RGCxGC toolbox: An R-package for data processing in comprehensive two-dimensional gas chromatography-mass spectrometry

• Published in: Microchemical journal Vol. 156; p. 104830
• Main Authors: Quiroz-Moreno, Cristian, Furlan, Mayra Fontes, Belinato, João Raul, Augusto, Fabio, Alexandrino, Guilherme L., Mogollón, Noroska Gabriela Salazar
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2020
• Subjects: Comprehensive two-dimensional gas chromatography; Exploratory analysis; Metabolomics; Signal processing; Toolbox; Signal processing; Metabolomics; Exploratory analysis; Comprehensive two-dimensional gas chromatography; Toolbox
• ISSN: 0026-265X; 1095-9149
• DOI: 10.1016/j.microc.2020.104830

•End-to-end pipeline for two-dimensional gas chromatography-mass spectrometry from preprocessing to multivariate analysis.•Acceptable frame times for large number of samples for batch analysis.•GC × GC–MS data is used for machine learning with supervised and unsupervised learning.•Performance of the toolbox is exemplified with three datasets. Comprehensive two-dimensional gas chromatography (GC×GC) offers detailed chemical information about volatile and semivolatile analytes from complex samples. However, the high complexity of the data structure encourages the development of new tools for a more efficient data handling and analysis. Although some tools have already been presented to overcome this challenge, there is still need for improvement. In this manuscript, we present a toolbox containing a pipeline for end-to-end basic GC×GC data processing which can be used for both, signal pre-processing and multivariate data analysis. The pre-processing algorithms perform signal smoothing, baseline correction, and peak alignment, while the multivariate analysis is done through Multiway Principal Component Analysis (MPCA). The software is capable to prepare the chromatographic data for further applications with other chemometric tools, e.g.: cluster analysis, regression, discriminant analysis, etc. The performance of this new software was tested on in-house experimental dataset and on two other published datasets.