Theoretically-derived molecular descriptors important in human intestinal absorption

Area of Science:

• Pharmacokinetics
• Computational Chemistry
• Drug Discovery

Background:

• Human intestinal absorption (IA) is a critical factor in drug efficacy and bioavailability.
• Predicting IA is essential for efficient drug development, reducing costly experimental failures.
• Quantitative structure-activity relationships (QSAR) are widely used to model pharmacokinetic properties.

Purpose of the Study:

• To develop a quantitative structure-human intestinal absorption relationship model.
• To identify key molecular descriptors influencing drug intestinal absorption using artificial neural network (ANN) modeling.
• To select the most significant descriptors for predicting intestinal absorption.

Main Methods:

• Collected data for 86 drug compounds and their experimentally-derived intestinal absorption values from literature.
• Calculated 57 global molecular descriptors for each compound.
• Employed a supervised artificial neural network (ANN) with a radial basis transfer function.
• Utilized a genetic algorithm for feature selection of molecular descriptors.
• Developed a genetic neural network (GNN) model with 15 input descriptors.

Main Results:

• The developed GNN model successfully correlated molecular descriptors with experimental human intestinal absorption values.
• The model identified lipophilicity, conformational stability, and intermolecular interactions (polarity, hydrogen bonding) as the most impactful descriptors.
• The best GNN model incorporated 15 selected molecular descriptors.

Conclusions:

• Artificial neural network modeling provides a robust method for predicting human intestinal absorption.
• Molecular descriptors related to lipophilicity, stability, and intermolecular forces are crucial for determining drug absorption.
• The developed GNN model can aid in the early-stage assessment of drug candidates' intestinal absorption potential.