Skin Sensitization: Reaction Mechanistic Applicability Domains for Structure-Activity Relationships Aynur O. Aptula,* ,† Grace Patlewicz, ‡ and David W. Roberts § SEAC, Unilever Colworth, Sharnbrook, Bedford, MK44 1LQ, England, European Chemicals Bureau TP582, Institute for Health and Consumer Protection, Joint Research Centre, European Commission, Via Fermi, 21020 Ispra (VA), Italy, and School of Pharmacy and Chemistry, Liverpool John Moores University, Byrom Street, Liverpool, L3 3AF, England Received March 16, 2005 The prediction of skin sensitization potential with minimum animal testing is currently of great importance in light of forthcoming legislation. A number of structure-activity relation- ships for skin sensitization have been published over the years, but their applicability has often been limited to structural classes. The concept of an applicability domain for a quantitative structure-activity relationship [(Q)SAR] is increasingly being viewed as key for the predictive application of (Q)SARs. This is particularly the case for skin sensitization if more widely applicable SARs are to be developed. In this paper, we analyze a recently published chemical data set for skin sensitization, apply reaction mechanistic criteria to domain classification, and evaluate the structure-activity trends observed within each of these mechanistic domains. Introduction Allergic contact dermatitis (ACD) is an eczematous skin disease affecting a significant minority of the general population worldwide, which can have a serious impact on quality of life (1, 2). At present, the only validated approaches to conclusively identify skin sensitization hazard are in vivo models such as the local lymph node assay (LLNA) (3, 4). There is an urgent need to develop novel approaches or risk assessment strategies to replace animal testing especially in view of legislative initiatives such as REACH (registration, evaluation, authorization of chemicals) as well as the 7th Amendment to the Cosmetics Directive (which poses a ban on animal testing for cosmetic ingredients by 2009, although some tests are exempted until 2013) (5, 6). Quantitative structure- activity relationships [(Q)SARs] are viewed as one of the most cost effective alternatives to estimate toxicity effects of chemicals since they have the potential to save time and money and minimize the use of animal testing. However, although the new and current EU legislation provides the possibility of using (Q)SARs instead of in vivo testing, there are neither accepted (Q)SARs for toxicological end points nor detailed guidance on how to develop (Q)SARs for risk assessment. As a consequence, uptake of (Q)SARs by competent authorities such as regulatory agencies has been very limited. Several initia- tives have recently emerged to increase the acceptance of (Q)SARs. The main principles for the validity of (Q)- SARs were identified in a workshop organized by CEFIC/ ICCA in Setubal in 2002 and were subsequently evalu- ated by OECD [as part of the ad hoc expert group for (Q)SARs] (7). These are now referred to as the “OECD principles”, which read as follows: “To facilitate the consideration of a (Q)SAR model for regulatory purposes, it should be associated with the following information: -a defined endpoint -an unambiguous algorithm -a defined domain of applicability -appropriate measures of goodness-of-fit, robustness, and predictivity -a mechanistic interpretation, if possible.” (Q)SARs, which fulfill these criteria, may in principle be applicable within the regulation practice to predict mammalian end points. Of these principles, perhaps the one that provokes the most debate is the domain of applicability. The domain of applicability aims to provide some guidance on the scope and boundaries for the use of a given (Q)SAR. Recently, a chemical data set for evaluation of alterna- tive approaches was published (8) listing skin sensitiza- tion potential, expressed as LLNA EC3% values, for 41 compounds covering a diverse range of structures. The EC3 is the dose required to give a 3-fold stimulation between treated and control groups in the LLNA. The EC3 is taken as a quantifier of the sensitization potential. These compounds were classified according to chemical class, e.g., ketone, aliphatic aldehyde, halogenated com- pound, etc., but no clear relationship between structure and sensitization potential was apparent from this clas- sification. The aim of this paper is to apply a chemical mechanism of action approach to reclassify these com- pounds into appropriate applicability domains. Approaches to Chemical Classification There are various ways in which a diverse set of chemicals can be grouped into classes according to some criterion of similarity (9). For example, they can be classified according to similarity in connectivity patterns (this may be useful, for example, where biological activity dependent on noncovalent binding to a specific receptor is being modeled), according to the presence of specific * To whom correspondence should be addressed. Tel: + 44 1234 264823. Fax: + 44 1234 264722. E-mail: nora.aptula@unilever.com. † Unilever Colworth. ‡ European Commission. § Liverpool John Moores University. 1420 Chem. Res. Toxicol. 2005, 18, 1420-1426 10.1021/tx050075m CCC: $30.25 © 2005 American Chemical Society Published on Web 08/20/2005