One third of the drugs used in the clinic today were synthesized not by chemists or biotechnologists but by plants or microorganisms. Most of these "natural products" come from a few genera of soil and marine bacteria that have long been known for their prolific chemistry. In addition to antibiotics, that group includes compactin, the grandfather of the entire drug class of statins, which Japanese microbiologists cultured from mold in rice sampled from a grain shop in Kyoto; sirolimus, an immunosuppressant used to prevent organ rejection following transplants, which Brazilian scientists uncovered in soil samples from Easter Island; and doxorubicin, a widely used anticancer drug, which Italian researchers isolated from the soil that surrounds Castel del Monte, a 13th-century castle.

The classical process of discovering drugs from microorganisms suffers from two challenges: it is slow—microbial isolates are painstakingly cultivated and tested one by one—and it yields limited possibilities for new drugs. We know about the limits from pioneering efforts to sequence the genomes of a few well-known drug-producing bacteria. These efforts reveal a gap between the small number of natural products each strain makes when cultured in the laboratory and the large number of drug-producing genes in each genome (enough to make three dozen or more molecules). The reasons for this deficit are not well understood but probably have something to do with the artificial conditions of lab culture, in which each species is typically cultivated in isolation.

Recent advances have raised the prospect of a resurgence in natural-product discovery, albeit from an unanticipated source. Our lab has developed a new algorithm to scan bacterial genome sequences for drug-producing genes. This tool accelerates the process of natural-product discovery by making it possible to perform the hardest step on the computer: prioritizing bacterial species and even specific gene clusters that are most likely to encode a novel drug. Using this approach, we found, to our surprise, a large number of drug-producing genes in the human microbiota, a group of organisms that has not previously been mined for biologically active small molecules. Other labs have developed complementary techniques based on recent advances in synthetic biology. A set of drug-producing genes can now be “refactored” in a way that greatly increases the likelihood that the genes can be expressed in an alternative, lab-friendly host.

These advances suggest two tantalizing prospects for developing drugs from the microbiome. The first would be to systematically mine the microbiome for new medicines, building on our unanticipated finding that gut, skin, oral and urogenital bacteria may be prolific producers of natural products. The second possibility is to use the microbes themselves as drugs—most likely in the form of an artificial bacterial community. Recent clinical efforts have shown that fecal transplantation is remarkably safe and engraftment of the new community surprisingly common, suggesting that a single consortium might engraft and function in a large subset of the human population. If a well-studied community of bacteria were used instead of a fecal sample, however, the cocktail of small molecules it produces could be optimized for a specific condition. Research suggests that such an approach might hold particular promise for the treatment of inflammatory bowel disease and perhaps metabolic disorders.

Small molecules that mediate many of the biological effects of the microbiota on the host could ultimately prove to be one of the richest sources for new drugs.