Summary

Bedaquiline (BDQ) is the newest approved anti-TB drug, yet resistance is increasing rapidly, especially in South African communities where BDQ has been in use since 2012. We developed a statistically robust, pip-installable tool, catomatic, enabling users to rapidly generate mutation-effect catalogues for resistance prediction. Applied to 11,867 clinical M. tuberculosis isolates with WGS and BDQ susceptibility data, our approach achieved high sensitivity and produced the largest BDQ resistance catalogue to date. We confirm that mmpL5 and mmpS5 are not associated with resistance, highlight the clinical significance of minor alleles in Rv0678, and validate that the most common resistance-conferring variants are loss-of-function mutations in Rv0678. The method is fast, flexible, and features a user-friendly API, making it adaptable for diverse applications, from experimental screening to diagnostic development. This work advances understanding of BDQ resistance and introduces sustainable catalogue creation for BDQ and other TB drugs.