An evaluation of mathematical models for predicting skin permeability

Area of Science:

• Pharmacology
• Computational Chemistry
• Dermatology

Background:

• Predicting skin permeability is crucial for transdermal drug delivery.
• Mathematical models range from empirical to deterministic, with ongoing debate on their efficacy.
• Current trends favor complex molecular descriptors over simpler parameters.

Purpose of the Study:

• To evaluate and compare the predictive performance of various mathematical models for skin permeability.
• To identify the most accurate models using a comprehensive experimental dataset.
• To understand the common features and limitations of effective skin permeability models.

Main Methods:

• A comprehensive experimental dataset of skin permeability for 124 compounds was compiled.
• Seven different mathematical models were compared.
• Model performance was assessed based on predictive accuracy.

Main Results:

• Mitragotri's deterministic model provided the best skin permeability predictions.
• The quantitative structure-permeability relationship (QSPR) model by Potts and Guy was the second best.
• More complex empirical QSPR models using numerous molecular descriptors showed unsatisfactory predictive performance.

Conclusions:

• Deterministic and simpler QSPR models, like Mitragotri's and Potts and Guy's, are superior for predicting skin permeability.
• These effective models share common assumptions regarding the lipid matrix pathway and utilize parameters like octanol-water partition coefficient and molecular size.
• Complex QSPR models with many collinear descriptors are less effective due to unclear mechanistic relationships.