Skin sensitization prediction using quantum chemical calculations: a theoretical model for the S(N)Ar domain

It is widely accepted that skin sensitization begins with the sensitizer in question forming a covalent adduct with a protein electrophile or nucleophile. We investigate the use of quantum chemical methods in an attempt to rationalize the sensitization potential of chemicals of the S(N)Ar reaction domain. We calculate the full reaction profile for 23 chemicals with experimental sensitization data. For all quantitative measurements, we find that there is a good correlation between the reported pEC3 and the calculated barrier to formation of the low energy product or intermediate (r(2) = 0.64, N = 12) and a stronger one when broken down by specific subtype (r(2) > 0.9). Using a barrier cutoff of ∼10 kcal/mol allows us to categorize 100% (N = 12) of the sensitizers from the nonsensitizers (N = 11), with just 1 nonsensitizer being mispredicted as a weak sensitizer (9%). This model has an accuracy of ∼96%, with a sensitivity of 100% and a specificity of ∼91%. We find that the kinetic and thermodynamic information provided by the complete profile can help in the rationalization process, giving additional insight into a chemical's potential for skin sensitization.