Classification of drugs in absorption classes using the classification and regression trees (CART) methodology.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Pharmacokinetics

Background:

• Quantitative structure-activity relationship (QSAR) studies are crucial for drug discovery.
• Predicting drug absorption is vital for oral bioavailability.
• Molecular descriptors encode structural and physicochemical properties.

Purpose of the Study:

• To evaluate Classification and Regression Trees (CART) for QSAR modeling of drug absorption.
• To assess the predictive power of CART models using molecular descriptors.
• To identify key molecular descriptors influencing drug absorption.

Main Methods:

• Developed CART models using 141 drug-like molecules.
• Employed over 1400 molecular descriptors (2D and 3D) as predictors.
• Validated model performance using cross-validation and an external test set.
• Utilized variable ranking to determine descriptor importance.

Main Results:

• CART models demonstrated high descriptive and predictive accuracy for drug absorption.
• Cross-validation and external test set validation confirmed model robustness.
• Variable ranking highlighted n-octanol/water partition coefficient (logP) and polar surface area (PSA) as highly important descriptors.
• CART successfully identified known key descriptors for intestinal absorption.

Conclusions:

• CART is a valuable tool for QSAR analysis in drug discovery.
• The models accurately predict drug absorption based on molecular properties.
• logP and PSA are critical determinants of drug absorption, as confirmed by CART analysis.