Tuberculosis (TB) kills more people globally than any other infectious disease, and progress towards its elimination is stagnating amid concerns about prevalence of drug resistance. Different lineages of M. tuberculosis differ in both their intrinsic drug susceptibility and their propensity to acquire antimicrobial resistance. Most studies of resistance use datasets enriched for drug-resistant isolates, creating an ascertainment bias that limits our understanding of how resistance evolves in the wider bacterial population. In this joint project between the UK Health Security Agency (UKHSA) and the University of Birmingham, we will analyse a national collection of M. tuberculosis genomes from England. The collection brings together drug-susceptible and drugresistant isolates with linked enhanced surveillance records and epidemiological data. We will use existing resistance catalogues to characterise drug resistance across lineages and apply genome-wide association methods to investigate genetic factors associated with resistance emergence, pathogenicity and heteroresistance. To address gaps in these catalogues, we will also examine whether advanced machine learning approaches such as genome and protein language models can improve resistance prediction for rare and novel genotypes. The student will develop skills in microbial genomics, bioinformatics and applied machine learning through investigating how emerging machine learning methods can support effective surveillance of TB and other pathogens.