Bedaquiline (BDQ) and clofazimine (CFZ) are core drugs for treatment of multidrug resistant tuberculosis (MDR-TB), however, our understanding of the resistance mechanisms for these drugs is sparse which is hampering rapid molecular diagnostics. To address this, we employed a unique approach using experimental evolution, protein modelling, genome sequencing, and minimum inhibitory concentration data (MIC) combined with genomes from a global strain collection of over 14,151 Mycobacterium tuberculosis complex isolates. Overall, 189 genomic variants causing elevated BDQ and/or CFZ MICs could be discerned, with 175 (95.1%) variants affecting the transcriptional repressor (Rv0678) of an efflux system (MmpS5-MmpL5). Structural modelling of Rv0678 suggests four major mechanisms that confer resistance: impairment of DNA binding, reduction in protein stability, disruption of protein dimerization, and reduction in affinity for its fatty acid ligand. These modelling and experimental techniques will improve personalized medicine in an impending drug resistant era.