antimicrobial resistance citizen science clinical microbiology publication tuberculosis

New publication: Automated detection of bacterial growth on 96-well plates for high-throughput drug susceptibility testing of M. tuberculosis

In this Microbiology paper we show how a Python package, called the Automated Mycobacterial Detection Growth Algorithm (AMyDGA for short), can be used to independently read a 96-well plate designed for determining the minimum inhibitory concentration of 14 different anti-tubercular drugs. AMyGDA is reproducible and shows promising levels of accuracy. Where it fails, it does in known ways, for example when there is little bacterial growth, or there are artefacts in the image, such as air bubbles, shadows or condensation.

You can download the software. Included are 15 images for testing that allow you to reproduce some of the figures in the paper. AMyGDA was discussed in an earlier post and also underpins the BashTheBug citizen science project since it allows the image of each 96-well plate to be segmented. The BashTheBug volunteers recently completed a million classifications.

The international CRyPTIC tuberculosis consortium is already using AMyGDA to quality control the readings used by the laboratory scientists; discrepants are sent to BashTheBug for adjudication.

By Philip Fowler

Philip W Fowler is a computational biophysicist studying antimicrobial resistance working at the John Radcliffe Hospital in Oxford.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.