Last week, Scientific American reported on the decline of nature-derived compounds in the pharmaceutical pipeline, in part because of the cost of isolating and identifying compounds that may have already been described and are therefore not patentable.

“You may invest a year of work screening these new compounds,” says computational biologist Pavel Pevzner of the University of California, San Diego. “Then a year later you figure out this work is wasted because somebody on the other side of the planet discovered it 10 years ago.”

This week in the journal Nature Methods, Pevzner and his colleagues provide a partial solution to this problem, letting computers sequence potential drugs analogous to the way they have untangled DNA sequences and proteins. “This area...remains the last fortress where computational techniques do not exist,” he says. (Scientific American is part of the Nature Publishing Group.)

Specifically, the new algorithm reconstructs the sequence of ring-shaped nonribosomal peptides (NRPs)—natural compounds that account for most antibiotics, including penicillin and anticancer drugs. Nine out of the top 20 bestselling drugs were either inspired by or derived from NRPs.

These molecules consist of several hundred different types of amino acids and are typically identified using nuclear magnetic resonance imaging, a tricky, expensive and time-consuming procedure.

With the new method, scientists run these compounds through a mass spectrometer, which breaks their rings apart into little chunks and measures their sizes. Then, a computer program uses this signature to deduce a molecule’s amino acid sequence, helping researchers weed out compounds that are similar to those previously described and sequenced.

Pevzner’s team uses the method to reconstruct a compound called “cyanopeptide X,” which they had previously sequenced using nuclear magnetic resonance imaging in 2007.

John Vederas, a chemist at the University of Alberta in Edmonton who was not involved in the study, is optimistic about the new method and praises the researchers for addressing a challenging dilemma. He notes that the algorithm is currently limited to NRPs that have previously been described and contain only a single ring. “As the database builds,” he says, “and there are more known structures, this will become more and more powerful.”

Watch last week’s slide show of nature-derived drugs