CIME4R: Exploring iterative, AI-guided chemical reaction optimization campaigns in their parameter space

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Data Science

Background:

• Chemical reaction optimization (RO) generates large, high-dimensional datasets, challenging current analysis tools.
• The integration of artificial intelligence (AI) in RO introduces further complexity in understanding model predictions and guiding the optimization process.

Purpose of the Study:

• To introduce CIME4R, an open-source interactive web application designed for analyzing reaction optimization data and AI predictions.
• To facilitate scientists' comprehension of reaction parameter spaces, RO process evolution, critical factors, and AI model decisions.
• To enhance human-AI collaboration and trust calibration in AI-guided reaction optimization workflows.

Main Methods:

• Development of CIME4R, an open-source interactive web application.
• Testing and evaluation of CIME4R with domain experts through three case studies.
• Focus on interactive analysis of RO data and AI model predictions.

Main Results:

• CIME4R enables users to comprehend reaction parameter spaces and track RO process development.
• The application aids in identifying critical reaction factors and understanding AI model predictions.
• Domain experts utilized CIME4R to gain insights from past RO campaigns and inform future experimental decisions.

Conclusions:

• CIME4R is the first open-source interactive web application specifically for reaction optimization analysis.
• The tool supports informed decision-making by combining human expertise with computational precision in AI-guided RO.
• CIME4R has the potential to improve scientific workflows in reaction optimization and foster an open-source community.