Applying support-vector machine learning algorithms toward predicting host-guest interactions with cucurbit[7]uril.

Area of Science:

• Supramolecular Chemistry
• Computational Chemistry
• Chemical Biology

Background:

• Machine learning (ML) applications in supramolecular chemistry are underexplored.
• Predicting molecular interactions is crucial for designing chemical technologies and drug delivery systems.

Purpose of the Study:

• To evaluate the utility of kernel-based support vector machine learning (SVM) for predicting small molecule binding coefficients with cucurbit[7]uril (CB[7]).
• To explore the potential of ML in supramolecular chemistry and drug delivery applications.

Main Methods:

• Utilized density functional theory (DFT) calculations for training data.
• Employed kernel-based support vector machine learning (SVM) algorithms.
• Experimentally validated ML predictions for drug binding.
• Performed mass transfer simulations to assess drug release.

Main Results:

• SVMs demonstrated predictive ability for small molecule-CB[7] binding coefficients.
• Accurate predictions for TAK-580 (strong binding) and selumetinib (poor binding) were experimentally validated.
• Cucurbit[8]uril (CB[8]) showed preferential binding to selumetinib, indicating tunable release potential.
• Demonstrated independent tuning of TAK-580 release by CB[7] without affecting selumetinib release.

Conclusions:

• Machine learning approaches are effective for predicting small molecule recognition by macrocycles.
• ML holds significant potential for advancing supramolecular chemistry and the development of sophisticated drug delivery systems.
• Findings support the broader utility of ML in chemical technology development.