Faecal library for water contamination
Posted 25 October 2016
With a so-called 'Library of Poo', machine learning and a credit-card sized device, US researchers have developed a highly accurate way of determining the source of water contamination.
Microbial ecologists from Lawrence Berkeley National Laboratory say the PhyloChip
technology could detect more than 60,000 species of bacteria and archaea, and distinguish between those that pose risks to human health and those that do not.
“With the PhyloChip, in an overnight test we can get a full picture of the microorganisms in any given sample,” said one of the lead researchers, Eric Dubinsky.
“Instead of targeting one organism, we're essentially getting a fingerprint of the microbial community of potential sources in that sample. So it gives us a more comprehensive picture of what's going on. It's a novel way of going about source tracking.”
Dubinsky said the technology is an improvement on current testing processes, which involve collecting water samples, culturing the bacteria overnight, and then checking the growth levels of E. coli and Enterococcus – presumed indicators of fecal contamination.
Existing methodology also fails to distinguish where the bacteria came, be it from humans, cows, sewage or decaying vegetation.
The PhyloChip identifies microbes based on variations of a specific gene, with no culturing needed.
But Berkeley had to build an extensive 'Library of Poo' to help pinpoint the source of particular microbial communities, according to the second lead researcher on the project, Gary Andersen.
“We had Laleh Coté, an intern at the time and now a lab employee, run around and basically collect poop from all sorts of animals,” he said.
"What we've done since then is develop a reference library of the microbial communities that occur in different types of poop – we have cows, horses, raccoons, humans, different types of birds, pigs, sea lions and other animals, as well as sewage and septage. We used that library to develop a model."
The new method takes the unknown sample and compares it against this microbial reference library.
"We've used the PhyloChip in a way that it hasn't been used before, by using machine learning models to analyse the data in order to detect and classify sources," Andersen said.
"It's essentially giving you a statistical probability that a microbial community came from a particular source."