Combination of Statistical Methods and Fourier Transform Ion Cyclotron Resonance Mass Spectrometry for More Comprehensive, Molecular-Level Interpretations of Petroleum Samples

Abstract

Complex petroleum mass spectra obtained by Fourier-transform ion cyclotron resonance mass spectrometry (FTICR MS) were successfully interpreted at the molecular level by applying principle component analysis (PCA) and hierarchical clustering analysis (HCA). A total of 40 mass spectra were obtained from 20 crude oil samples using both positive and negative atmospheric pressure photoionization (APPI). Approximately 400,000 peaks were identified at the molecular level. Conventional data analyses would have been impractical with so much data. However, PCA grouped samples into score plots based on their molecular composition. In this way, the overall compositional difference between samples could be easily displayed and identified by comparing score and loading plots. HCA was also performed to group and compare samples based on selected peaks that had been grouped by PCA. Subsequent heat map analyses revealed detailed compositional differences among grouped samples. This study demonstrates a promising new approach for studying multiple, complex petroleum samples at the molecular level.

Application of Clustering Methods for Interpretation of Petroleum Spectra from Negative-Mode ESI FT-ICR MS

“…In a previous study of positive and negative-mode atmospheric pressure photo ionization spectra, 13 sulfur and nitrogen contents and TAN values were important in grouping samples. In this study, nitrogen content was not important for grouping samples.…”

“…Detailed information on data preprocessing can be found in the previous literature. 13 Briefly explained, each sample was analyzed in triplicate for statistical significance. Triplicate spectra were later combined into a single spectrum, resulting in 10 mass spectra and an equal number of peak lists.…”

“…Detailed descriptions of the methods have been reported previously. 13,18 Briefly, principle component analysis is used to reduce the dimensionality of data while retaining the variation in the original data as much as possible. Principal components (PCs) are generated by linear combination of the original data set.…”

“…Thus, the clustering pattern was very similar to that obtained previously using PCA. 13 The clustering of samples is shown in Figure 4c. In the first cluster, each of the black and red lines represents crude01 and 03 samples, respectively.…”

“…As no prior knowledge about the data is required, clustering is known as an unsupervised method. Although the two clustering methods, PCA and HCA, were successfully used in previous study, 13 no systematic investigations to find the best method for petroleomic studies have yet been reported.…”