What this research found
Rifampicin resistance in tuberculosis is normally diagnosed by sequencing a short stretch of the rpoB gene, which means variants outside that window get missed. Every rifampicin record in the WHO 2023 Mycobacterium tuberculosis mutation catalogue was mapped onto the rifampicin-bound RNA polymerase structure, scored for how the substitution would disrupt drug binding, and distilled into a 10-variant surveillance and phenotyping panel. Distance from the drug tracks resistance closely: variants the WHO grades as resistance-associated sit a mean of 4.79 Å from rifampicin, against 26.88 Å for those of uncertain significance.
- Proximity to the drug separates the WHO confidence grades sharply. Resistance-associated variants average 4.79 Å from rifampicin (n = 127), uncertain ones 26.88 Å (n = 672), and unassociated ones 31.55 Å (n = 6), a difference with a Kruskal-Wallis p-value of 2.6e-62.
- Structural distance correlates negatively with epidemiological strength: Spearman rho = -0.4907 against the resistance odds ratio across 139 variants, and rho = -0.4091 against resistant-isolate counts across all 805 mapped variants.
- The final panel holds 10 variants across 9 distinct positions, split evenly between five classic high-confidence markers and five rare, borderline, or unresolved ones. S450L anchors the high end with an odds ratio of 245.31 from 10,859 resistant and 226 susceptible isolates, and sits 2.6 Å from the drug.
- Two variants at position 491 fall outside the rifampicin resistance-determining region that standard molecular assays amplify. I491F, at an odds ratio of 2.40 from 116 resistant and 88 susceptible isolates, is a diagnostic escaper missed by assays such as Xpert, while I491V remains unresolved with 14 resistant isolates and no susceptible ones on record.
- Most rpoB variation is predicted to be mechanistically inert. Of 779 missense variants modelled, 646 scored as neutral or benign, leaving 56 conformational disruptions, 45 steric clashes, 22 electrostatic repulsions, and 10 lost polar contacts or hydrogen bonds.
How it was done
All 1,502 rifampicin rpoB records in the WHO 2023 catalogue, pinned to a specific published release, were parsed down to 822 protein-coding variants with resolvable codon positions. A numbering calibration established that catalogue position N corresponds to UniProt position N plus 6, at 100% concordance across 506 residues, after which 805 variants were aligned onto chain C of the rifampicin-bound RNA polymerase structure. For each one the minimum heavy-atom distance to the bound drug was computed and combined with side-chain volume, hydropathy, formal charge at physiological pH, hydrogen-bond polarity, and proline or glycine conformational flags to assign a disruption mechanism. The 10 nominated variants were paired with a three-part laboratory control set: wild-type H37Rv as the susceptible baseline, S450L as the high-level resistance benchmark, and L430P, at an odds ratio of 2.13, as a low-level benchmark for calibrating assay sensitivity.
Data sources
- WHO 2023 Mycobacterium tuberculosis mutation catalogue — 48,152 rows, 1,502 rifampicin rpoB entries
- Protein Data Bank entry 5UHB — rifampicin-bound RNA polymerase, chain C with 1,126 resolved residues
- UniProt P9WGY9 (M. tuberculosis RpoB) and P0A8V2 (Escherichia coli RpoB) for cross-species numbering
Limitations
Disruption mechanisms are inferred from residue physicochemistry and distances in a single static crystal structure rather than measured, so the panel is a nomination for laboratory phenotyping rather than a validated assay. Most catalogue variants remain of uncertain significance, and one panel member has an undefined odds ratio because no susceptible isolate carrying it has been recorded.
Figures from this analysis
How this research was produced
K-Dense Web planned and ran this genomics investigation end to end — gathering the sources, carrying out the analysis, producing the figures, and drafting the report. The full session transcript, including every intermediate step, is available to view.


