Screening Cyclodextrin Complexes for Bisphenols with High Binding Performance Based on the Data-Driven Model.

Area of Science:

• Supramolecular Chemistry
• Computational Chemistry
• Materials Science

Background:

• Cyclodextrins (CDs) possess unique cavity structures enabling molecular recognition and binding.
• Modifying hydroxyl groups on CDs allows tuning of their binding capacities for specific applications.
• Designing modified CDs with desired binding properties is crucial for advancing chemical, pharmaceutical, and material sciences.

Purpose of the Study:

• To develop a data-driven model for predicting and assisting in the design of cyclodextrin hosts.
• To introduce a novel structure representation method for cyclodextrin/guest complexes.
• To validate the model's performance and demonstrate its utility in identifying high-affinity cyclodextrin binders.

Main Methods:

• Development of a data-driven predictive model utilizing a novel cyclodextrin/guest structure representation.
• Validation of the model through cross-validation (Q²=0.801) and test set evaluation (R²test=0.841).
• Application of the model in conjunction with fluorescence experiments to screen and characterize cyclodextrin hosts for bisphenol binding.

Main Results:

• The data-driven model demonstrated high predictive accuracy for cyclodextrin binding properties.
• Several cyclodextrin hosts with strong binding capacities for bisphenols were successfully identified, synthesized, and characterized.
• The model achieved a controlled average absolute error of 0.605 M⁻¹, confirming its feasibility for molecular design and data supplementation.

Conclusions:

• The proposed data-driven model serves as a valuable tool for theoretical assistance in cyclodextrin complex design.
• This approach can accelerate the discovery and optimization of cyclodextrins for industrial applications and scientific research.
• The study highlights the potential of computational methods to drive innovation in supramolecular chemistry and materials science.