New Web tool may speed drug discovery

Local scientists have created what they hope will become the Google of drug discovery: a free, Web-based search engine that quickly finds potential new compounds to treat particular diseases.

As the website went live yesterday, the team released three papers in top scientific journals, demonstrating the technique's promise. In initial testing, the search engine identified a potential leukemia drug. A clinical trial for that drug, which is already approved for another use, is likely to begin in the next few months -- a testament to the speed of the new approach.

The new tool, dubbed the ``connectivity map," works by quickly matching drugs and diseases that have opposite effects on some of the genes inside a human cell -- the set of instructions that, when turned on, tell the cell what to do. A match would indicate that giving the drug might reverse the effects of the disease, turning off genes the disease turns on, and vice versa.

The utility of the database needs to be verified with a larger study, but scientists said it promises to accelerate drug discovery at a time when developing a new drug is frustratingly slow and expensive.

``The industry is always going to be enthusiastic about any tool that helps in the discovery process, and this looks like a very exciting tool that has real potential," said Janice Reichert , a senior research fellow at the Tufts Center for the Study of Drug Development .

Scientists at the Broad Institute of Harvard and MIT, which created the web site and search engine, said they have already begun work to expand the database to include more than 10 times as many compounds. And, they said, they were hopeful that other laboratories would eventually join in, creating a massive, cooperative venture similar to the Human Genome Project, allowing scientists from around the world to efficiently aggregate their genetic data and insights.

``It is a major intellectual and practical tool," said Eric S. Lander, director of the Broad Institute. ``I imagine a world, five years from now, where everyone who is working on a potential drug will, as the first thing they do, quickly look it up" to see what diseases it might work on.

The work, published by the journals Science and Cancer Cell, suggests that it is possible to dramatically simplify some aspects of biology, and thus speed the search for cures, according to Dr. Todd R. Golub , who led the research at the Broad Institute and is also a cancer researcher at the Dana-Farber Cancer Institute and Children's Hospital Boston. Different types of cells typically respond slightly differently to drugs, but the team found that analyzing just a few types of cells, such as skin cells, was enough to make connections to diseases that affect organs as diverse as the brain and the prostate.

To assemble the database, the scientists used state-of-the-art technology that rapidly reads the activity of all the approximately 22,000 genes in a human cell. The team created distinctive genetic signatures for each drug and disease. Putting a drug on a cell, for example, might dramatically alter the activity of a few dozen genes or even just a few, leaving the rest relatively unchanged. The signature is the list of genes whose activity was boosted or turned down by the drug. The same was done with diseases; the gene activity in a diseased cell usually looks very similar to that of a healthy cell, except for a few genes.

All these signatures quickly add up to an overwhelming amount of data -- the boon and bane of the genetic age of biomedical research. What gives the software its value is its Google-like ability to sift through this vast array of information for a few relevant facts.

Lander said he had been unsure whether the new approach would generate meaningful connections, giving it ``about a one in three" chance of success.

The most striking demonstration of the technology was the identification of a leukemia drug, an effort led by Dr. Scott A. Armstrong of Children's Hospital and the Dana-Farber. Patients with the disease are treated with steroids, which helps kill the cancerous blood cells. But the cells of some patients resist this treatment. Armstrong found that some of the genes of resistant blood cells were boosted compared with genes in the normal blood cells. The database matched this signature to a drug used to prevent organ rejection in transplant patients. Called rapamycin , the drug appears to make the cancerous blood cells vulnerable to steroids, according to one of the papers.

Another finding suggested a new strategy for attacking prostate cancer, potentially leading to new drugs.

The pressing question now, said the researchers and other scientists, is precisely how powerful the search engine can be, and whether this will justify the work of expanding the database. In the database unveiled yesterday, the team included 164 compounds, but for the next round the team wants to gather genetic signatures for virtually all drugs approved by the US Food and Drug Administration, some 2,000 compounds, as well as trying more variations in the doses and types of cells used to generate the signatures.

``This is a tour de force, but it is a long way to the [next] stage," said Dr. Christopher P. Austin , director of the National Institutes of Health Chemical Genomics Center.

If the additional work yields promising results, said Austin and others, it is likely that many others will join the project.

The Paul G. Allen Family Foundation , started by one of the founders of Microsoft, has provided $750,000 to start the follow-up project, Golub said.

Gareth Cook can be reached at cook@globe.com