Peptidic natural products, or PNPs, are groupings of amino acids that can make for effective antibiotics like penicillin and vancomycin. These chemical compounds are produced by microbes to kill off competing organisms and can exist everywhere, including in soil and in the human body.
VarQuest, an algorithm created by researchers at Carnegie Mellon University, the University of California San Diego and Saint Petersburg State University in Russia, has given scientists a way to quickly identify previously undiscovered PNPs.
Hosein Mohimani, assistant professor of computational biology in CMU's school of computer science, said that since the time of Alexander Fleming—who made his historic discovery of penicillin almost by accident— scientists have been looking for new ones.
"We are taking advantage of this warfare in the microbial world to extract these natural products and use them as antibiotics to kill human pathogens," said Mohimani.
He said that over the last decade, scientists have mostly been using an instrument called a mass spectrometer to try and find the PNPs within promising samples of material. The basic process is that a scientist will feed a sample into this instrument, which then sorts all of the present chemical compounds by weighing them very precisely.
For each compound, the mass spectrometer will spit out what Mohimani describes as a kind of "fingerprint" that reflects its structure. However, that fingerprint doesn't say exactly what that compound is, so when scientists catalog these results into databases, they face the challenge of picking out the PNPs from everything else.
"The field of natural product discovery is turning into a big data field where there are large-scale data sets collected from microbial samples," said Mohimani. "[We are] searching against billions of data points created by mass spectrometers."
Mohimani said that previously, it would have taken a computer hundreds of years to search through all of the world's existing data points using a "brute force" strategy: checking each data point, one by one, to see if it's a PNP or not.
VarQuest was able to comb through this same amount of information in just a few hours. Mohimani said that the algorithm arranges the information in these databases in a way that makes it easy to find "fingerprints" that share certain patterns which scientists have already associated with PNPs.
"This is analogous to the way that Google's search engine searches the whole web for a specific pattern given by the user," said Mohimani. "We were able to find thousands of novel variants of known antibiotics."
He said these new discoveries could give scientists a leg up in fighting bacteria like MRSA that are rapidly mutating and growing resistant to existing antibiotics.
