YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

Compute for the Cure

By Kevin Davies

Nov. 15, 2006 | More bleak news emanated from the pharmaceutical sector last month, as big and small pharmas, including AstraZeneca, GlaxoSmithKline, and Aspreva, reported failures of once-promising drug candidates in late-stage clinical trials. As more and more drug candidates wind their way through the tortuous approval process, only to stumble at the last hurdle, the moral is clear: the industry hast to find more systematic, intelligent, expeditious methods to identify new drugs (and targets) or repurposing existing molecules.

Thus the latest resource from the Broad Institute is particularly timely. The use of high-throughput gene expression methods as a disease diagnostic and cellular readout is well known. But what the Broad’s Todd Golub, Justin Lamb and their colleagues have done is create a new database that allows gene expression profiles of cells treated with various drugs to be queries with gene signatures representing different disease states or other compounds. Computational comparison offers a seductive new approach to identify new drugs for disease, as well as re-purposing existing drugs.

Dubbed by some commentators the “Google of drug discovery,” the Connectivity Map (C-Map) was outlined in a recent paper in Science*, and a pair of accompanying publications in Cancer Cell. In the same spirit as the human genome project, the C-Map database can be mined freely by researchers over the Internet  ( As Golub told NPR’s Science Friday, one of the hopes of the C-Map is it “will allow scientists to discover new, previously unrecognized functions for old drugs that we know are very safe.”

Freedom of Expression
For the proof-of-concept study in Science, Lamb and colleagues spent about a year gathering expression data on 164 drugs (or “perturbagens”) on a few cell lines, chiefly a well-characterized breast cancer cell line called MCF7. The drug-induced gene expression signatures were captured typically six hours after treatment, using Affymetrix GeneChip microarrays. For each tested drug, therefore, the C-Map contains a rank-ordered list of expression changes for some 22,000 genes.

Against this database, users can try to match a “query signature” — a list of genes that are reproducibly up- or down-regulated under various conditions, for example treatment by a particular drug, or tissue affected in a certain disease. (The strength of the match is calculated using a pattern-matching approach based on the Kolmogorov-Smirnov statistic.)

Although early days, Lamb, Golub, and colleagues offer several intriguing examples of the C-Map in action. In most instances, the query signature was taken from publications by other groups, demonstrating the potential breadth of applications and community involvement. For example, the C-Map:

•            Identified two putative HDAC inhibitors based on connections to known inhibitors, vorinostat and trichostatin A.

•           Noted a high correlation between a gene signature of diet-induced obesity and the effect of PPAR_ agonists.

•            Revealed a highly negative connection between a 25-gene signature of affected Alzheimer’s disease brain and DAPH, which is under active investigation as a potential AD therapeutic.

Two further applications of the C-Map are covered in detail in a pair of papers in Cancer Cell by Golub’s group. In one paper, the authors were able to uncover a mechanism of action for gedunin, a plant extract identified in a screen of androgen receptor inhibitors but about which little else was known. Gedunin turns out to interfere with an important cellular quality control mechanism mediated by a heat shock protein.

And in a second paper, a study led by Guo Wei et al. showed that the Wyeth immunosuppressant drug sirolimus (also known as rapamycin), which is most commonly used to prevent patients reject organ transplants, is a promising candidate for overcoming drug resistance in acute lymphoblastic leukemia. The sirolimus signature closely matched one associated with dexamethasone sensitivity. “The result from the Connectivity Map immediately suggests that sirolimus should be tested in a clinical trial of ALL patients with dexamethasons resistance,” the authors write.

Golub says an immediate goal is to develop gene signatures for all of the estimated 1,400 drugs approved by the FDA, a task that will likely take a year or two. In the meantime, the C-Map offers a wealth of connections into genes, drugs and diseases, validating known links and revealing new leads and mechanisms. It will surely boost efforts into identifying new therapeutics for disease as well as new uses for existing drugs.

 * Lamb, J. et al. Science 313, 1929-1935 (2006); Hieronymus, H. et al. Cancer Cell 10, 321-330 (2006); Wei, G. et al. Cancer Cell 10, 331-342 (2006).

 Email Kevin Davies.

 Subscribe to Bio-IT World  magazine.


View Next Related Story
Click here to login and leave a comment.  


Add Comment

Text Only 2000 character limit

Page 1 of 1

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.