Times are changing but order matters: transferable prediction of small-molecule liquid chromatography retention times.

A two-step machine-learning method enables transferable prediction of small-molecule retention times across chromatographic conditions without target-system training data.

• Why it matters: Accurate retention time prediction is crucial for small-molecule separation, but existing models struggle with transferability due to changes in chromatographic conditions, limiting their practical use.
• What they did: The authors developed a two-step approach: first predicting retention order indices considering chromatographic conditions, then mapping these indices to absolute retention times, tested on reversed-phase chromatography.
• The result: This method surpasses existing models trained on target datasets and reveals how specific chromatographic conditions can significantly alter retention order, improving robustness and applicability.