Accuracy, linearity, and statistical differences in comparative quantification in untargeted plant metabolomics using LC-ESI-Orbitrap-MS

Abstract High-resolution mass spectrometers, particularly when paired with liquid chromatography, are the instrument of choice for untargeted metabolomics approaches. Instruments, such as the Orbitrap, offer high sensitivity, selectivity, and exceptional mass accuracy, though they pose certain technical challenges, complicating absolute and comparative quantification. Consequently, method validation is crucial to ensure reliable results, as untargeted metabolomics approaches require the detection and quantification of a large number of metabolites in a broad dynamic range. Methods can be assessed using performance characteristics like accuracy and linearity to ensure analytical reliability. This study evaluates the suitability of untargeted metabolomics methods for discovery-based investigations. A stable isotope–assisted strategy was used with wheat extracts analyzed by a Q Exactive HF Orbitrap. Results showed that 70% of all detected 1327 metabolites displayed non-linear effects in at least one of the nine dilution levels employed. However, when considering fewer levels, 47% of all metabolites demonstrated linear behavior in at least four levels (i.e., a difference factor of 8). Moreover, the analysis further suggests that the observed abundances in less concentrated samples and those outside the linear range were mostly overestimated compared to expected abundances, but hardly ever underestimated. Consequently, during statistical analysis, which is an important step in prioritizing detected metabolites and correlating them with the biological hypothesis, the number of false-positives was not inflated, but the number of false-negatives might be increased. Generally, (non-)linear behavior did not correlate with specific compound classes or polarity, suggesting non-linearity is not easily predictable based on chemical structures. Graphical Abstract