It demonstrates how whole genome sequencing can be used to accurately predict drug susceptibility for the four first-line anti-tubercular drugs (isoniazid, rifampicin, pyrazinamide and ethambutol) and, crucially, how the accuracy only degrades slowly as the local prevalence of resistance increases.
All the 10,000 samples had qualitative drug susceptibility profiles gathered using the MGIT system which simply reports whether a sample is resistant or sensitive to a drug. CRyPTIC is currently collecting >30,000 samples using a 96-well plate containing 14 different drugs and therefore will be reporting quantitative data (i.e. minimum inhibitory concentrations). So, whilst this is a good first step, I expect the project to produce a wide range of even more exciting studies in the next few years.
Since our ability to draw inferences from correlations between the genetic and phenotypic data depend critically on minimising the errors in the data, I have been heavily involved developing methods to allow us to quality control the measurements made in the different consortium laboratories. These are
- AMyGDA. This is software that automatically analyses photographs of the 96 well plates CRyPTIC is using. You can read more here. A paper is accepted and I’ll update this post when I have a link.
- BashTheBug. This is a Citizen Science project hosted on the Zooniverse where anyone can classify a few (or a thousand) images taken from the 96 well plates.
These independent sets of measurements are then combined with those collected in the laboratories and a consensus is reached, hopefully minimising errors, improving signal to noise and reducing therefore the number of samples we need before we can start e.g. identifying new genes and genetic variants that can confer resistance to anti-TB drugs.
In short, watch this space!