Fig 1.
MALDI-TOF-MS spectrum of PEG castor oil.
The spectrum shows many regularly spaced peaks, which make up several overlapping distributions from different ethoxylated components.
Fig 2.
Steps of investigation are shown through two routes, where multivariate analysis of the full MALDI-MS spectrum is compared to analysis of the extracted distribution features.
Table 1.
Number of batches of ethoxylates analysed.
Fig 3.
PCA score plots of MALDI data of polysorbate 80.
Scores plot for the first two principal components obtained from full MALDI spectra unscaled (a) and scaled (b) of three different types of Polysorbate 80, showing that types 2 and 3 are indistinguishable. Observations are coloured by type.
Fig 4.
PCA score plots of extracted distribution features of MALDI data of polysorbate 80.
Scores plot for the first two principal components obtained from features extracted from the distributions of peaks identified in the mass spectra of Polysorbate 80. The plot shows that there is distinction between all types of Polysorbate 80. Observations are coloured by type.
Table 2.
PLS classification results of test data for PEG castor oil using extracted distribution features.
Table 3.
PLS classification results of test data for PEG castor oil using the full spectrum.
Fig 5.
Distributions isolated within the MALDI-MS spectrum of a typical batch of PEG castor oil.
Plots relate to (a) type 1 and (b) type 2.
Fig 6.
Partial least squares regression (PLS-R) analysis of PEG Castor oil using extracted features of MALDI-MS data.
The scores plot for the first two PLS latent variables is shown in (a) with the values predicted by the PLS-R model using 3-fold cross-validation in (b). Here Type 1 and Type 2 are encoded as 1.0 and 2.0 respectively. The dotted line represents a predicted value of 1.5, taken as the boundary between classes.
Fig 7.
Variable importance in projection (VIP) plots from PLS-R of PEG Castor oil MALDI data.
(A) Full spectrum analysis, showing important variables as MALDI peaks (m/z); (B) Extracted features analysis showing more informative results, based on general distribution shifts (distribution maxima m/z) and ratios of distributions and therefore constituents.
Table 4.
Types of Oleth-20 analysed.
Fig 8.
Distributions obtained from the peak maxima for the two highest intensity ethoxylated constituents of Oleth-20, coloured according to type.
(a) shows the distribution attributed to ethoxylated oleyl alcohol and (b) that assigned to ethoxylated cetyl alcohol.