Pred-Skin

Pred-Skin predicts human skin sensitization potential of chemical compounds by integrating multiple QSAR models into a consensus naïve Bayes framework using diverse human, in vivo, in chemico, and in vitro data.

Key Features:

• Extensive Database Integration: Incorporates human repeat insult patch tests, human maximization tests, mouse local lymph node assays (LLNA), and data from direct peptide reactivity assay (DPRA), KeratinoSens, and the human cell line activation test (H-CLAT).
• Advanced Modeling Techniques: Builds individual QSAR models using open-source tools and QSAR best practices and integrates them into a consensus naïve Bayes model.
• High Predictive Accuracy: Reports external prediction performance with a correct classification rate of 89%, sensitivity of 94%, positive predicted value of 91%, specificity of 84%, and negative predicted value of 89%.
• Validation Against Known Sensitizers: Correctly identified nine out of eleven cosmetic ingredients known to cause skin sensitization that were excluded from the training set.

Scientific Applications:

• Alternative to Animal Testing: Provides in silico assessments of human skin sensitization that can serve as an alternative to animal-based assays.
• Safety Assessment of Industrial Chemicals: Supports safety assessment workflows for industrial chemicals by predicting sensitization potential.
• Regulatory Decision Support: Aids researchers and regulators in making informed decisions on chemical safety to support regulatory acceptance and public health protection.

Methodology:

Integrates human, in vivo, in chemico, and in vitro data to develop individual QSAR models using open-source tools and QSAR best practices, validates these models, and combines their outputs via a consensus naïve Bayes model.