Chemometrics, Comprehensive Two-Dimensional gas chromatography and “omics” sciences: Basic tools and recent applications

Highlights

• Chemometrics allows trend recognition that would not be discernible using manual assessment.

• GC × GC data structure, pre-processing steps and algorithms restrictions are discussed.

• Chemometrics strategies applied in omics fields and its misuse are evaluated.

Abstract

The advent of Comprehensive Two-dimensional Gas Chromatography (GC × GC) as a practical and accessible analytical tool had a considerable impact on analytical procedures associated to the so-called “omics” sciences. Specially when GC × GC is hyphenated to mass spectrometers or other multichannel detectors, in a single run it is possible to separate, detect and identify up to thousands of metabolites. However, the resulting data sets are exceedingly complex, and retrieving proper biochemical information from them demands powerful statistical tools to deal effectively with the massive amount of information generated by GC × GC. Nevertheless, the obtention of results valid on a chemical and biological standpoint depends on a deep understanding by the analyst of the fundamentals both of GC × GC and chemometrics. This review focuses on the basics of contemporary, fundamental chemometric tools applied to proccessing of GC × GC obtained from metabolomic, petroleomic and foodomic analyses. Here, we described the fundamentals of pattern recognition methods applied to GC × GC. Also, we explore how different detectors affect data structure and approaches for better data handling. Limitations regarding data structure and deviations from linearity are stressed for each algorithm, as well as their typical applications and expected output.

Introduction

Regardless of the line of research followed, the analyst should aim for an analysis that is quick, efficient and as clean as possible. The proper selection of techniques both for sample preparation and for the analysis of the extracts is a significant step to reduce overall time and cost. Comprehensive and deep knowledge about intermolecular interactions among all chemical species involved, as well as of the nature of the sample, is relevant for the optimization of these parameters [1]. Therefore, the selection of sensitive and compatible techniques is paramount in ensuring both adequate detectability and use of reduced amounts of sample for study. When the analysis is focused on the so-called “omics” sciences, the use of robust, high-sensitivity and high throughput techniques is not only desirable but mandatory. One of the omics, Metabolomics, is based on analytical strategies applied with the objective of identifying and in some cases quantifying changes in the endogenous metabolites of biological systems in response to factors that can be intrinsic and extrinsic [2]. Complete metabolomic strategies deal with both primary and secondary metabolites that have broad physical and chemical properties such as lipids, sugars, amino acids, etc. Thus, given the great complexity and variation in the classes of compounds involved, the metabolome becomes an even greater analytical challenge when compared to other omics approaches [3].

Consolidated techniques such as conventional liquid or gas chromatography, despite their high separation power and sensitivity, may not be sufficient for the resolution and identification of the compounds present in samples related to metabolomic analysis, or to provide data sets that are useful to retrieve the information relevant to the characterization of these samples. In this context, multidimensional chromatographic techniques and especially Comprehensive Two-dimensional Gas Chromatography (GC × GC) can be considered the most significant advance in the area of separations since the invention of capillary chromatography in 1958 [4,5]. The latter is based on systems fitted with two chromatographic columns coated with stationary phases of distinct separation mechanisms - the first and second dimension (¹D and ²D) columns, connected in tandem by an interface called modulator [6]. The addition of a second dimension for separation results in an increase in the overall system peak capacity n corresponding to n₁ × n₂ (where n₁ and n₂ are the peak capacities of the ¹D and ²D columns, respectively). This corresponds to a geometric increment, which is much higher than that possible in conventional GC using typical strategies to improve selectivity and efficiency – for example, a twofold increase in peak capacity in GC demands an impracticable increment in analysis time [4]. Therefore, extremely complex samples that were previously evaluated generically by observing bulk properties can now be studied in detail at the molecular level with the application of GC × GC: the composition of these samples can be understood in a simple and structured way. Another aspect regarding the application of GC × GC to “omics” sample sets relates to the data processing methods to be applied. In conventional chromatography, it is common to use as input for information retrieval data sets consisting of tables containing information about retention time, peak area and MS identification of the detected analytes. However, in GC × GC, the slicing of chromatographic bands during the modulation process and the improvement in the amount of detected peaks makes this process non-trivial. Its performance requires increased attention during the reconstitution of the multidimensional chromatographic space to avoid the association between peaks other than a single compound. In addition, considering the size of the information extracted from a single sample on a discrete GC × GC run, it is impossible to manually interpret hundreds of distinct chromatographic peaks within an acceptable time window. Therefore, the verification of classical chromatographic parameters such as peak area, resolution and sample fingerprinting is no longer a trivial step due to the structure of the data generated by GC × GC: the high density of data generated requires the use of data processing techniques in which the search for information is based on strategies that demand reduced intervention of the analyst - typically, multivariate chemometric tools.

The use of these chemometric techniques applied to GC × GC data allows the verification of trends or recognition of patterns that would not be discernible using conventional manual inspection and assessment. In this context, GC × GC users rely on chemometric techniques of data analysis to extract reliable and accurate sample information. In this paper, we intend to review the most popular chemometric tools used for pattern recognition in GC × GC, addressing also the application of these techniques to GC × GC analysis of “omics” sample sets, either for exploration of the systems or for sample classification.