The incidence of severe adverse drug reactions is approximately 7% in hospital patients. In many cases the adverse event is difficult to predict or even explain on the basis of the known pharmacology of the causative agent. It is not infrequent in the context of multiple drug therapy, which complicates identification of the culprit. This can present a major problem in the management of chronic diseases such as tuberculosis or epilepsy. Pharmacogenetics offers one approach to reducing the incidence of drug-induced adverse reactions but has recognised limitations as a predictive tool and little role in diagnosis. Proteomics may have some predictive value but is likely to be of greater use in diagnosis-for example by recognising a drug signature in an accessible tissue. This may be possible on a blood sample or biopsy taken at presentation. Alternatively an in vitro assay that replaced rechallenging the patient with a drug would be helpful. The goal is to identify the causative drug permitting resumption of treatment with a safer alternative.