Web-BO: towards increased accessibility of Bayesian optimisation (BO) for chemistry

Area of Science:

• Chemical Informatics
• Computational Chemistry
• Machine Learning in Chemistry

Background:

• Traditional chemical discovery relies on time- and resource-intensive trial-and-improvement methods.
• The vastness of chemical space necessitates advanced search strategies for efficient optimization.
• Bayesian optimization (BO) offers a powerful framework for guided chemical searches but lacks accessibility.

Purpose of the Study:

• To develop a user-friendly graphical user interface (GUI) for Bayesian optimization (BO) in chemical discovery.
• To enhance the accessibility of data-driven optimization tools for chemists.
• To integrate BO capabilities with existing electronic lab notebooks.

Main Methods:

• Development of Web-BO, a GUI-based platform for chemical optimization tasks.
• Demonstration of Web-BO's performance using an open-source dataset and emulator.
• Integration of Web-BO with the datalab electronic lab notebook.

Main Results:

• Web-BO provides a GUI to simplify the implementation of Bayesian optimization in chemistry.
• The platform successfully demonstrates performance for chemical optimization tasks.
• Seamless linking with electronic lab notebooks is achieved, improving workflow integration.

Conclusions:

• Web-BO significantly improves the accessibility of Bayesian optimization for chemical research.
• The GUI-based approach democratizes data-driven optimization in chemistry.
• This tool facilitates efficient discovery of chemical systems with desired properties.