Automation and AI-Powered Prediction in Chromatographic Separation.

Area of Science:

• Analytical Chemistry
• Computational Chemistry

Background:

• Chromatography, including TLC, CC, GC, and HPLC, is vital for chemical separations but often relies on empirical optimization, hindering reproducibility.
• The integration of laboratory automation and artificial intelligence (AI) offers a path to overcome these limitations.

Purpose of the Study:

• To develop a unified framework for AI-assisted chromatography, enabling predictive and programmable separations.
• To demonstrate the transformation of chromatography into a predictive science through automation, machine learning, and cross-method transfer.

Main Methods:

• Utilizing robotic systems for reproducible data acquisition in TLC and CC.
• Developing machine-learning models, including graph neural networks for enantioseparation, incorporating mechanistic constraints.
• Implementing multimodal frameworks for GC and uncertainty quantification for HPLC.

Main Results:

• Creation of transferable predictive models linking TLC R_f values to CC retention volumes.
• Accurate prediction of GC retention under dynamic conditions using molecular features and heating programs.
• Development of chirality-aware models for HPLC enantioseparation, providing separation probabilities.

Conclusions:

• AI-assisted chromatography, integrating automation and machine learning, significantly enhances prediction accuracy, interpretability, and transferability across methods.
• This unified framework accelerates chemical discovery and improves reproducibility by enabling predictive and programmable chromatographic separations.