Replicate-anchored calibration of within-host single nucleotide variant detection in Mycobacterium tuberculosis whole genome sequencing
BackgroundWithin host diversity (WHD) in Mycobacterium tuberculosis is increasingly used to infer transmission and treatment response, but it lies at low allele frequencies where sequencing error and reference mapping artifact can be mistaken for signal. MethodsIn a cohort in Accra, Ghana, we measured WHD using deep short read whole genome sequencing and LoFreq with a depth-conditional detection limit and an artifact-aware mask and assessed reproducibility using within-patient biological replicates. FindingsAmong 83 single strain infections, about a fifth of called variants were recurrent artifact, and the corrected WHD burden was modest (median 16 variants per patient). WHD was higher in M. africanum than in lineage 4 (median 26 versus 14; p = 0.002) and was not explained by host or clinical features. Over the first two months of treatment the within host bacterial population drifted and cleared without detectable positive selection. Patients with an unfavorable two-month outcome carried higher baseline WHD (median 20.5 versus 14.0; p = 0.014). Shared within host variants did not mark transmission, and a single sputum undercaptured diversity. InterpretationWHD in early treatment is modest, lineage-patterned, and drift-dominated, and reading it for clinical or epidemiological signal depends first on measuring it defensibly. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSWe searched PubMed, Semantic Scholar, and Scopus (via Consensus) through July 2026 for within-host Mycobacterium tuberculosis diversity, using terms combining "within-host / minority / intra-host variants", "sequencing error / artifact", "transmission", "treatment outcome", and "M. africanum / West Africa". Within-host M. tuberculosis diversity is real but small and inconsistently measured, and low frequency calls are strongly shaped by sequencing and mapping artifact[1-5]. Shared minority variants have been proposed to sharpen transmission inference[1], while baseline diversity has not predicted treatment outcome[6]. Artifact-aware within-host data that include M. africanum, which causes a large share of West African tuberculosis, are lacking[7,8]. Added value of this studyUsing a depth-conditional empirical limit of detection with a monomorphic site false positive check, we provide an artifact-aware within-host diversity measurement in a Ghanaian cohort: a modest burden (median 16 iSNVs) an M. africanum elevation, weak host association, no transmission signal, and marked single-sputum undercapture. Implications of all the available evidenceWithin-host diversity holds public health promise as densely captured minority variants could refine transmission chain reconstruction, flag mixed infection, and surface emerging drug resistance earlier than consensus genomes allow. Realizing that promise requires sampling and measurement dense enough to recover the signal that single sputum specimens miss.