Quantitative prediction of ionization effect on human skin permeability

Area of Science:

• Pharmacokinetics and Drug Delivery
• Computational Chemistry
• Dermal Science

Background:

• Skin permeability of active ingredients is significantly influenced by ionization in solution.
• Existing prediction models often fail to account for these ionization effects.
• Accurate prediction of skin penetration considering ionization is crucial for drug development.

Purpose of the Study:

• To develop reliable prediction models for in vitro human skin permeability coefficients that incorporate ionization effects.
• To create a comprehensive dataset for training and validating these models.
• To enable high-throughput screening and optimization of active ingredients for dermal delivery.

Main Methods:

• Curated a novel dataset of 322 in vitro human skin permeability coefficients for diverse permeants.
• Generated thousands of computational descriptors, including LogD (octanol-water distribution coefficient at a specific pH).
• Employed nonlinear support vector regression (SVR) and Gaussian process regression (GPR) with greedy descriptor selection.

Main Results:

• The SVR model demonstrated slightly superior performance over GPR.
• Achieved high external validation metrics: R²=0.94, RMSE=0.29, MAE=0.21.
• LogD was identified as a key descriptor for predicting ionization effects on skin permeability.

Conclusions:

• The developed SVR and GPR models accurately predict skin permeability considering ionization effects.
• These models satisfy stringent statistical validation criteria.
• The models facilitate high-throughput evaluation and optimization of active ingredients for dermal applications across various pH conditions.