This so-called "Connectivity Map" will be able to reveal links among drugs, genes and diseases, the researchers wrote in the latest issue of Science.
The findings, reported Thursday by US scientists and also published in the journal Cancer Cell, may accelerate the search for new drugs to treat diseases, by predicting the molecular actions of novel therapeutic compounds and revealing how existing drugs can be newly applied to treat diseases such as cancer.
A key challenge in biomedicine is to connect each human disease with drugs that effectively treat it, and to understand the molecular basis for such drugs' effects.
To solve this problem systematically, the researchers described the effects of drugs and diseases in the common language of "genomic signatures", meaning the full complement of genes that the drugs turn on and off.
The researchers collected gene-expression profiles from cultured human cells treated with 164 bioactive small molecules, including the anti-cancer drug gedunin, estrogen and certain antipsychotics.
They also collected genetic signatures from cells affected by conditions such as diet-induced obesity, Alzheimer's disease, and a drug-resistant form of leukemia.
Using pattern-matching software to mine these data, the researchers identified potential mechanisms of drug action, confirmed previous applications of known drugs, and discovered potential new uses for known drugs.
"This is a powerful discovery tool for the scientific community," said Justin Lamb, the lead author of the Science paper and a senior scientist at the Broad Institute's Cancer programme.
"By analysing just a small fraction of available drugs, we have already confirmed several biological connections between drugs and human disease, and made entirely new ones too."
One of the surprising results emerging from the connectivity map involves gedunin, a plant derivative that, despite a long history of medicinal use, is not well understood molecularly.
Another key finding suggests a new way to overcome drug resistance in cancer, according to the researchers.
Using the connectivity map, another team led by Scott Armstrong, an assistant professor at Harvard Medical School, identified the FDA-approved immunosuppressant drug sirolimus, as a therapeutic candidate for overcoming drug resistance in a form of human leukemia.