This study investigated the capability of electromembrane extraction (EME) as a general technique for peptides, by extracting complex pools of peptides comprising in total of 5953 different substances, varying in size from seven to 16 amino acids. Electromembrane extraction was conducted from a sample adjusted to pH 3.0 and utilized a liquid membrane consisting of 2-nitrophenyl octyl ether and carvacrol (1:1 w/w), containing 2% (w/w) di(2-ethylhexyl) phosphate. The acceptor phase was 50 mM phosphoric acid (pH 1.8), the extraction time was 45 min, and 10 V was used. High extraction efficiency, defined as a higher peptide signal in the acceptor than the sample after extraction, was achieved for 3706 different peptides. Extraction efficiencies were predominantly influenced by the hydrophobicity of the peptides and their net charge in the sample. Hydrophobic peptides were extracted with a net charge of +1, while hydrophilic peptides were extracted when the net charge was +2 or higher. A computational model based on machine learning was developed to predict the extractability of peptides based on peptide descriptors, including the grand average of hydropathy index and net charge at pH 3.0 (sample pH). This research shows that EME has general applicability for peptides and represents the first steps toward in silico prediction of extraction efficiency.
Keywords: LC‐MS; electromembrane extraction; machine learning; peptides; sample preparation.
© 2024 The Author(s). Journal of Separation Science published by Wiley‐VCH GmbH.