Answers are grounded in this paper and cite their sources.
Rifampicin (RIF) resistance is used as a proxy for multidrug-resistant tuberculosis, but whether resistant Mycobacterium tuberculosis isolates lacking rpoC / rpoA compensation form an identifiable subset has rarely been tested with open genotype–phenotype data. Reanalyzing Comprehensive Resistance Prediction for Tuberculosis: an International Consortium data, we corrected three data-preparation hazards: conflation of RIF with the related drug rifabutin; isolates lacking a mutation record treated as confirmed wild-type; and a cluster of biologically implausible per-gene mutation counts, which were winsorized. After correction, only 239 isolates (0.53% of the RIF-phenotyped pool) had both a valid phenotype and mutation data, a subset that may not be randomly selected by lineage or geography; however, its resistance rate (30.1%) closely matched that of the wider population (30.3%; p = 0.95). Presence of rpoB was associated with resistance (odds ratio [OR] = 12.90, 95% confidence interval [CI]: 3.89–42.75), and the presence of compensatory rpoC or rpoA mutation was even more strongly associated (OR = 8.28, 95% CI: 4.06–16.88). A logistic-regression classifier reached an area under the curve of 0.75 (95% CI: 0.61–0.89). Eight resistant isolates carried rpoB without compensation and are nominated as candidates for future validation, not a validated finding. The scale of the corrections needed, a 24.6% drug-code misassignment and a genotype file covering 1.3% of phenotyped isolates, shows why reanalyses of public resistance datasets need provenance checks.
No takes yet. Share an insight, caveat, or question.
Vijayakumar et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: