A Database for Disease
A genetic “roadmap” will help to find treatments for diseases, by looking at the signatures that drugs leave behind.
A newly developed genetic “roadmap” promises to streamline the drug discovery process. Called the Connectivity Map, this public database matches drug compounds with diseased cells and the processes occurring within them.
“The reason it’s so difficult to find those disease and drug connections is that the languages in which they are conventionally described are very different,” says Justin Lamb, senior scientist at the Broad Institute in Cambridge, MA. “A physician would describe a disease in terms of its physical symptoms, whereas a chemist would describe drug actions in terms of binding that chemical to a particular protein.” The researchers want to bridge that gap using a common language: gene-expression signatures.
At any point in time, some genes in a cell are expressed, or “on”, while others are not. And a cell’s particular profile of activity is known as its gene-expression signature. When cells are exposed to a drug, that signature changes: some genes that were expressed are turned off and vice-versa. And different drugs leave different signatures. It is these signatures that the researchers used to build the Connectivity Map.
Lamb and his colleagues conducted a pilot study on a select number of compounds and cell types to create the first installment of the map, reported recently in the journal Science. They chose 164 molecules that were biologically active, including drugs approved by the FDA and compounds commonly used as tools in the lab. They tested the molecules on four types of cancer cells–breast cancer, prostate cancer, leukemia, and melanoma–looking at how the compounds affected gene expression in those cells.
The researchers did the analysis using DNA microarrays made by the company Affymetrix. These tiny glass chips are coated with thousands of short sequences of DNA that refer to parts of the human genome that often differ between individuals. For a given drug or cell type, the chips produce a unique pattern corresponding to the particular genes expressed. For example, the hormone estrogen might cause breast cancer cells to express certain genes, but have no effect in a prostate cancer cell, and that difference would be visible on the DNA chip.
The researchers then developed a computer program to compare the signatures to each other and rate the strength of the connections. The data from even this relatively small number of cell types and compounds, Lamb says, has yielded two new findings, described in papers in the journal Cancer Cell.
One compound, a plant-derivative called gedunin, was identified through a conventional screening method as interfering with the hormone androgen in prostate cancer cells, which is an important strategy in treating the disease. But the exact mechanism of how gedunin blocked the androgen signaling pathway wasn’t clear. When the scientists searched the Connectivity Map for compounds that had similar activity to gedunin, though, they found matches to compounds that inhibit heat shock proteins and thus suppress androgen receptor activity.
The other finding involved a specific type of leukemia that was resistant to traditional chemotherapy. A team led by Scott Armstrong, an assistant professor at Harvard Medical School and Children’s Hospital in Boston, determined the signature of the drug-resistant cells, queried the Connectivity Map, and found a match to sirolimus, a drug currently used to prevent rejection after organ transplantation. When they tested the drug in the lab, the scientists found that it re-sensitized the leukemia cells to chemotherapy, reversing the drug resistance.
“That was a particularly gratifying example for all of us,” Lamb says, “because sirolimus is already FDA-approved for another indication. That means that this compound is known to be safe and tolerated by humans, and the path to clinical evaluation of sirolimus can probably be tested in the clinic much more quickly.”
The team plans to expand the map to cover all 1,400 or so drugs approved by the FDA, an effort that should take between one and two years. “We wanted to make data that was broadly useful, so that requires a systematic approach to data generation,” Lamb says. “And then if you can make that database accessible to the world in a way which is easy for the world to interact with it, that would solve a lot of problems for a lot of people.”
The work done by the Broad scientists “is right on target,” says Gregory Riggins, associate professor of neurosurgery, oncology, and genetic medicine at Johns Hopkins University. “This kind of group effort and approach is therapeutic-directed and is what is needed from the research community.”
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today