Examples of unwanted variation when characterising dissolved organic matter using direct injection electrospray mass spectrometry and chemometrics

We describe a methodology for comparing and contrasting a large number of dissolved organic matter samples based on their electrospray ionization mass spectral characteristics. The analytical process and especially the electrospray process give rise to unwanted variation, which needs to be either mitigated, compensated for or monitored, and until now only few examples of multivariate comparison of large sets of dissolved organic matter mass spectral fingerprints where also unwanted variation is monitored have been published. The methodology comprises a tier of processes: (I) establishing a sample set including facilitator and quality control samples; (II) establishing pre-concentration/dilution factors to put samples to the same concentration domain; (III) chemical analysis by direct injection electrospray ionisation time-of-flight mass spectrometry; (IV) data pre-processing - normalisation and variable weighting - to emphasize variation relevant to the objective of the investigation; and (V) multivariate comparison and differentiation of a large number of samples by principal component analysis. The implementation of the methodology was demonstrated on two sample sets: terrestrial and oceanic dissolved organic matter from North America and Antarctica (set 1), and dissolved organic matter from Danish groundwaters (set 2). Samples were re-constituted from dry matter (set 1) or worked up directly (set 2). Artefacts originating from concentration effects were attempted removed by pre-concentration/dilution prior to chemical analysis, pre-concentration/dilution factors determined by either UV-(set 1) or negative mode ESI-MS screening (set 2). Normalisation factors and variable weights were established by use of facilitator sample sets. Artefacts related to unwanted variation induced by remaining concentration effects, sampling-, storage- and analytical processes are described and were monitored by independent quality control sample sets.