Highly effective drugs for treating TB were introduced over 30 years ago, yet deaths from the disease continue to increase. New tools are needed, including drugs with activity against multi-drug resistant strains of Mycobacterium tuberculosis. Agents that reduce the duration and complexity of the current therapy would have a major impact on compliance and overall cure rate. In recent years, our understanding of the tubercle bacillus and its interaction with the human host has improved dramatically, particularly with the publication in 1998 of the complete genome sequence of M. tuberculosis H37Rv. New genetic tools have been developed and we can now ascertain the function of individual genes. Thus, many potential drug targets have been identified and a number demonstrated to be essential. Several lead compounds have been found, as well as a potential drug candidate, the nitroimidazopyran PA-824. A far greater effort is needed to translate basic research into drug discovery programmes. High throughput screening and rational design must be employed to find lead compounds acting against well-validated targets and a substantial increase in resources devoted to medicinal chemistry is required to take these leads and turn them into drugs. Models of mycobacterial persistence, in which compounds with potent sterilizing activity can be rapidly analysed, must be characterized. Finally, surrogate markers that give an early indication of treatment outcome would facilitate clinical trials.