Scope and limitation of some approaches to predicting contact hypersensitivity

Allergic contact dermatitis (ACD) is the endpoint of a complex sequence of events. Construction of a predictive computational assay requires a combination of statistical analyses, semi-empirical modelling, chemical acumen and a good understanding of prohapten activation. Main events on the path to the altered self are chemical and microbial conversion on and in the stratum corneum (SC), epidermal penetration, non-specific binding, and finally hapten-protein conjugation. This report reviews the principal approaches currently taken in the prediction of ACD, using computer-based SAR, QSAR and expert systems, and also includes the BIOSAR models, which combine statistical analysis with chemical knowledge. Specifically designed to predict the ACD potential of chemicals in man, the BIOSAR approach of two-value regression analysis achieved a success rate of 88% when trained on a set of 72 compounds of wide diversity, with a sensitivity of 83% and specificity of 100%.