Correlation between the structure and skin permeability of compounds.

Area of Science:

• Computational chemistry
• Pharmacokinetics
• Toxicology

Background:

• Skin permeability is a critical factor in drug absorption and toxicity.
• Developing accurate predictive models for skin permeability is essential for drug discovery.
• Existing quantitative structure-activity relationship (QSAR) models have limitations in accuracy and sample size.

Purpose of the Study:

• To develop a robust quantitative structure-activity/toxicity relationship (QSAR/QSTR) model for predicting skin permeability.
• To apply support vector machine (SVM) and genetic algorithms for enhanced model performance.
• To validate the model's predictive power on a large dataset of 274 compounds.

Main Methods:

• A three-descriptor quantitative structure-activity/toxicity relationship (QSAR/QSTR) model was developed.
• Support vector machine (SVM) algorithm was employed for nonlinear modeling.
• Genetic algorithms were used for feature selection and optimization.
• A dataset of 274 compounds was utilized for training and testing.

Main Results:

• The optimal SVM model achieved a coefficient of determination (R²) of 0.946 and root mean square (rms) error of 0.253 for the training set (139 compounds).
• The model demonstrated strong performance on the test set (135 compounds) with R² of 0.872 and rms error of 0.302.
• The developed SVM model exhibited superior statistical performance compared to existing literature models, particularly with a larger test set.

Conclusions:

• The study successfully developed a nonlinear QSAR model for skin permeability using SVM.
• The SVM-based model provides a more accurate and reliable prediction of skin permeability.
• This approach enhances the prediction of compound behavior in biological systems, aiding drug development and safety assessment.