Resolving fluorescence spectra of Maillard reaction products formed on bovine serum albumin using parallel factor analysis.

Formation of Maillard reaction products (MRPs) is increasingly studied by the use of fluorescence spectroscopy, and most often, by measuring single excitation/emission pairs or use of unresolved spectra. However, due to the matrix complexity and potential co-formation of fluorescent oxidation products on tryptophan and tyrosine residues, this practice will often introduce errors in both identification and quantification. The present study investigates the combination of fluorescence excitation emission matrix (EEM) spectroscopy and parallel factor analysis (PARAFAC) to resolve the EEMs into its underlying fluorescent signals, allowing for better identification and quantification of MRPs. EEMs were recorded on a sample system of bovine serum albumin incubated at 40 °C for up to one week with either glucose, methylglyoxal or glyoxal added. Ten unique PARAFAC components were resolved, and assignment was attempted based on similarity with fluorescence of pure standards of MRPs and oxidation products and reported data from literature. Of the ten fluorescent PARAFAC components, tyrosine and buried and exposed tryptophan were resolved and identified, as well as the formation of specific MRPs (argpyrimidine and Nα-acetyl-Nδ-(5-methyl-4-imidazolon-2-yl)ornithine) and tryptophan oxidation products (kynurenine and dioxindolylalanine). The formation of the PARAFAC resolved protein modifications were qualitatively validated by liquid chromatography-mass spectrometry.

Detection of protein oxidation products by fluorescence spectroscopy and trilinear data decomposition: Proof of concept.

Current analytical methods studying protein oxidation modifications require laborious sample preparation and chromatographic methods. Fluorescence spectroscopy is an alternative, as many protein oxidation products are fluorescent. However, the complexity of the signal causes misinterpretation and quantification errors if single emission spectra are used. Here, we analyzed the entire fluorescence excitation-emission matrix using the trilinear decomposition method parallel factor analysis (PARAFAC). Two sample sets were used: a calibration set based on known mixtures of tryptophan, tyrosine, and four oxidation products, and a second sample set of oxidized protein solutions containing UV-illuminated ß-lactoglobulin. The PARAFAC model succeeded in resolving the signals of the model systems into the pure fluorophore components and estimating their concentrations. The estimated concentrations for the illuminated ß-lactoglobulin samples were validated by liquid chromatography-mass spectrometry. Our approach is a promising tool for reliable identification and quantification of fluorescent protein oxidation products, even in a complex protein system.

Using deep learning to evaluate peaks in chromatographic data.

Analysis of untargeted gas-chromatographic data is time consuming. With the earlier introduction of the PARAFAC2 (PARAllel FACtor analysis 2) based PARADISe (PARAFAC2 based Deconvolution and Identification System) approach in 2017, this task was made considerably more time-efficient. However, there are still a number of manual steps in the analysis which require data analytical expertise. One of these is the need to define whether or not each PARAFAC2 resolved component represents a peak suitable for integration. As the peaks may change in both shape and location on the elution time-axis, this presents a problem which cannot be readily solved by applying a linear classifier, such as PLS-DA (Partial Least Squares regression for Discriminant Analysis). As part of our ongoing efforts to further automate analysis of Gas Chromatography with Mass Spectrometry (GC-MS), we therefore explore a convolutional neural network classifier, capable of handling these shifts and variations in shape. The theory of convolutional neural networks and application on vector samples is briefly explained, and the performance is tested against a PLS-DA classifier, a shallow artificial neural network and a locally weighted regression model. The models are built on a training set with PARAFAC2 resolved components from eight different aroma related GC-MS runs with a total of over 70,000 elution profile samples, and validated using another, independent, GC-MS dataset. Based on Receiver Operating Characteristic curves (ROC) and manual analysis of the misclassified cases, it is shown that the convolutional network consistently outperforms the competing models, yielding an Area Under the Curve (AUC) value of 0.95 for peak classification. Examples are given illustrating that this new approach provides convincing means to automatically assess and evaluate modelled elution profiles of chromatographic data and thereby remove this laborious manual step.