New paper: Evaluating 12 WGS analysis pipelines for MBTC

Ruan Spies did a careful systematic analysis of the publicly-available pipelines that claimed to process raw genetics files from M. tuberculosis complex samples, including the pipeline available via EIT’s GPAS which we have written.

The preprint was posted back in January 2025 and the paper has just been published in the Lancet Microbe.

There is a real range: 28 pipelines were identified but over half (16) were excluded for reasons ranging from being unable to install or execute (n=7), to no longer having a functioning website (n=2) or having a bug that was not fixed at the time of the study (n=1). This was a shame and illustrates the value in both properly funding academic software and using appropriate software engineering skills.

Of the 12 that could be evaluated, six required local compute to be run and the other six processed samples in the cloud. Predicting whether a sample was resistant or susceptible to a range of antibiotics is the academically difficult exercise yet most of the 12 achieved similar results. Only four were able to determine if a sample was epidemiologically related to other samples seen.

Obviously I am biased, but it was pleasing to see GPAS do well, partly on the non-functional requirements like availability, accessibility, scalability, privacy, security, and sustainability. For example, GPAS was the only pipeline based in the cloud that removed human reads from the FASTQ files before the file left the users network, an increasingly important consideration with the advent of GDPR.