Developing a quantitative structure-property relationships (QSPR) model using Caco-2 cell bioavailability indicators...

Area of Science:

• Pharmacology and Toxicology
• Computational Chemistry
• Food Science

Background:

• Bioavailability (BA) is crucial for drug and functional ingredient efficacy.
• Assessing BA indicators like epithelial barrier function, apparent permeability (P_app), and efflux ratio is vital.
• Traditional methods for BA assessment can be time-consuming and resource-intensive.

Purpose of the Study:

• To measure BA indicators for 84 phytochemicals using Caco-2 cells.
• To develop predictive models for BA indicators using machine learning and QSPR.
• To utilize Isomeric Simplified Molecular Input Line Entry System (SMILES) for molecular descriptor generation.

Main Methods:

• Caco-2 cell assays were employed to measure epithelial barrier function, P_app, and efflux ratio.
• Phytochemicals were encoded into molecular descriptors using Isomeric SMILES via PaDEL-Descriptor and alvaDesc.
• Quantitative structure-property relationship (QSPR) models were built using machine learning algorithms.
• High-performance liquid chromatography (HPLC) was used for phytochemical analysis.

Main Results:

• The QSPR models demonstrated high predictive accuracy for transepithelial electrical resistance (TEER), P_app, and efflux ratio.
• For P_app, R² values reached 0.95 for the training set and 0.91 for the test set.
• The model for efflux ratio explained 92% of the variance, with R²_Test of 0.85.

Conclusions:

• A predictive system for bioavailability indicators (TEER, P_app, efflux ratio) was successfully developed using QSPR models.
• This predictive system can accelerate the discovery of novel functional ingredients and pharmaceutical compounds.
• The study highlights the potential of computational approaches in predicting phytochemical behavior.