To address the challenge of comparing entire chromatograms, three tile-based comparative analysis methods (Fisher ratio, pairwise, and relative variance ranking) were applied to data collected by comprehensive two-dimensional gas chromatography with vacuum ultraviolet spectroscopy detection (GC × GC-VUV). The tile-based scheme, originally developed for time-of-flight mass spectrometry (TOFMS) detection, was adapted herein to rank analytes on a hit list according to the metric for each method. We analyzed a GC × GC-VUV data set comprising 14 gas oils from five sample classes: straight-run (SR), light cycle oil (LCO), coker (CK), hydroconverted (HDC), and hydrotreated (HDT). Three replicates each of HDT and LCO gas oil GC × GC-VUV data were analyzed by Fisher ratio analysis with 459 out of 557 analyte hits exhibiting at least a 2-fold change, which were visualized by 2D chromatogram projections. Additionally, multivariate curve resolution-alternating least-squares was applied to obtain the pure VUV spectrum of each analyte, which was readily classified into five compound classes (saturates, olefins, mono-, di-, and triaromatics) and projected onto the 2D separation space to visualize the positions of these compound classes. Next, pairwise analysis, either between replicates of the same sample or between samples within the same sample class was studied. A concentration ratio >1.2 was determined as the 99% confidence level threshold for distinguishing analyte hits from background variation. Finally, relative variance analysis of all 24 gas oil chromatograms combined with principal component analysis enabled unsupervised differentiation of the five gas oil sample classes.
Ma et al. (Mon,) studied this question.