P027

1:University of Oxford

The rise of antimicrobial resistance (AMR) necessitates accurate and rapid diagnostics. Especially in Mycobacterium tuberculosis infections, whole genome sequencing (WGS) has become an established diagnostic method. But disagreement between the resistance prediction through genotypic, such as WGS, and phenotypic approaches can hinder effective treatment and surveillance. Within-host bacterial diversity, particularly the presence of subpopulations harbouring resistance-conferring alleles, may contribute to these inconsistencies. 

This study investigated the impact of incorporating minor mutations in WGS-based rifampicin resistance prediction in a dataset of 38,787 M. tuberculosis samples. We examined the presence of resistance mutations in homogenous samples as well as in subpopulations (heteroresistance). The sensitivity and specificity of resistance classification were evaluated with and without the inclusion of minor mutations for prediction. 

Identifying resistant minor mutations in the samples led to a significant improvement in sensitivity, with a small trade-off for specificity. Further analysis revealed distinct clusters of samples based on the number of minor variants present, pointing towards different mechanisms of resistance acquisition, i.e. resistance emerging due to within-host evolution or resistance acquired through a secondary infection.

Through our analysis, we have demonstrated the importance of considering within-host diversity when evaluating drug resistance in M. tuberculosis. Acknowledging and utilizing the predictive power of resistant minor mutations can significantly improve the accuracy of genotypic resistance classification. This enhanced accuracy can facilitate better-informed treatment decisions and more effective surveillance strategies. These findings raise the question of whether minor alleles should be routinely considered in resistance prediction algorithms.