Phenolic Acid-β-Cyclodextrin Complexation Study to Mask Bitterness in Wheat Bran: A Machine Learning-Based QSAR...

Area of Science:

• Biochemistry
• Computational Chemistry
• Materials Science

Background:

• Cyclodextrins, particularly beta-cyclodextrin, are widely used for encapsulating hydro-sensitive molecules across various industries.
• Beta-cyclodextrin is crucial for masking the bitterness of phenolic compounds, such as those found in wheat bran.
• A need exists for predictive models to assess beta-cyclodextrin's bitterness masking efficacy for diverse phenolic compounds.

Purpose of the Study:

• To develop robust predictive models for assessing beta-cyclodextrin's bitterness masking capabilities.
• To investigate the binding interactions between beta-cyclodextrin and phenolic acids.
• To identify key molecular descriptors influencing the binding affinity.

Main Methods:

• Utilized a dataset of 20 phenolic acids docked into the beta-cyclodextrin cavity to determine binding constants.
• Integrated docking data with topological, topographical, and quantum-chemical features.
• Employed machine learning, specifically a combination of genetic algorithm (GA) and multiple linear regression (MLR), to build Quantitative Structure-Activity Relationship (QSAR) models.

Main Results:

• Developed three distinct ML/QSAR models for predicting binding constants with high accuracy.
• Achieved excellent performance metrics, including correlation coefficients of 0.969 for training and 0.984 for test sets.
• Identified molecular features positively contributing to binding affinity, such as six-membered rings, branching, electronegativity, and polar surface area.

Conclusions:

• The developed ML/QSAR models provide a reliable tool for predicting beta-cyclodextrin binding affinity with phenolic compounds.
• Understanding the key molecular features influencing binding can guide the selection of compounds for effective bitterness masking.
• This study enhances the application of cyclodextrins in pharmaceuticals, food science, and agriculture by providing predictive insights.