# Comparing end-user diagnostic outputs from a commercial tNGS pipeline for Mycobacterium tuberculosis drug resistance detection

## TL;DR

This study shows that updating a bioinformatics pipeline improves the detection of drug-resistant tuberculosis without changing lab methods.

## Contribution

The study demonstrates that bioinformatic updates alone can significantly enhance diagnostic accuracy for drug-resistant TB.

## Key findings

- The updated pipeline showed substantial agreement with the original pipeline for drug resistance calls.
- Significant increases in sensitivity and diagnostic yield were observed for specific anti-TB drugs using the updated pipeline.
- Bioinformatic enhancements alone improved diagnostic performance without requiring wet-lab changes.

## Abstract

Targeted next-generation sequencing has emerged as a rapid solution for diagnosing drug-resistant TB (DR-TB) directly from clinical specimens. Updating the bioinformatics software component can lead to rapid improvements in diagnostic performance. We compared the diagnostic performance of an updated bioinformatic pipeline output to the original pipeline output for the Oxford Nanopore Technology (ONT) TB Drug Resistance Test.

A total of 721 sediment samples were evaluated for 13 anti-TB drugs using phenotypic drug susceptibility testing and whole genome sequencing. Sequencing data outputs previously analysed using the original pipeline were re-analysed using an updated pipeline and compared.

There were no significant differences in successful sequencing results, and direct comparison of DR-TB call agreement was substantial (κ > 0.7) between the original and updated pipeline outputs. Diagnostic accuracy relative to the composite reference standard was compared, and significant (P value < 0.05) increases in sensitivity and diagnostic yield, using the updated pipeline, were identified for streptomycin, pyrazinamide, bedaquiline, and clofazimine.

Comparison of the updated pipeline to the original pipeline revealed significant improvements in diagnostic performance, demonstrating that bioinformatic enhancements alone – without wet-lab modifications – can substantially boost sensitivity and diagnostic yield for DR-TB. These findings underscore the critical role of continuous pipeline optimisation in the evolving resistance landscape to enhance real-time clinical decision-making.