General strategies to increase the repeatability in non-target screening by liquid chromatography-high resolution mass spectrometry

This article focuses on the data evaluation of non-target high-resolution LC-MS profiles of water samples. Taking into account multiple technical replicates, the difficulties in peak recognition and the related problems of false positive and false negative findings are systematically demonstrated. On the basis of a combinatorial approach, different models involving sophisticated workflows are evaluated, particularly with regard to the repeatability. In addition, the improvement resulting from data processing was systematically taken into consideration where the recovery of spiked standards emphasized that real peaks of interest were barely or not removed by the derived filter criteria. The comprehensive evaluation included different matrix types spiked with up to 263 analytical standards which were analyzed repeatedly leading to a total number of more than 250 injections that were incorporated in the assessment of different models of data processing. It was found that the analysis of multiple replicates is the key factor as, on the one hand, it provides the option of integrating valuable filters in order to minimize the false positive rate and, on the other hand, allows correcting partially false negative findings occurring during the peak recognition. The developed processing strategies including replicates clearly point to an enhanced data quality since both the repeatability as well as the peak recognition could be considerably improved. As proof of concept, four different matrix types, including a wastewater treatment plant (WWTP) effluent, were spiked with 130 isotopically labeled standards at different concentration levels. Despite the stringent filter criteria, at 100 ng L-1 recovery rates of up to 93% were reached in the positive ionization mode. The proposed model, comprising three technical replicates, filters less than 5% and 2% of the standards recognized at 100 and 500 ng L-1, respectively and thus indicates the general applicability of the presented strategies. (C) 2016 Elsevier B.V. All rights reserved.