Reliable Peak Selection for Multisample Analysis with Comprehensive Two-Dimensional Chromatography

Augmented visualization by computer vision and chromatographic fingerprinting on comprehensive two-dimensional gas chromatographic patterns: Unraveling diagnostic signatures in food volatilome

Computer Vision is an approach of Artificial Intelligence (AI) that conceptually enables "computers and systems to derive useful information from digital images" giving access to higher-level information and "take actions or make recommendations based on that information". Comprehensive two-dimensional chromatography gives access to highly detailed, accurate, yet unstructured information on the sample's chemical composition, and makes it possible to exploit the AI concepts at the data processing level (e.g., by Computer Vision) to rationalize raw data explorations. The goal is the understanding of the biological phenomena interrelated to a specific/diagnostic chemical signature. This study introduces a novel workflow for Computer Vision based on pattern recognition algorithms (i.e., combined untargeted and targeted UT fingerprinting) which includes the generation of composite Class Images for representative samples' classes, their effective re-alignment and registration against a comprehensive feature template followed by Augmented Visualization by comparative visual analysis. As an illustrative application, a sample set originated from a Research Project on artisanal butter (from raw sweet cream to ripened butter) is explored, capturing the evolution of volatile components along the production chain and the impact of different microbial cultures on the finished product volatilome. The workflow has significant advantages compared to the classical one-step pairwise comparison process given the ability to realign and pairwise compare both targeted and untargeted chromatographic features belonging to Class Images resembling chemical patterns from many different samples with intrinsic biological variability.

PARAFAC2×N: Coupled decomposition of multi-modal data with drift in N modes

Analysis of GC×GC-TOFMS data for large numbers of poorly-resolved peaks, and for large numbers of samples remains an enduring problem that hinders the widespread application of the technique. For multiple samples, GC×GC-TOFMS data for specific chromatographic regions manifests as a 4^th order tensor of I mass spectral acquisitions, J mass channels, K modulations, and L samples. Chromatographic drift is common along both the first-dimension (modulations), and along the second-dimension (mass spectral acquisitions), while drift along the mass channel is for all practical purposes nonexistent. A number of solutions to handling GC×GC-TOFMS data have been proposed: these involve reshaping the data to make it amenable to either 2^nd order decomposition techniques based on Multivariate Curve Resolution (MCR), or 3^rd order decomposition techniques such as Parallel Factor Analysis 2 (PARAFAC2). PARAFAC2 has been utilised to model chromatographic drift along one mode, which has enabled its use for robust decomposition of multiple GC-MS experiments. Although extensible, it is not straightforward to implement a PARAFAC2 model that accounts for drift along multiple modes. In this submission, we demonstrate a new approach and a general theory for modelling data with drift along multiple modes, for applications in multidimensional chromatography with multivariate detection. The proposed model captures over 99.9% of variance for a synthetic data set, presenting an extreme example of peak drift and co-elution across two modes of separation.

Extending the breadth of saliva metabolome fingerprinting by smart template strategies and effective pattern realignment on comprehensive two-dimensional gas chromatographic data

Comprehensive two-dimensional gas chromatography with time-of-flight mass spectrometry (GC × GC-TOFMS) is one the most powerful analytical platforms for chemical investigations of complex biological samples. It produces large datasets that are rich in information, but highly complex, and its consistency may be affected by random systemic fluctuations and/or changes in the experimental parameters. This study details the optimization of a data processing strategy that compensates for severe 2D pattern misalignments and detector response fluctuations for saliva samples analyzed across 2 years. The strategy was trained on two batches: one with samples from healthy subjects who had undergone dietary intervention with high/low-Maillard reaction products (dataset A), and the second from healthy/unhealthy obese individuals (dataset B). The combined untargeted and targeted pattern recognition algorithm (i.e., UT fingerprinting) was tuned for key process parameters, the signal-to-noise ratio (S/N), and MS spectrum similarity thresholds, and then tested for the best transform function (global or local, affine or low-degree polynomial) for pattern realignment in the temporal domain. Reliable peak detection achieved its best performance, computed as % of false negative/positive matches, with a S/N threshold of 50 and spectral similarity direct match factor (DMF) of 700. Cross-alignment of bi-dimensional (2D) peaks in the temporal domain was fully effective with a supervised operation including multiple centroids (reference peaks) and a match-and-transform strategy using affine functions. Regarding the performance-derived response fluctuations, the most promising strategy for cross-comparative analysis and data fusion included the mass spectral total useful signal (MSTUS) approach followed by Z-score normalization on the resulting matrix.

Chemical fingerprinting strategies based on comprehensive two-dimensional gas chromatography combined with gas chromatography-olfactometry to capture the unique signature of Piemonte peppermint essential oil (Mentha x piperita var Italo-Mitcham)

Accurate, reliable, and informative mapping of untargeted and targeted components across many samples is here performed by combining off-line GC-Olfactometry (GC-O) and comprehensive two-dimensional gas chromatography (GC×GC) coupled to time-of-flight mass spectrometry with variable ionization energy (TOF MS featuring Tandem Ionization™). In particular, untargeted and targeted (UT) features patterns are processed by chromatographic fingerprinting, giving differential priority to potent odorants' retention-times regions. Distinguishing peppermint essential oil (EO) from Piedmont (Italy - Mentha × piperita L. var. Italo-Mitcham - Menta di Pancalieri EO), with its unique sensory fingerprint (i.e., freshness and long-lasting sweetness), from high-quality peppermint EOs produced in other areas poses a great challenge. Chromatographic UT fingerprinting provided a great chemical dimensionality by mapping more than 350 peak-regions at 70 eV and 135 at 12 eV. From them, 95 components were identified and responses compared to available literature. Then, potent odorants, detected by GC-O using the aroma extraction dilution analysis (AEDA), were tracked over the chromatographic space and tentatively identified. With the highest flavor dilution (FD), 1,8-cineole (eucalyptus, fresh, camphoraceous); menthone (minty, herbaceous); and menthofuran (minty, musty, petroleum-like) were highlighted. Responsible for creamy and coumarinic notes were the diasteroisomers of (3,6)-dimethyl-4,5,6,7-tetrahydrobenzo[b]-furan-2(3H)-one (i.e., menthofurolactones), detected in higher relative abundance in Pancalieri EOs. By prioritizing the investigation of volatiles on higher LogFD retention regions, including 131 untargeted/targeted features, Pancalieri EOs were separately clustered from United States samples. Besides pre-targeted analytes, additional untargeted features were post-processed for identification within marker chemicals. Myrtenyl methyl ether, ethyl 3-methyl butanoate, propyl-2-methylbutanoate, and (E)-2-hexenal were putatively identified. Of the "unknown - knowns" with diagnostic roles, all metadata were collected including low energy spectra at 12 eV, which were found to be highly complementary to 70 eV spectra.

Exploring extra dimensions to capture saliva metabolite fingerprints from metabolically healthy and unhealthy obese patients by comprehensive two-dimensional gas chromatography featuring Tandem Ionization mass spectrometry

This study examines the information potential of comprehensive two-dimensional gas chromatography combined with time-of-flight mass spectrometry (GC×GC-TOF MS) and variable ionization energy (i.e., Tandem Ionization™) to study changes in saliva metabolic signatures from a small group of obese individuals. The study presents a proof of concept for an effective exploitation of the complementary nature of tandem ionization data. Samples are taken from two sub-populations of severely obese (BMI > 40 kg/m²) patients, named metabolically healthy obese (MHO) and metabolically unhealthy obese (MUO). Untargeted fingerprinting, based on pattern recognition by template matching, is applied on single data streams and on fused data, obtained by combining raw signals from the two ionization energies (12 and 70 eV). Results indicate that at lower energy (i.e., 12 eV), the total signal intensity is one order of magnitude lower compared to the reference signal at 70 eV, but the ranges of variations for 2D peak responses is larger, extending the dynamic range. Fused data combine benefits from 70 eV and 12 eV resulting in more comprehensive coverage by sample fingerprints. Multivariate statistics, principal component analysis (PCA), and partial least squares discriminant analysis (PLS-DA) show quite good patient clustering, with total explained variance by the first two principal components (PCs) that increases from 54% at 70 eV to 59% at 12 eV and up to 71% for fused data. With PLS-DA, discriminant components are highlighted and putatively identified by comparing retention data and 70 eV spectral signatures. Within the most informative analytes, lactose is present in higher relative amount in saliva from MHO patients, whereas N-acetyl-D-glucosamine, urea, glucuronic acid γ-lactone, 2-deoxyribose, N-acetylneuraminic acid methyl ester, and 5-aminovaleric acid are more abundant in MUO patients. Visual feature fingerprinting is combined with pattern recognition algorithms to highlight metabolite variations between composite per-class images obtained by combining raw data from individuals belonging to different classes, i.e., MUO vs. MHO.

Graphical abstract

Untargeted/Targeted 2D Gas Chromatography/Mass Spectrometry Detection of the Total Volatile Tea Metabolome

Identifying all analytes in a natural product is a daunting challenge, even if fractionated by volatility. In this study, comprehensive two-dimensional gas chromatography/mass spectrometry (GC×GC-MS) was used to investigate relative distribution of volatiles in green, pu-erh tea from leaves collected at two different elevations (1162 m and 1651 m). A total of 317 high and 280 low elevation compounds were detected, many of them known to have sensory and health beneficial properties. The samples were evaluated by two different software. The first, GC Image, used feature-based detection algorithms to identify spectral patterns and peak-regions, leading to tentative identification of 107 compounds. The software produced a composite map illustrating differences in the samples. The second, Ion Analytics, employed spectral deconvolution algorithms to detect target compounds, then subtracted their spectra from the total ion current chromatogram to reveal untargeted compounds. Compound identities were more easily assigned, since chromatogram complexities were reduced. Of the 317 compounds, for example, 34% were positively identified and 42% were tentatively identified, leaving 24% as unknowns. This study demonstrated the targeted/untargeted approach taken simplifies the analysis time for large data sets, leading to a better understanding of the chemistry behind biological phenomena.

Highly Informative Fingerprinting of Extra-Virgin Olive Oil Volatiles: The Role of High Concentration-Capacity Sampling in Combination with Comprehensive Two-Dimensional Gas Chromatography

The study explores the complex volatile fraction of extra-virgin olive oil by combining high concentration-capacity headspace approaches with comprehensive two-dimensional gas chromatography, which is coupled with time of flight mass spectrometry. The static headspace techniques in this study are: (a) Solid-phase microextraction, with multi-polymer coating (SPME- Divinylbenzene/Carboxen/Polydimethylsiloxane), which is taken as the reference technique; (b) headspace sorptive extraction (HSSE) with either a single-material coating (polydimethylsiloxane—PDMS) or a dual-phase coating that combines PDMS/Carbopack and PDMS/EG (ethyleneglycol); (c) monolithic material sorptive extraction (MMSE), using octa-decyl silica combined with graphite carbon (ODS/CB); and dynamic headspace (d) with either PDMS foam, operating in partition mode, or Tenax TA™, operating in adsorption mode. The coverage of both targeted and untargeted 2D-peak-region features, which corresponds to detectable analytes, was examined, while concentration factors (CF) for a selection of informative analytes, including key-odorants and off-odors, and homolog-series relative ratios were calculated and the information capacity was discussed. The results highlighted the differences in concentration capacities, which were mainly caused by polymer-accumulation characteristics (sorptive/adsorptive materials) and its amount. The relative concentration capacity for homologues and potent odorants was also discussed, while headspace linearity and the relative distribution of analytes, as a function of different sampling amounts, was examined. This last point is of particular interest in quantitative studies where accurate data is needed to derive consistent conclusions.

Untargeted and Targeted Fingerprinting of Extra Virgin Olive Oil Volatiles by Comprehensive Two-Dimensional Gas Chromatography with Mass Spectrometry: Challenges in Long-Term Studies

Comprehensive two-dimensional gas chromatography coupled with mass spectrometric detection (GC × GC-MS) offers an information-rich basis for effective chemical fingerprinting of food. However, GC × GC-MS yields 2D-peak patterns (i.e., sample 2D fingerprints) whose consistency may be affected by variables related to either the analytical platform or to the experimental parameters adopted for the analysis. This study focuses on the complex volatile fraction of extra-virgin olive oil and addresses 2D-peak patterns variations, including MS signal fluctuations, as they may occur in long-term studies where pedo-climatic, harvest year, or shelf life changes are studied. The 2D-pattern misalignments are forced by changing chromatographic settings and MS acquisition. All procedural steps, preceding pattern recognition by template matching, are analyzed and a rational workflow defined to accurately realign patterns and analytes metadata. Signal-to-noise ratio (SNR) detection threshold, reference spectra extraction, and similarity match factor threshold are critical to avoid false-negative matches. Distance thresholds and polynomial transform parameters are key for effective template matching. In targeted analysis (supervised workflow) with optimized parameters, method accuracy reaches 92.5% (i.e., % of true-positive matches) while for combined untargeted and targeted ( UT) fingerprinting (unsupervised workflow), accuracy reaches 97.9%. Response normalization also is examined, evidencing good performance of multiple internal standard normalization that effectively compensates for discriminations occurring during injection of highly volatile compounds. The resulting workflow is simple, effective, and time efficient.

Characterization of Key Aroma Compounds in a Commercial Rum and an Australian Red Wine by Means of a New Sensomics-Based Expert System (SEBES)-An Approach To Use Artificial Intelligence in Determining Food Odor Codes

Although to date more than 10 000 volatile compounds have been characterized in foods, a literature survey has previously shown that only 226 aroma compounds, assigned as key food odorants (KFOs), have been identified to actively contribute to the overall aromas of about 200 foods, such as beverages, meat products, cheeses, or baked goods. Currently, a multistep analytical procedure involving the human olfactory system, assigned as Sensomics, represents a reference approach to identify and quantitate key odorants, as well as to define their sensory impact in the overall food aroma profile by so-called aroma recombinates. Despite its proven effectiveness, the Sensomics approach is time-consuming because repeated sensory analyses, for example, by GC/olfactometry, are essential to assess the odor quality and potency of each single constituent in a given food distillate. Therefore, the aim of the present study was to develop a fast, but Sensomics-based expert system (SEBES) that is able to reliably predict the key aroma compounds of a given food in a limited number of runs without using the human olfactory system. First, a successful method for the quantitation of nearly 100 (out of the 226 known KFOs) components was developed in combination with a software allowing the direct use of the identification and quantitation data for the calculation of odor activity values (OAV; ratio of concentration to odor threshold). Using a rum and a wine as examples, the quantitative results obtained by the new SEBES method were compared to data obtained by applying an aroma extract dilution analysis and stable isotope dilution assays required in the classical Sensomics approach. A good agreement of the results was found with differences below 20% for most of the compounds considered. By implementing the GC × GC data analysis software with the in-house odor threshold database, odor activity values (ratio of concentration to odor threshold) were directly displayed in the software pane. The OAVs calculated by the software were in very good agreement with data manually calculated on the basis of the data obtained by SIDA. Thus, it was successfully shown that it is possible to characterize key food odorants with one single analytical platform and without using the human olfactory system, that is, by "artificial intelligence smelling".

Differentiation of Rums Produced from Sugar Cane Juice (Rhum Agricole) from Rums Manufactured from Sugar Cane Molasses by a Metabolomics Approach

A large set of volatiles (a metabolome) was isolated by SAFE distillation from 25 high priced rums prepared from sugar cane juice (SCJ) and 26 high priced rums manufactured from sugar cane molasses (SCM). The volatile fractions were first analyzed by comprehensive two-dimensional gas chromatography time-of-flight mass spectrometry (GC×GC-TOF-MS), and the "comprehensive template matching fingerprinting" was used to extract the entire features present in the respective set of volatile compounds. After raw data pretreatment, chemometrics was used to locate marker compounds. Following, a sparse-partial-least-squares discriminant analysis ( sPLS-DA) and a partial-least-squares discriminant analysis (PLS-DA) were applied to a training data set for creating a model. The model was validated using leave-one-out cross validation and tested over an independent data set to evaluate its predictive power. The characteristic fingerprint resulted in a 100% correct classification of sugar cane juice rums, thus achieving the first aim of locating markers for these higher quality rums. Then, past-processing identification within the discriminant features was done to characterize 12 significant marker compounds as 1-decanol, γ-dodecalactone, ethyl 3-methylbutanoate, ethyl nonanoate, 3-furancarboxaldehyde, 1-hexanol, β-ionone, 2- and 3-methylbutanol, methyl decanoate, 3-octanol, and 2-undecanone. Quantitation of eight selected markers by stable isotope dilution assays confirmed higher concentrations in SCJ compared to SCM and served as the final proof to differentiate both types of spirits.

Headspace solid-phase microextraction coupled to comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry for the analysis of aerosol from tobacco heating product

A method involving headspace solid-phase microextraction (HS-SPME) and comprehensive two-dimensional gas chromatography coupled to time-of-flight mass spectrometry (GC×GC-TOFMS) was developed and optimised to elucidate the volatile composition of the particulate phase fraction of aerosol produced by tobacco heating products (THPs). Three SPME fiber types were studied in terms of extraction capacity and precision measurements. Divinylbenzene polydimethylsiloxane appeared as the most efficient coating for these measurements. A central composite design of experiment was utilised for the optimization of the extraction conditions. Qualitative and semi-quantitative analysis of the headspace above THP aerosol condensate was carried out using optimised extraction conditions. Semi-quantitative analyses of detected constituents were performed by assuming that their relative response factors to the closest internal standard (itR) were equal to 1. Using deconvoluted mass spectral data (library similarity and reverse match >750) and linear retention indices (match window of ±15 index units), 205 peaks were assigned to individual compounds, 82 of which (including 43 substances previously reported to be present in tobacco) have not been reported previously in tobacco aerosol. The major volatile fraction of the headspace contained ketones, alcohols, aldehydes, alicyclic hydrocarbons alkenes, and alkanes. The method was further applied to compare the volatiles from the particulate phase of aerosol composition of THP with that of reference cigarette smoke and showed that the THP produced a less complex chemical mixture. This new method showed good efficiency and precision for the peak areas and peak numbers from the volatile fraction of aerosol particulate phase for both THP and reference cigarettes.

Comprehensive Chemical Fingerprinting of High-Quality Cocoa at Early Stages of Processing: Effectiveness of Combined Untargeted and Targeted Approaches for Classification and Discrimination

This study investigates chemical information of volatile fractions of high-quality cocoa (Theobroma cacao L. Malvaceae) from different origins (Mexico, Ecuador, Venezuela, Columbia, Java, Trinidad, and Sao Tomè) produced for fine chocolate. This study explores the evolution of the entire pattern of volatiles in relation to cocoa processing (raw, roasted, steamed, and ground beans). Advanced chemical fingerprinting (e.g., combined untargeted and targeted fingerprinting) with comprehensive two-dimensional gas chromatography coupled with mass spectrometry allows advanced pattern recognition for classification, discrimination, and sensory-quality characterization. The entire data set is analyzed for 595 reliable two-dimensional peak regions, including 130 known analytes and 13 potent odorants. Multivariate analysis with unsupervised exploration (principal component analysis) and simple supervised discrimination methods (Fisher ratios and linear regression trees) reveal informative patterns of similarities and differences and identify characteristic compounds related to sample origin and manufacturing step.

Cultivable gut bacteria provide a pathway for adaptation of Chrysolina herbacea to Mentha aquatica volatiles

BACKGROUND

A chemical cross-talk between plants and insects is required in order to achieve a successful co-adaptation. In response to herbivory, plants produce specific compounds, and feeding insects respond adequately7 to molecules produced by plants. Here we show the role of the gut microbial community of the mint beetle Chrysolina herbacea in the chemical cross-talk with Mentha aquatica (or watermint).

RESULTS

By using two-dimensional gas chromatography-mass spectrometry we first evaluated the chemical patterns of both M. aquatica leaf and frass volatiles extracted by C. herbacea males and females feeding on plants, and observed marked differences between males and females volatiles. The sex-specific chemical pattern of the frass paralleled with sex-specific distribution of cultivable gut bacteria. Indeed, all isolated gut bacteria from females belonged to either α- or γ-Proteobacteria, whilst those from males were γ-Proteobacteria or Firmicutes. We then demonstrated that five Serratia marcescens strains from females possessed antibacterial activity against bacteria from males belonging to Firmicutes suggesting competition by production of antimicrobial compounds. By in vitro experiments, we lastly showed that the microbial communities from the two sexes were associated to specific metabolic patterns with respect to their ability to biotransform M. aquatica terpenoids, and metabolize them into an array of compounds with possible pheromone activity.

CONCLUSIONS

Our data suggest that cultivable gut bacteria of Chrysolina herbacea males and females influence the volatile blend of herbivory induced Mentha aquatica volatiles in a sex-specific way.

Cluster-based comparison of the peptide mass fingerprint obtained by MALDI-TOF mass spectrometry. A case study: long-term stability of rituximab

We evaluated the use of the peptide mass fingerprint (PMF) obtained by matrix assisted laser desorption and ionization (MALDI) time-of-flight mass spectrometry (TOF-MS) to track changes in the structure of a protein. The first problem we had to overcome was the inherent complexity of the PMF, which makes it difficult to compare. We dealt with this problem by developing a cluster-based comparison algorithm which takes into account the proportional error made by the mass spectrometer. This procedure involves grouping together similar masses in an intelligent manner, so that we can determine which data correspond to the same peptide (any slight differences can be explained as experimental errors), and which of them are too different and thus more likely to represent different peptides. The proposed algorithm was applied to track changes in a commercially available monoclonal antibody (mAb), namely rituximab (RTX), prepared under the usual hospital conditions and stored refrigerated (4 °C) and frozen (-20 °C) for a long term study. PMFs were obtained periodically over three months. For each checked time, five replicates of the PMFs were obtained in order to evaluate the similarities between them by means of the occurrences of the particular peptides (m/z). After applying the algorithm to the PMF, different approaches were used to analyse the results. Surprisingly, all of them suggested that there were no differences between the two storage conditions tested, i.e. the RTX samples were almost equally well preserved when stored refrigerated at 4 °C or frozen at -20 °C. The cluster-based methodology is new in protein mass spectrometry and could be useful as an easy test for major changes in proteins and biopharmaceutics for diverse applications in industry and other fields, and could provide additional stability data in relation to the practical use of anticancer drugs.