A quantitative in silico model for predicting skin sensitization using a nearest neighbours approach within...

Area of Science:

• Toxicology
• Computational Chemistry
• Dermatology

Background:

• Allergic contact dermatitis results from dermal chemical exposure.
• Assessing chemical skin sensitization is crucial for public health and regulatory compliance.
• European Union Regulation 1223/2009 mandates an end to animal testing for cosmetic ingredients, necessitating alternative methods.

Purpose of the Study:

• To develop and validate a quantitative in silico model for predicting chemical skin sensitization potency.
• To provide a non-animal testing alternative for assessing the risk of allergic contact dermatitis.
• To support regulatory frameworks like the EU Regulation 1223/2009 by offering reliable predictive tools.

Main Methods:

• A Nearest Neighbours in silico model was created using 1096 murine local lymph node (LLNA) assay results.
• The model predicts the EC3 value (effective concentration for a threefold increase in stimulation index) based on similar compounds within the same mechanistic pathway.
• Model performance was evaluated using internal and external datasets, assessing accuracy against various classification systems (e.g., GHS).

Main Results:

• The in silico model demonstrated strong predictive capability for the Globally Harmonized System (GHS) skin sensitization categories.
• It correctly classified 64% of chemicals in an external test set.
• Discrepancies included 25% over-prediction and 11% under-prediction of GHS categories in unseen data.

Conclusions:

• The developed Nearest Neighbours model offers a reliable, non-animal method for predicting skin sensitization potency.
• This tool can aid in risk assessment and regulatory compliance, particularly concerning allergic contact dermatitis.
• Further refinement could improve the prediction accuracy for GHS classifications, enhancing chemical safety evaluations.