Predicting the Host-Guest Binding Gibbs Free Energy for Anion Guests

Area of Science:

• Supramolecular Chemistry
• Computational Chemistry

Background:

• Host-guest chemistry is crucial for developing functional molecular systems in drug delivery, sensing, and adaptive materials.
• Experimental challenges in host-guest system development include complex synthesis and achieving selective binding.

Purpose of the Study:

• To present a computational model for predicting Gibbs free energies of binding in host-guest systems.
• To overcome experimental limitations in discovering new hosts and guests.

Main Methods:

• A two-component machine learning model was developed.
• The model utilizes electrostatic interactions and 3D molecular fingerprints.
• A curated dataset of 247 host-guest systems was employed.

Main Results:

• The model achieved an R² of 0.77 and a mean average error of 5.3 kJ/mol on an external test set.
• The computational approach successfully distinguished between strong and weak binders for a mixHC[8] macrocycle.
• The model's predictions aligned with newly acquired experimental data.

Conclusions:

• The developed computational model enables efficient high-throughput screening for host-guest discovery.
• This approach accelerates the design of novel functional molecular systems.
• It addresses key experimental challenges in supramolecular chemistry.