18Efforts to eradicate tuberculosis are hampered by the rise and spread of antibiotic 19 resistance. Several large-scale projects have aimed to specifically link clinical mutations to 20 resistance phenotypes, but they were limited in both their explanatory and predictive powers. 21 Here, we combine functional genomics and phylogenetic associations using clinical strain 22 genomes to decipher the architecture of isoniazid resistance and search for new resistance 23 determinants. This approach has allowed us to confirm the main target route of the antibiotic, 24 determine the clinical relevance of redox metabolism as an isoniazid resistance mechanism 25 and identify novel candidate genes harboring resistance mutations in strains with previously 26 unexplained isoniazid resistance. This approach can be useful for characterizing how the 27 tuberculosis bacilli acquire resistance to new antibiotics and how to forestall them. 28 29 45 we can detect rifampicin resistance with a 92% sensitivity, but the figure drops to 87% for 46 isoniazid and 58% for ethambutol. 6 However, there is still a knowledge gap as the catalog of 47 62 63 One way to unveil the genetic basis of resistance is by means of functional genomics, like 64 transposon mutagenesis approaches. This technique involves the genetic alteration of every 65 gene in the genome for explicit genotype-phenotype associations, 8,9 thus revealing more 66 genetic determinants than regular association studies do. This approach successfully 67 overcomes the shortcomings of genetic association studies: it can be used in a prospective 68 way, as it involves the systematic generation and testing of resistance mutants; it can detect 69 both genes with large and small effects on resistance; and it explicitly detects genes that 70 increase sensitivity when disrupted, thus indicating which genes are most promising for 71 treatments to prevent or reverse the evolution of resistance. However, transposon 72 mutagenesis alters the gene by disrupting it, highly informative about the biology of 73 resistance but limited in clinical explanation potential as most type of mutations found in 74 clinical resistance of M. tuberculosis are single nucleotide polymorphisms. Conversely, the 75 low diversity of the MTBC, clonality and the fact that clinical resistance is encoded in the 76 encoded in the chromosome makes Mycobacterium tuberculosis amenable for phylogenetic 77 association tests 10 which can determine which mutations are associated with resistance in 78 the bacterial phylogeny and are thus clinically relevant. 79 80In this paper, we provide a combined approach that uses functional genomics and 81 phylogenetic inference from clinical data to provide an in-depth picture of resistance to the 82 first-line antibiotic isoniazid. Isoniazid is a well-studied drug, yet we are still unable to 83 determine the causal mutation in around 6% of resistant strains, 11 although some 84 researchers have reported up to 25% in certain settings. 12 Here, we systematically 85 determi...