A method detection limit for the analysis of natural organic matter via Fourier transform ion cyclotron resonance mass spectrometry.

Fourier Transform Ion Cyclotron Resonance mass spectra (FT-ICR-MS) of natural organic matter are complex and consist of several thousands of peaks. The corresponding mass to charge ratios (m/z) and signal intensities result from analytes and noise. The most commonly applied way of distinguishing between analyte and noise is a fixed signal-to-noise ratio below which a detected peak is considered noise. However, this procedure is problematic and can yield ambiguous results. For example, random noise peaks can occur slightly above the signal-to-noise threshold (false positives), while peaks of low abundance analytes may occasionally fall below the fixed threshold (false negatives). Thus, cumulative results from repeated measurements of the same sample contain more peaks than a single measurement. False positive and false negative signals are difficult to distinguish, which affects the reproducibility between replicates of a sample. To target this issue, we tested the feasibility of a method detection limit (MDL) for the analysis of natural organic matter to identify peaks that can reliably be distinguished from noise by estimating the uncertainty of the noise. We performed 556 replicate analyses of a dissolved organic matter sample from the deep North Pacific on a 15 T FT-ICR-MS; each of these replicate runs consisted of 500 cumulated broadband scans. To unambiguously identify analyte peaks in the mass spectra, the sample was also run at time-consuming high-sensitivity settings. The resulting data set was used to establish and thoroughly test a MDL. The new method is easy to establish with software help, does only require the additional analysis of replicate blanks (low time increase), and can implement all steps of sample preparation. Especially when analysis time does not allow for replicate runs, major merits of the MDL are reliable removal of false positive (noise) peaks and better reproducibility, while the risk of losing analytes with low signal intensities (false negative) is comparatively low. When replicate analyses are feasible, the removal of all singly detected peaks is further recommended, as these have the highest probability of being noise peaks. We suggest that the here proposed detection limit should become routine in FT-ICR-MS data processing.