Artificial Intelligence-Based Quantitative Structure-Property Relationship Model for Predicting Human Intestinal...

Area of Science:

• Pharmaceutical Sciences
• Computational Chemistry
• Drug Discovery

Background:

• Oral medications dominate the pharmaceutical market, necessitating effective drug absorption through intestinal walls.
• Predicting drug absorption is crucial for efficient candidate screening and reducing drug development timelines.
• Existing prediction algorithms often focus solely on solubility, neglecting permeability.

Purpose of the Study:

• To develop an AI-based system for predicting human intestinal absorption (HIA) of active pharmaceutical ingredients (APIs).
• To enhance the accuracy of identifying highly permeable molecules for oral drug delivery.
• To utilize APIs with serotonergic activity as a relevant dataset for permeability prediction.

Main Methods:

• Developed a hierarchical AI system combining classification and regression models.
• Focused on predicting drug permeability as a key indicator of intestinal bioavailability.
• Utilized a dataset of APIs with serotonergic activity for model training and validation.

Main Results:

• The AI system achieved high accuracy in classifying molecules with high permeability.
• External validation demonstrated the correct selection of 38% of highly permeable molecules with zero false positives.
• The combined model approach expanded the capability to identify highly permeable compounds.

Conclusions:

• The AI-based system offers a promising tool for early-stage oral drug screening.
• Accurate in silico prediction of drug permeability can significantly aid drug discovery and development.
• The developed models and datasets are publicly available to facilitate further research.