Meeting Abstract
Recently, amphibian skin bacteria have been isolated and sequenced as part of an initiative to develop probiotic-based approaches to mitigate infection by the pathogenic fungus Batrachochytrium dendrobatidis (Bd) in wild amphibians. Bioinformatics is increasingly used to infer the inhibitory ability of bacteria based on short-read amplicon data (~250 bp of the 16S rRNA gene) by matching these sequences to a database of Bd-inhibitory isolates. However, anti-fungal predictions made solely via sequence-based similarity are rarely validated. Here, we matched sequences of the bacterial 16S rRNA gene of 22 isolates from the skin of two endangered Atelopus frog species to sequences in the database, and used the database to predict the anti-Bd ability of each isolate. We then performed in vitro challenge assays to determine whether the isolates behaved as predicted. We found that the database only predicted the outcome of the challenge assays for 55% of the isolates, suggesting a fair amount of uncertainty in using sequence matching to assign anti-Bd properties of amphibian skin bacteria without actual challenge assay data.
