Times are changing but order matters: Transferable prediction of small molecule liquid chromatography retention times

Thousands of publications on the prediction of small molecule retention times were published during the last decades. The ultimate goal is, without doubt, the transferable prediction of retention times: We want to train a model on a certain set of compounds from one dataset and then use the model to predict retention times for a different set of compounds from another dataset. Unfortunately, retention times may change massively, even for nominally identical chromatographic conditions. Retention order is much better retained, yet even the retention order of compounds may change if chromatographic conditions vary. Here, we systematically study what chromatographic conditions result in notable changes in retention order. We then present a machine learning model that can predict retention order or, more precisely, a retention order index, taking into account chromatographic conditions. Finally, we show how to map the retention order index to retention times. Disentangling these two task finally enables retention time prediction across chromatographic conditions and compound classes.