New artificial neural network models for risk assessment of skin sensitization using amino acid derivative assay, KeratinoSens™, human cell line activation test and in silico structural alert parameter

Area of Science:

• Toxicology
• Computational Chemistry
• Dermatology

Background:

• Estimating the point of departure (PoD) is vital for next-generation risk assessment (NGRA) of skin sensitization.
• The murine local lymph node assay (LLNA) is the gold standard for skin sensitization potential, providing EC3 values for dermal quantitative risk assessment (QRA).

Purpose of the Study:

• To develop artificial neural network (ANN) models for predicting LLNA EC3 values.
• To enhance these models by integrating data from the Amino Acid Derivative Reactivity Assay (ADRA) to broaden applicability.
• To enable predictive QRA for skin sensitization without animal testing.

Main Methods:

• Artificial neural network (ANN) analysis was employed to build prediction models.
• Descriptors from ADRA, using molar and gravimetric concentrations, were correlated with LLNA EC3 values.
• ANN models incorporated parameters from GL497-adopted methods and were compared with existing models and integrated judgment criteria.

Main Results:

• ADRA-derived descriptors showed significant correlations with LLNA EC3 values.
• The developed ANN models demonstrated a strong correlation with LLNA EC3 values.
• Predicted EC3 values from molar and gravimetric concentrations correlated well, as did predictions from ANN models using ADRA versus DPRA.

Conclusions:

• ANN models integrating ADRA data can accurately predict EC3 values, serving as reliable PoDs for QRA.
• The prediction accuracy of these ANN models, combined with GHS classification criteria, is comparable to established methods.
• This study facilitates broader QRA for skin sensitization, reducing reliance on animal testing and promoting safer chemical assessments.