Detecting mixed Mycobacterium tuberculosis infections and differences in drug susceptibility with WGS data

Tuberculosis is a pulmonary disease caused by the pathogen Mycobacterium tuberculosis and is the second leading cause of death from an infectious disease. Individuals infected by multiple strains referred to as a mixed infection are associated with poor treatment outcomes when the infecting strains differ in their susceptibility against antibiotics. Studies aimed to detect mixed infections are likely to underestimate the true prevalence of mixed infections, because conventional genotypic methods have limited sensitivity to distinguish TB strains. Tools are needed to distinguish strains at a finer resolution and allow the simultaneous detection of multiple strains. Whole genome sequencing yields more information and therefore provides increased resolution to also distinguish closely related strains. In this study, I detect mixed infections by using a large number of group-specific SNP markers obtained from sequence data. I associated SNPs to clusters in the MTBC phylogenetic tree to obtain cluster-specific SNP markers that allow detecting and estimating frequencies of present strain types at different levels in the phylogeny. The prevalence of mixed infections was found to be ~10% of which half were mixed at the sub-lineage level. Approximately 15% of the mixed infections were found to have ambiguous SNPs corresponding to locations that are known to cause drug resistance, indicating the presence of MTB populations with differing drug susceptibility profiles. The results show that patient-level multi-drug resistance can be caused by multiple strains each with their own resistance to a particular drug. This work illustrates the dire need for high-resolution molecular diagnostics that can pinpoint the exact nature of the infection.BioinformaticsIntelligent SystemsElectrical Engineering, Mathematics and Computer Scienc