A team of researchers has developed a new way of representing chemical genomic data that could prove to be an effective tool for drug discovery.
In this week’s issue of Science, researchers from the Broad Institute of MIT and Harvard published a proof of concept for a so-called Connectivity Map — a database of 564 gene-expression profiles derived from screening 164 small molecules against several different cell lines.
The data set, while limited, proved informative. In a series of examples outlined in the Science paper, the researchers showed that these genomic signatures could be used to recognize drugs with common mechanisms of action, discover unknown mechanisms of action, and identify potential new uses for existing drugs.
For example, users armed with a gene-expression signature from any gene-expression platform can query the Connectivity Map, whose expression profiles are rank-ordered according to their differential expression.
The Connectivity Map uses a pattern-matching algorithm to compare the query signature with each of the reference signatures, and scores each profile based on where query genes tend to appear relative to the reference profile.
If up-regulated genes appear near the top and down-regulated genes appear near the bottom of a particular reference signature, the query signature has positive connectivity with that profile. If it’s vice versa, then the query signature exhibits negative connectivity. All instances in the database are ranked based on their connectivity score for each query.
In one example, the plant derivative gedunin was found in a separate screening study to disrupt androgen receptor activity, although it was not known how. By querying the gene-expression signature for gedunin against the Connectivity Map, the researchers found that it had high connectivity with several heat-shock protein 90 inhibitors — a surprising finding, given that gedunin is not structurally similar to these compounds. The researchers were then able to experimentally prove that gedunin nearly eliminated AR by inhibiting HSP90.
In another example, the researchers used the Connectivity Map to identify Wyeth’s immunosuppressant drug Rapamune, known generically as sirolimus and rapamycin, as a potential therapeutic for dexamethasone-resistant acute lymphoblastic leukemia patients based on its similarity to a gene-expression profile for dexamethasone sensitivity.
The authors note that the “encouraging” results of the pilot study support “the generation of an expanded Connectivity Map as a community resource project in the spirit of other genomic efforts.”
To be sure, there are a number of uncertainties associated with ramping up this kind map of to genomic scale, including the number of cell types and small molecules that ought to be studied; the optimal number of concentrations, time points, and replicates that would be required; and whether current analytical tools are sufficient to ensure the statistical accuracy of the results. But the authors propose an initial goal of “all FDA-approved drugs and inhibitory RNAs targeting a large collection of genes in perhaps 10 diverse cell lines.”
Justin Lamb, lead author on the Science paper and a senior scientist at the Broad, told BioInform that while the researchers have not yet secured funding for scaling up the effort, “it’s absolutely our aspiration to do that, and to explore as much of [the] biological space and as much of [the] chemical space as is possible and as is practical.”
Lamb said that the Broad team has already “built the infrastructure that would make this a scalable exercise” by adopting the Affymetrix High Throughput Array system, which uses 96-well plates to simultaneously process 96 GeneChip arrays.
Mapping the 1,300 or so FDA-approved drugs would be a good next step, Lamb said, because those compounds all have known biological activity and are known to be safe and tolerated for human use. “The beauty of that, of course, is that the path to clinical evaluation of those compounds for new indications is relatively short,” he said.
Lamb said that the Connectivity Map currently includes hyperlinks to the ChemBank database hosted by the Broad Chemical Biology program, and that there are plans in the works to improve the integration between the two resources.
The Connectivity Map probably won’t adopt too many features of other chemical databases, however. “It was a serious design decision that we made right at the start to not to try and replicate other small molecule annotation databases because there are very many of those already,” Lamb said.
The Connectivity Map is freely available at http://www.broad.mit.edu/cmap/.