Scientists link gene behavior with disease and drugs

Connectivity Map opens doors to early prediction of effective or detrimental drug candidates.

Randall C Willis
BOSTON—As described in Science and Cancer Cell, researchers have developed a system that correlates gene expression with human dis­eases and small molecules. In the process, the Connectivity Map (cmap) has opened doors to early prediction of effective or detrimental drug candidates.
 
"The Connectivity Map works much like a Google search to discover connec­tions among drugs and dis­eases," said Dr. Todd Golub, director of the cancer pro­gram at the Broad Institute, in announcing the project. "These connections are notoriously difficult to find in part because drugs and diseases are characterized in completely different scien­tific languages."
 
The researchers monitored changes in gene expression of human cells treated with 164 small-molecule com­pounds—so-called perturba­gens—that reflected a range of activities and included FDA-approved drugs and non­drug bioactive molecules. They then compared these profiles with gene expression signatures from various human disease states.
 
What makes this effort dis­tinct from other efforts to tie gene expression with disease is how the researchers worked with the data.
 
"Rather that trying to correlate disease/drug-induced states with gene expression, the basic idea of cmap is to use gene expression data to connect disease states with the actions of small molecules," says Dr. Justin Lamb, a Broad research­er. "In other words, we use gene expression changes as a transitory feature, and often don't even look at the lists of genes that underlie the connections."
 
The theory ran that if a given disease increased expression in 10 genes and decreased activity of 15 genes, the best drug candi­date would be the compound that decreased and increased expres­sion in the same 10 and 15 genes, respectively.

While the researchers acknowl­edge their repertoire of cells, com­pounds or dosing regimens were not comprehensive, they were still able to identify compounds with high cmap scores for their abil­ity to either mimic or suppress a handful of diseases. For example, a compound that had a high score against Alzheimer's disease had been shown elsewhere to reverse fibril formation in vitro.
 
"There are circumstances where using the correct cellular context and small molecule concentration, for example, are absolutely criti­cal to finding strong connectivity," Lamb says. "But what is remark­able, I think, is that the approach does not seem unduly sensitive to these parameters. We find many connections largely independent of cell type, for example."
 
According to Lamb, the team is quite open to outside assistance in their efforts to expand cmap, whether it comes from other academic groups or government agencies. Likewise, he says they would be delighted by any interest from pharmaceutical or biotech­nology companies. As an incentive to all—and as a general belief—the group is making cmap publicly available through a web site host­ed by the Broad.
 
"Our number one priority is to expand our dataset to include profiles for the 1400 or so drugs approved for human use by the FDA, because of their proven safe­ty and tolerability," he adds. "By the same argument, we'd also dear­ly like to include late-stage clinical efficacy failures in our set."

Randall C Willis

Subscribe to Newsletter
Subscribe to our eNewsletters

Stay connected with all of the latest from Drug Discovery News.

February 2023 Front Cover

Latest Issue  

• Volume 19 • Issue 2 • February 2023

February 2023

February 2023 Issue