Highly predictive and interpretable models for PAMPA permeability.

Area of Science:

• Pharmacology
• Computational Chemistry
• Drug Discovery

Background:

• Cell membrane permeability is a key factor influencing oral drug absorption and bioavailability.
• Predicting drug permeability in silico is essential for efficient drug design and development.
• Existing methods for permeability assessment can be time-consuming and resource-intensive.

Purpose of the Study:

• To develop and validate a highly predictive in silico model for assessing drug molecule permeability.
• To provide medicinal chemists with a tool for designing novel compounds with improved drug-like properties.
• To accelerate the lead optimization phase in drug discovery through accurate permeability predictions.

Main Methods:

• A large dataset of 7488 compound entries (5435 unique molecules) measured by parallel artificial membrane permeability assay (PAMPA) was utilized.
• Support Vector Regression (SVR) and Support Vector Classification (SVC) models were developed using customized molecular descriptors.
• Model performance was evaluated using the area under the curve of the receiver operating characteristic (AUC-ROC).

Main Results:

• The SVR model predicted quantitative permeability data for 1364 compounds with an AUC-ROC of 0.90.
• The SVC model accurately predicted permeability for the remaining data with an AUC-ROC of 0.88.
• The developed models demonstrate high predictive accuracy for cell membrane permeability.

Conclusions:

• The in silico SVR model is a powerful and predictive tool for medicinal chemists.
• This computational approach facilitates the design and synthesis of drug candidates with enhanced permeability.
• The tool aids in accelerating lead optimization and improving the overall drug discovery process.