Rapid drug resistance prediction in positive <i>Mycobacterium tuberculosis</i> clinical samples using an extensive targeted next-generation sequencing panel

Tuberculosis (TB) remains a global health challenge, exacerbated by the emergence of drug-resistant <i>Mycobacterium tuberculosis</i> strains. Most methods for drug susceptibility testing (DST) are culture-dependent and time consuming, possibly delaying optimal TB-treatment. This study aimed to develop an extensive targeted next-generation sequencing (tNGS) approach for rapid genotypic DST directly from clinical samples. We designed a tNGS panel comprising 30 amplicons targeting 19 genomic regions associated with resistance to 20 antibiotics. This method was applied to 71 smear-positive (0-3+) pulmonary TB clinical samples collected at the Portuguese National Reference Laboratory. DNA was extracted and amplified using multiplex PCRs, followed by sequencing on Oxford Nanopore Technologies MinION platform. Sequencing data were using TB-Profiler and the tNGS results compared to phenotypic DST and whole genome sequencing (WGS) data from corresponding isolates. The tNGS demonstrated high concordance with both phenotypic and WGS-based DST across different sample types and smear positivity levels. For first-line drugs, tNGS showed 88% categorical agreement (CA) with pDST, increasing to 97% when excluding undetermined results. Compared to WGS across all analysed antibiotics, tNGS achieved 92% CA, increasing to >99% when excluding undetermined results. Validation of the tNGS panel showed 90% (1,895/2,076) of amplicons reaching >10x coverage at all analysed positions and 43 (61%) samples with all complete amplicons above this threshold. Non-specific amplification of contaminant bacterial DNA was minimal, with most mapped off-target reads being of human origin. This method enables comprehensive resistance prediction directly from clinical samples and signifies an important development in TB diagnostics and resistance monitoring.