文摘
We developed a computational technique to assist in thelarge-scale identification of charged metabolites. Theelectrophoretic mobility of metabolites in capillary electrophoresis-mass spectrometry (CE-MS) was predictedfrom their structure, using an ensemble of artificial neuralnetworks (ANNs). Comparison between relative migrationtimes of 241 various cations measured by CE-MS andpredicted by a trained ANN ensemble produced a correlation coefficient of 0.931. When we used our techniqueto characterize all metabolites listed in the KEGG liganddatabase, the correct compounds among the top threecandidates were predicted in 78.0% of cases. We suggestthat this approach can be used for the prediction of themigration time of any cation and that it represents apowerful method for the identification of uncharacterizedCE-MS peaks in metabolome analysis.