Optimization of the Data Treatment Steps of a Non-targeted LC-MS-Based Workflow for the Identification of Trace Chemical Residues in Honey

Non-targeted screening (e.g., suspected-target) is emerging as an attractive tool to investigate the occurrence of contaminants in food. The sample preparation and instrument analysis steps are known to influence the identification of analytes with non-targeted workflows, especially for complex matrices. However, for methods based on mass spectrometry, the impact of the post-analysis data treatment (e.g., feature extraction) on the capacity to correctly identify a contaminant at trace level is currently not well understood. The aim of the study was to investigate the influence of seven post-analysis data treatment parameters on the non-targeted identification of trace contaminants in honey using high-performance liquid chromatography coupled to hybrid quadrupole time-of-flight mass spectrometry (HPLC-QTOF-MS). Seven compounds reported as veterinary drugs for honeybees were applied as model compounds. Among the parameters studied, the expansion window for chromatogram extraction and the average scans included in the spectra influenced significantly the identification process results. The optimized data treatment was applied to the non-targeted screening of veterinary drugs, pesticides, and other contaminants in 55 honey samples as a proof of concept. Among the 43 compounds included in a library of honey-related compounds that was used for screening, eight compounds were tentatively identified in at least one honey sample. The tentative identity of two of these compounds (tylosin A and hydroxymethylfurfural) was further confirmed with analytical standards.