a staff of 15 scientists who are a rare mix of information technology and biology experts. These workers, called bioinformaticians, spend their days hunched at computers writing software algorithms and searching databases for the mysteries locked in human cells. The clock is ticking for Caruso’s project at the Lion Bioinformatics Research Institute (LBRI), which is scouring reams of genetic data to fulfill a five-year $100 million deal to identify 500 potential drug targets for German drug giant Bayer. So far, they’ve nailed 400.

PDF Download

The practice of bioinformatics comes at a critical time for the drug industry.b The average new drug now takes 15 years and $880 million to bring to market, according to a new report from the Boston Consulting Group. And even if the reality isn’t quite that bad, another challenge looms: Within three years, an expected 200 prescription drug patents will expire, leading generic drug makers to flood the market with cheaper replacements that threaten drug makers’ profits. In this environment, anything that makes the drug-development process more efficient is going to get a hearing. If the approach works in drugs, it’s likely other industries whose growth hinges on new-product development will seek to apply bioinformatics’ concepts, especially its emphasis on fast evaluation of complex information.

Bayer, which sells about $4 billion in drugs each year, builds deals with companies such as LBRI to supplement the company’s own drug R&D spending, which was $1.3 billion in 2001. Overall, about 14% of drug companies’ R&D dollars nowadays are spent on outside deals, according to Accenture. Bayer itself has committed some $2 billion to these outside partnerships in recent years.

LBRI customized its technology platform specifically for Bayer, whose internal biologists can log on to get research updates from labs in Japan, England, Germany and the U.S. While LBRI’s data is protected from outsiders by a firewall, the company installed two dedicated lines to Bayer’s West Haven, Conn., research site and a 64K line to its German headquarters. LBRI’s scientists work on computers running both Windows NT and Linux operating systems and write code in the Java and PERL programming languages.

They search for targets largely by using Lion’s workhorse—a proprietary database system called SRS, which provides access to 480 public databases including gene sequence databases, protein sequence databases and databases that track how genes react in the body to other stimuli. While scientists during the 1990s were challenged by the lack of genetic information, they have had the opposite problem since mid-2000, when the Human Genome Project arrived, promising information on 30,000 genes. Not surprisingly, though, the flood of new Genomic information is “99% junk,” as one bioinformatician puts it, at least from the perspective of what researchers can use. Which is where the software interfaces come in, tying together as they do numerous databases, along with database tools and data-query technologies.

It’s not easy to identify a potential drug target. In the Cambridge lab, biologists use computers to analyze how the protein in one possibly carcinogenic gene interacts with proteins in other genes. The aim is to discover any abnormality that can lead to disease. Once the abnormality is identified, scientists then can rapidly test chemical compounds that might treat it.

Bayer scientists use LBRI’s results to narrow down the drug targets they pursue. The Boston Consulting Group believes technology such as LBRI’s can cut two years or more off the “discovery” stage—the first and most expensive stage of drug development (see chart, facing page). “It used to take several years to isolate the relationship between one gene and a particular disease,” says IDC analyst Mike Swenson. “Now it’s done in days.”

Without LBRI and its other partners, Bayer would be “under-resourced for the task at hand,” says Ken Kupfer, manager of Bayer’s scientific information data management group in Berkeley, Calif. “The whole justification for this project is that instead of working on the same targets we’ve worked on for 30 to 40 years, we’ll do more.”

Kupfer remembers the days when the programmers and analysts worked separately from the biologists and chemists. Now, the two sides are converging. Yet there are only about 1,000 bioinformaticians in the world—often Ph.D.s who command salaries of $100,000 a year and up. Bayer typically hires one or two such high-level experts for its research centers.

The research that scientists are doing for Bayer isn’t only intended to increase the speed of drug development. It’s also about increasing the number of drug possibilities. Bayer aims to produce 20 new drug trial candidates this year through its research—and out of that, deliver two new drugs to help the company recover from last year’s damaging recall of cholesterol drug Baycol/Lipobay. The drug was pulled off the market after being linked to 40 deaths.

Besides Lion, Bayer also has alliances with Millennium Pharmaceuticals and CuraGen Corp. Bayer and CuraGen are sharing $1.3 billion in development costs over 15 years in an effort to develop treatments for diabetes and obesity. Essentially, what Bayer is hoping is that CuraGen can help it identify unpromising development efforts early—and kill them.

The CuraGen deal also differs from LBRI’s in that CuraGen goes further to validate targets for Bayer. A scientist can discover a gene, determine its sequence, identify its genetic characteristics and proteins and then associate the gene with a disease or a patient population using CuraGen’s platform. “The problem with drug development today is there’s so much information out there and you can’t pull it together if you can’t automate it,” says CuraGen CIO John Murphy.

To that end, CuraGen’s platform not only lets scientists do database research, but also conduct experiments on computers in real time on a secure portal CuraGen operates.

After CuraGen’s job is done, Bayer takes over, using chemistry to develop compounds that will bind to the defined targets to treat the disease. The goal is to bring 12 drugs to clinical trial in 15 years as a result of the partnership.

While Bayer won’t be in a position to judge the success of the work its partners are doing until the end of the decade—a function of how long development cycles will last and when Bayer entered these deals—it has clear goals in mind. If LBRI, CuraGen and Bayer’s other partners can’t help Bayer hit its targets of bringing 20 drug candidates to clinical trial and two new drugs to market every year, Bayer’s multibillion dollar bioinformatics initiative will have fallen short.