Development of a Hierarchical Support Vector Regression-Based In Silico Model for Caco-2 Permeability

Area of Science:

• Pharmacokinetics and Drug Metabolism
• Computational Chemistry
• Cell Biology

Background:

• Drug absorption is crucial for drug discovery and development.
• The Caco-2 cell model is a standard in vitro tool for assessing intestinal drug absorption.
• Predicting intestinal permeability is complex due to passive and active transport mechanisms.

Purpose of the Study:

• To develop a predictive quantitative structure-activity relationship (QSAR) model for Caco-2 cell permeability.
• To effectively model both passive diffusion and transporter-mediated active transport of drugs.
• To provide a tool for early-stage drug discovery and development.

Main Methods:

• Utilized a hierarchical support vector regression (HSVR) machine learning approach.
• Generated a QSAR model based on chemical structure and Caco-2 permeability data.
• Validated the model's performance using training, test, and outlier datasets, along with statistical criteria.

Main Results:

• The HSVR model demonstrated strong agreement with experimental Caco-2 permeability values.
• The model's predictive capability was confirmed through rigorous statistical validation and mock testing.
• The developed QSAR model accurately captures complex drug transport mechanisms.

Conclusions:

• The HSVR-based QSAR model is a reliable tool for predicting Caco-2 permeability.
• This predictive model can significantly assist in the drug discovery and development pipeline.
• The study highlights the utility of advanced machine learning in pharmacokinetic assessments.