Year: 2026
Type: preprint
Abstract: Purification is essential for isolating compounds for reliable characterization and subsequent experiments; however, suitable chromatographic conditions are often still selected through trial and error. Herein, we present a data-driven framework that recommends a stationary phase or separation mode, solvent, and solvent ratio to support thin-layer chromatography (TLC) screening and column chromatography (CC) purification. Using purification records from publications and laboratory notebooks, we developed models to predict TLC and CC conditions. Models using product structures alone or in combination with reactant and agent information were compared across prediction tasks and data sources. Experimental evaluation of 11 previously unreported compounds demonstrated that ranked recommendations could guide TLC screening and the selection of CC purification conditions. Additionally, a web application was developed to provide users with flexible access to the prediction models. The proposed framework supports the selection of practical purification conditions and provides a foundation for integrating purification planning into future autonomous experimental workflows.
Source: ChemRxiv
Language: English
Subfield: Materials Chemistry
Field: Materials Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7220687930