Many microorganisms are "unculturable," or at least not able to grow in known media. Now, a new tool enables researchers to predict what nutrients organisms need to thrive in the lab, eliminating most of the guesswork involved in setting up new cultures. Read more...
The team combed through recipe files and extracted details such as ingredient lists and salt concentrations to create a Known Media Database (KOMODO) that includes more than 18,000 strain-media combinations as well as more than 3000 media variants and compound concentrations.
Oberhardt and his colleagues then leveraged the database to predict which organisms would grow well in which media. By looking at media that support multiple microbes and applying the transitive property and a phylogeny-based filter, they were able to predict which microbes would grow in new in vitro experiments with approximately 83% accuracy.
“No one has really looked at this problem this way before, and we’ve been able to set a new framework,” Oberhardt said. “We’re not done yet. We’re going to improve the database by including more biogenetic data and ecological data. But we think this is a really good starting point, and we can use more sophisticated machine learning methods to help us create good metabolic models in the future.”
Oberhardt MA, Zarecki R, Gronow S, Lang E, Klenk HP, Gophna U, Ruppin E. Harnessing the landscape of microbial culture media to predict new organism-media pairings. Nat Commun. 2015 Oct 13;6:8493. doi: 10.1038/ncomms9493.