EME methods guided by machine learning!
In collaboration with scientists from the University of the Balearic Islands, we have developed the first machine learning framework for quantitative prediction and in silico optimization of electromembrane extraction (EME). Using data from 182 chemically diverse compounds and five generic liquid membranes, we trained neural network models capable of predicting extraction recoveries directly from molecular properties and experimental conditions. To make the technology accessible, we also developed EME Predictor, a free desktop application that recommends optimal extraction conditions for new analytes. This work demonstrates how EME method development can move beyond trial-and-error toward data-driven, predictive, and more sustainable analytical workflows...