Multi-block Data Analysis Research Articles

Enhancing mass spectrometry interpretability by ComDim-ICA multi-block analysis: Geographical and varietal traceability of Brazilian Coffea canephora

Mass spectrometry can gain analytical interpretability by studying complementarity and synergy between the data obtained by the same technique. To explore its potential in an untargeted metabolomic application, the objective of this work was to obtain organic and aqueous coffee extracts of three coffee Canephora groups produced in Brazil with distinctive aspects: geographical origin and botanical variety. Aqueous and organic extracts of roasted coffee beans were analyzed by direct infusion electrospray ionization mass spectrometry. Due to the large number of samples, the injector of the liquid chromatography system was used to automate the analysis. The column was removed, and a peak tube was added to connect the system directly to the mass spectrometer to inject both polar and nonpolar fractions of the coffee extracts individually. The technique provided characteristic fingerprinting mass spectra that not only allowed for differentiation of geographical origins but also between robusta and conilon botanical varieties. The mass spectra of the organic and water extracts represented two separate data blocks to be analyzed by the ComDim-ICA multi-block data analysis method. While the classical ComDim is based on applying PCA to the iteratively reweighted concatenated matrices, in the ComDim-ICA, the factorization is done using independent components analysis, which promotes specific improvements since it is based on extracting components that are statistically independent of one another. The results highlighted by ComDim-ICA show that both water and organic extracts contributed with important ions to the characterization of the coffee composition. However, the results revealed a high variability of metabolomic composition within each botanical variety (Robusta Amazônico and Conilon Capixaba) and geographical provenance (Rondônia indigenous-1, Rondônia non-indigenous-2 and Espírito Santo-3). Even so, water mass spectra differentiated the botanical variety Conilon from Robusta based on significant ions related to trigonelline, caffeic acid, caffeoylquinic acid, and methylpyridinium; both water and organic mass spectra differentiated Rondônia indigenous from Rondônia non-indigenous and Espírito Santo Conilon based on significant ions related to benzoic acid, pentose, coumaric acid, caffeine in the organic extract and malonic acid, pentose, caffeoylquinic acid, methyl pyridinium, caffeine, and sucrose present in the water extract. With the proposed approach acquiring ion fingerprints of different coffee extracts and their subsequent analysis by ComDim-ICA, new complementary chemical aspects of Brazilian Coffea canephora were put in evidence.

Evaluation of Eureka lemon quality during cold storage integrating hyperspectral imaging, 1H NMR and quality parameters in a multi-block data analysis perspective

To investigate diverse information associated with the same sample, it's crucial to develop a multi-block data analysis. This study utilized quality parameters (L*, a*, b*, hardness, weight loss, vitamin C, titratable acid and soluble solid content), visible near-infrared (Vis-NIR) spectra and proton nuclear magnetic resonance (1H NMR) fingerprints to characterize the quality of Eureka lemons at different storage periods. Common Dimension (ComDim) multi-block analysis was carried out to explore the relationships between these multiple data blocks and assess the contribution of each block. Specifically, the ComDim method calculated the significance, scores, and loadings for each dimension to reveal the synergistic effects among multiple data blocks. Meanwhile, a data-fusion based principal component analysis was performed but was not as efficient as ComDim in exploiting multiple data blocks. The ComDim results indicated that Eureka lemons stored for less than 120 days were linked to firmness, the spectral region of 479–772 nm, and 1H NMR signals (sucrose, fucose, fructose, and α-glucose). Lemons stored for over 150 days correlated with L*, a*, b* values, weight loss, nutritional parameters, the spectral regions of 400–479 nm and 772–1000 nm, and 1H NMR signals (ethanol). This work demonstrates that ComDim furnishes a new approach to comprehensively evaluate the quality of Eureka lemons across storage periods. It also provides guiding significance for transportation, refrigeration and preservation of fruits or other agricultural products.

Implementing multiblock techniques in a full-scale plant scenario: On-line prediction of quality parameters in a continuous process for different acrylonitrile butadiene styrene (ABS) products

BackgroundThe study explores the challenges of handling multiblock data of different natures (process and NIR sensors) for on-line quality prediction in a full-scale plant scenario, namely a plant operating in continuous on an industrial scale and producing different grade Acrylonitrile Butadiene Styrene (ABS) products. This environment is an ideal scenario to evaluate the use of multiblock data analysis methods, which can enhance data interpretation, visualization, and predictive performances. In particular, a novel multiblock extension of Locally Weighted PLS has been proposed by the authors, namely Locally Weighted Multiblock Partial Least Squares (LW-MB-PLS). Response-Oriented Sequential Alternation (ROSA) has also been employed to evaluate the diverse block relevance for the prediction of two quality parameters associated with the polymer. Data are split in blocks both according to sensor type and different plant sections, and different models have been built by incremental addition of data blocks to evaluate if early estimation of product quality is feasible. ResultsROSA method showed promising predictive performance for both quality parameters, highlighting the most influential plant sections through the selection of data blocks. The results suggested that both early and late-stage sensors play crucial roles in predicting product quality. A reasonable estimation of quality parameters before production completion has been achieved. On the other hand, the proposed LW-MB-PLS, while comparable in predictive performances, allowed reducing systematic prediction errors for specific products. SignificanceThis study contributes valuable insights for continuous production processes, aiding plant operators and paving the way for advancements in online quality prediction and control. Furthermore, it is implemented as a locally weighted extension of MB-PLS.

Proximal sensing approach for characterization of calcareous soils using multiblock data analysis

Application of visible-near infrared (VNIR) or mid-infrared (MIR) proximal sensing for prediction of soil attributes has proven successful in many cases and is currently extensively used. However, in some cases, predictions fail, especially for properties that do not have a strong spectral signature. Multiblock data analysis techniques are an extension of partial least squares regression (PLSR) that allows the incorporation of data obtained from different instruments into a single model. The techniques have not been used in soil characterization particularly in chemical and elemental soil analysis. Therefore, the aim of the study was to analyze the use of two multiblock algorithms, sequential orthogonalized partial least squares (SO-PLS) and response-oriented sequential alternation (ROSA), to improve the prediction of several soil attributes from VNIR and MIR spectra and their 2nd order derivatives in calcareous soils from Iran. A total of 130 soil samples were taken from 0 to 30 cm depth in Kamfiroz (Mulla Sadra Dam Watershed), Fars Province, Iran. The analyzed soil properties included soil organic matter (SOM), electrical conductivity (EC), carbonate calcium equivalent (CCE), clay and sand contents, and plant-available (DTPA-extractable) micronutrients such as manganese (Mn), iron (Fe), nickel (Ni), zinc (Zn), and copper (Cu). Multiblock models (SO-PLS and ROSA) were compared with single-block PLSR models obtained from VNIR, MIR and their 2nd order derivatives. The results indicated that SO-PLS improved the prediction performance for all studied properties compared to single-block PLSR models from VNIR, MIR or their 2nd order derivatives, and multiblock ROSA. The highest improvement was achieved for properties where single-block models were less accurate. However, the ROSA models were less accurate than single-block models. The best SO-PLS models were obtained for CCE (R2 = 0.96, RPIQ = 5.95, and NRMSE = 7.24%), followed by SOM (R2 = 0.91, RPIQ = 3.38, and NRMSE = 11.45%), log Fe (R2 = 0.89, RPIQ = 3.16, and NRMSE = 9.98%), Cu (R2 = 0.87, RPIQ = 3.21, and NRMSE = 15.51%), clay content (R2 = 0.85, RPIQ = 3.57, and NRMSE = 10.14%), Ni (R2 = 0.76, RPIQ = 2.60, and NRMSE = 28.25%), EC (R2 = 0.74, RPIQ = 1.63, and NRMSE = 20.94%), sand (R2 = 0.68, RPIQ = 1.82, and NRMSE = 27.58%), and Zn (R2 = 0.63, RPIQ = 1.65, and NRMSE = 18.14%). However, Mn was poorly predicted by all evaluated methods. The results confirmed the potential of the SO-PLS model to improve the prediction of soil properties using a combination of spectral information obtained with different instruments.

Tensor generalized canonical correlation analysis

Regularized Generalized Canonical Correlation Analysis (RGCCA) is a general statistical framework for multiblock data analysis. RGCCA enables deciphering relationships between several sets of variables and subsumes many well-known multivariate analysis methods as special cases. However, RGCCA only deals with vector-valued blocks, disregarding their possible higher-order structures. This paper presents Tensor GCCA (TGCCA), a new method for analyzing higher-order tensors with canonical vectors admitting an orthogonal rank-R CP decomposition. Moreover, two algorithms for TGCCA, based on whether a separable covariance structure is imposed or not, are presented along with convergence guarantees. The efficiency and usefulness of TGCCA are evaluated on simulated and real data and compared favorably to state-of-the-art approaches.

A multi-block data approach to assessing beef quality: ComDim analysis of hyperspectral imaging, 1H NMR, electronic nose and quality parameters data

Several factors affect the quality of beef. In the field of chemometrics, multi-block data analysis methods are useful for examining multiple sources of information from a sample. This study focuses on the application of ComDim, a multi-block data analysis method, to evaluate beef from different parts of hyperspectral spectrum and image texture information, 1H NMR fingerprints, quality parameters and electronic nose. Compared to principal component analysis (PCA) methods based on low-level data fusion, ComDim is more efficient and powerful, because it reveals the relationships between the methods and techniques studied, as well as the variability of beef quality across multiple metrics. The quality and metabolite composition of beef tenderloin and hindquarters were differentiated, with low L* value and high shear tenderloin distinguished from hindquarters with opposite characteristics. The proposed strategy demonstrates that ComDim approach can be used to characterize samples when different techniques describe the same set of samples.

Absorbance-transmission and fluorescence excitation-emission matrix (A-TEEM) with multi-block data analysis and machine learning for accurate intraregional classification of Barossa Shiraz wine

Authentication of wine can be considered at different scales, with classification according to country, province/state, or appellation/wine producing region. An absorbance-transmission and excitation-emission matrix (A-TEEM) technique was applied for the first time to examine intraregional differences, using Shiraz wines (n = 186) produced during three vintages from five subregions of Barossa Valley and from Eden Valley. Absorption spectra and EEM fingerprints were modelled as a multi-block data set for initial exploration with k-means cluster analysis and principal component analysis, and then with machine learning modelling using extreme gradient boosting discriminant analysis (XGBDA). Whereas some clustering was evident with the initial unsupervised approaches, classification with XGBDA afforded an impressive 100% correct class assignment for subregion and vintage year. Extending the utility and novelty of the A-TEEM approach, predictive models for chemical parameters (alcohol, glucose + fructose, pH, titratable acidity, and volatile acidity) were also validated using A-TEEM data with XGB regression.

Texture profile and short-NIR spectral vibrations relationship evaluated through Comdim: The case study for animal and vegetable proteins

Important parameters related to the acceptance of a food can be investigated by the texture profile analysis (TPA), which has as a drawback the time and sample-consuming characteristic. In this sense, the development of a method to integrate the texture parameters into a fast and non-destructive analysis, such as the near-infrared (NIR) spectroscopy, can contribute to better quality control for food products. In this sense, this study aimed to compare commercial samples of proteins from animal and plant sources to find the relation between NIR vibrations and TPA parameters through principal component analysis (PCA) and multiblock data analysis of common dimensions (ComDim). To do so, it was necessary to use a strategy to measure the TPA of powder proteins, which was done by producing a ¼ (m:m) flour: protein dough. The results indicated that animal proteins are related to adhesiveness and springiness and correlated to the NIR region from 1120 to 1180 nm and 1300–1460 nm. The plant proteins are characterized by their high cohesiveness, chewiness, gumminess, hardness, and resilience correlated to the NIR regions between 900 and 1120 nm, 1180–1220 nm, 1250–1295 nm, and 1460–1700 nm.

Rearrangement of incomplete multi-omics datasets combined with ComDim for evaluating replicate cross-platform variability and batch influence

Multi-omics studies can highlight the interrelationships among data across different layers of biological information. However, methods for the unsupervised analysis of multi-block data do not take the individual variability across batches into account and cannot deal with omics datasets when they present different numbers of replicates. We have explored three different data arrangement strategies to tackle these limitations. Several multi-block methods can be used to decipher the common variations across blocks and to determine the contribution of each block to each common component. In this study the ComDim method was used to compare these rearrangement strategies for three multi-omics datasets. We found that arranging the data using the ‘replicate by blocks’ strategy, where each block comprises data from only one replicate independently of its data type, provided the most insightful results. ComDim allowed the evaluation of the variability across the replicate blocks, confirming the existence of batch effects in some of the studies. Moreover, since the contributions of these batch effects were separated from the other contributions, the coordinated biological responses common across the different blocks was characterized for each data type.

Multi-omics analysis to decipher the molecular link between chronic exposure to pollution and human skin dysfunction

Environmental pollution is composed of several factors, namely particulate matter (PM2.5, PM10), ozone and Ultra Violet (UV) rays among others and first and the most exposed tissue to these substances is the skin epidermis. It has been established that several skin disorders such as eczema, acne, lentigines and wrinkles are aggravated by exposure to atmospheric pollution. While pollutants can interact with skin surface, contamination of deep skin by ultrafine particles or Polycyclic aromatic hydrocarbons (PAH) might be explained by their presence in blood and hair cortex. Molecular mechanisms leading to skin dysfunction due to pollution exposure have been poorly explored in humans. In addition to various host skin components, cutaneous microbiome is another target of these environment aggressors and can actively contribute to visible clinical manifestation such as wrinkles and aging. The present study aimed to investigate the association between pollution exposure, skin microbiota, metabolites and skin clinical signs in women from two cities with different pollution levels. Untargeted metabolomics and targeted proteins were analyzed from D-Squame samples from healthy women (n = 67 per city), aged 25–45 years and living for at least 15 years in the Chinese cities of Baoding (used as a model of polluted area) and Dalian (control area with lower level of pollution). Additional samples by swabs were collected from the cheeks from the same population and microbiome was analysed using bacterial 16S rRNA as well as fungal ITS1 amplicon sequencing and metagenomics analysis. The level of exposure to pollution was assessed individually by the analysis of polycyclic aromatic hydrocarbons (PAH) and their metabolites in hair samples collected from each participant. All the participants of the study were assessed for the skin clinical parameters (acne, wrinkles, pigmented spots etc.). Women from the two cities (polluted and less polluted) showed distinct metabolic profiles and alterations in skin microbiome. Profiling data from 350 identified metabolites, 143 microbes and 39 PAH served to characterize biochemical events that correlate with pollution exposure. Finally, using multiblock data analysis methods, we obtained a potential molecular map consisting of multi-omics signatures that correlated with the presence of skin pigmentation dysfunction in individuals living in a polluted environment. Overall, these signatures point towards macromolecular alterations by pollution that could manifest as clinical sign of early skin pigmentation and/or other imperfections.

A general strategy for setting up supervised methods of multiblock data analysis

A general strategy for setting up supervised methods of multiblock data analysis is outlined. This yields a unified framework where some already known methods are retreived and novel extensions are introduced. All the methods of analysis are based on the determination of latent variables associated with the various blocks of variables. They are derived from clear optimization criterion whose aim is to maximize either the sum of the covariances or the sum of squared covariances between the latent variable associated with the response variables and the block latent variables associated with the various explanatory datasets. New indices are proposed to help better interpreting the outcomes of the analyses. The methods are illustrated and compared based on simulated and real datasets.

Spectrofluorometric analysis combined with machine learning for geographical and varietal authentication, and prediction of phenolic compound concentrations in red wine

Fluorescence spectroscopy is rapid, straightforward, selective, and sensitive, and can provide the molecular fingerprint of a sample based on the presence of various fluorophores. In conjunction with chemometrics, fluorescence techniques have been applied to the analysis and classification of an array of products of agricultural origin. Recognising that fluorescence spectroscopy offered a promising method for wine authentication, this study investigated the unique use of an absorbance-transmission and fluorescence excitation emission matrix (A-TEEM) technique for classification of red wines with respect to variety and geographical origin. Multi-block data analysis of A-TEEM data with extreme gradient boosting discriminant analysis yielded an unrivalled 100% and 99.7% correct class assignment for variety and region of origin, respectively. Prediction of phenolic compound concentrations with A-TEEM based on multivariate calibration models using HPLC reference data was also highly effective, and overall, the A-TEEM technique was shown to be a powerful tool for wine classification and analysis.

Deep multiblock predictive modelling using parallel input convolutional neural networks

In the domain of chemometrics, multiblock data analysis is widely performed for exploring or fusing data from multiple sources. Commonly used methods for multiblock predictive analysis are the extensions of latent space modelling approaches. However, recently, deep learning (DL) approaches such as convolutional neural networks (CNNs) have outperformed the single block traditional latent space modelling chemometric approaches such as partial least-square (PLS) regression. The CNNs based DL modelling can also be performed to simultaneously deal with the multiblock data but was never explored until this study. Hence, this study for the first time presents the concept of parallel input CNNs based DL modelling for multiblock predictive chemometric analysis. The parallel input CNNs based DL modelling utilizes individual convolutional layers for each data block to extract key features that are later combined and passed to a regression module composed of fully connected layers. The method was tested on a real visible and near-infrared (Vis-NIR) large data set related to dry matter prediction in mango fruit. To have the multiblock data, the visible (Vis) and near-infrared (NIR) parts were treated as two separate blocks. The performance of the parallel input CNN was compared with the traditional single block CNNs based DL modelling, as well as with a commonly used multiblock chemometric approach called sequentially orthogonalized partial least-square (SO-PLS) regression. The results showed that the proposed parallel input CNNs based deep multiblock analysis outperformed the single block CNNs based DL modelling and the SO-PLS regression analysis. The root means squared errors of prediction obtained with deep multiblock analysis was 0.818%, relatively lower by 4 and 20% than single block CNNs and SO-PLS regression, respectively. Furthermore, the deep multiblock approach attained ∼3% lower RMSE compared to the best known on the mango data set used for this study. The deep multiblock analysis approach based on parallel input CNNs could be considered as a useful tool for fusing data from multiple sources.

Ecometabolic mixture design-fingerprints from exploratory multi-block data analysis in Coffea arabica beans from climate changes: Elevated carbon dioxide and reduced soil water availability

Ecometabolic mixture design-fingerprinting in coffee cultivated under climate change was chemically explored using ComDim. Multi-blocks were formed using UV, NIRS, 1H NMR, SWV, and FT-IR data. ComDim investigated all these different fingerprints according to the extractor solvent and in virtue of atmospheric CO2 increase. Ethanol and ethanol-dichloromethane showed the best separations due to CO2 environment. 1H NMR loading indicate increases of fatty acids, caffeine, trigonelline, and glucose in beans under current CO2 levels, whereas quinic acid/chlorogenic acids, malic acid, and kahweol/cafestol increased in beans under elevated CO2 conditions. SWV indicated quercetin and chlorogenic acid as important compounds in coffee beans cultivated under current and elevated CO2, respectively. Based on the ethanol and ethanol-dichloromethane fingerprints, k-NN correctly classified the beans cultivated under different carbon dioxide environments and water availabilities, confirming the existence of metabolic changes due to climate changes. SWV proved to be promising compared with widely used spectrometric methods.

Integrated 1H NMR fingerprint with NIR spectroscopy, sensory properties, and quality parameters in a multi-block data analysis using ComDim to evaluate coffee blends

Coffee quality is determined by several factors and, in the chemometric domain, the multi-block data analysis methods are valuable to study multiple information describing the same samples. In this industrial study, the Common Dimension (ComDim) multi-block method was applied to evaluate metabolite fingerprints, near-infrared spectra, sensory properties, and quality parameters of coffee blends of different cup and roasting profiles and to search relationships between these multiple data blocks. Data fusion-based Principal Component Analysis was not effective in exploiting multiple data blocks like ComDim. However, when a multi-block was applied to explore the data sets, it was possible to demonstrate relationships between the methods and techniques investigated and the importance of each block or criterion involved in the industrial quality control of coffee. Coffee blends were distinguished based on their qualities and metabolite composition. Blends with high cup quality and lower roasting degrees were generally differentiated from those with opposite characteristics.

CLUSTATIS: Cluster analysis of blocks of variables

The STATIS method is one of many strategies of analysis devoted to the unsupervised analysis of multiblock data. A new optimization criterion to define this method of analysis is introduced and an extension to the cluster analysis of several blocks of variables is discussed. This consists in a hierarchical cluster analysis and a partitioning algorithm akin to the K-means algorithm. Moreover, in order to improve the cluster analysis outcomes, an additional cluster called noise cluster which contains atypical blocks of variables is introduced. The general strategy of analysis is illustrated by means of two cases studies.

Multi-block data analysis for online monitoring of anaerobic co-digestion process

Anaerobic digestion is a chemical process whose purpose is to maximize biogas production whilst concomitantly treating organic waste mostly through co-digestion due to the variety of substrates. To avoid failures, the process requires the monitoring of several parameters and / or inhibitors. The existing strategies and methods used in the process monitoring still lack sensitivity and robustness, when taken individually. The current study investigated the use of sequential and orthogonalized partial least squares (SO-PLS) regression to relate these parameters to several blocks of data coming for near infrared spectroscopy, chemical routine analysis and kinetics of biogas production. The models produced were able to extract relevant information from each block’s data and discard redundancies. Moreover, to meet biogas plant operators’ requirements, variable selection was performed on the infrared blocks using a recent method: SO-CovSel. SO-CovSel is a method resulting from coupling SO-PLS and Covariance Selection (CovSel) method. The method has been demonstrated to be suitable for multi-response calibration purposes with infrared calibration. It has provided good predictions and an interesting interpretation of wavelengths involved in the monitoring of the relevant parameters of stability in anaerobic co-digestion.

Multiway generalized canonical correlation analysis.

Regularized generalized canonical correlation analysis (RGCCA) is a general multiblock data analysis framework that encompasses several important multivariate analysis methods such as principal component analysis, partial least squares regression, and several versions of generalized canonical correlation analysis. In this article, we extend RGCCA to the case where at least one block has a tensor structure. This method is called multiway generalized canonical correlation analysis (MGCCA). Convergence properties of the MGCCA algorithm are studied, and computation of higher-level components are discussed. The usefulness of MGCCA is shown on simulation and on the analysis of a cognitive study in human infants using electroencephalography (EEG).

Unsupervised multiblock data analysis: A unified approach and extensions

For the analysis of multiblock data, a unified approach of several strategies such as Generalized Canonical Correlation Analysis (GCCA), Multiblock Principal Components Analysis (MB-PCA), Hierarchical Principal Components Analysis (H-PCA) and ComDim is outlined. These methods are based on the determination of global and block components. The unified approach postulates, on the one hand, two link functions that relate the block components to their associated global components and, on the other hand, two summing up expressions to compute the global components from their associated block components. Not only several well-known methods are retrieved but we also introduce a variant of GCCA. More generally, we hint to other possibilities of extensions thus emphasizing the fact that the unified approach, besides being simple, is versatile. We also show how this approach of analysis although basically unsupervised could be adapted to yield a supervised method to be used for a prediction purpose. Illustrations on the basis of simulated and real case studies are discussed.

Understanding cold maceration in red winemaking: A batch processing and multi-block data analysis approach

The phenolic extraction optimization during the winemaking process is of crucial importance. The main aim of cold maceration is to enhance phenolic extraction prior to the fermentation process. The pre-fermentative diffusion of phenolic substances takes place at low temperatures and in the absence of ethanol. Variable results are often reported for cold maceration applications, which are ascribed to grape composition at harvest. The influence of different cold maceration treatments was evaluated following batch process modelling combined with hierarchal multiblock analysis. The winemaking process was divided in three blocks, namely: cold maceration, alcoholic fermentation and malolactic fermentation. Batch evolution and batch level models were assessed using PLS quantified phenolic composition. A combined effect of high SO2 levels and enzyme additions seemed to have a significant positive influence on wine phenolic levels, regardless of grape characteristics. Batch modelling provides a better process understanding through the identification of the relevant process variables (phenolics). Moreover, hierarchal multiblock zooming capability was used to understand the effect of cold maceration during the different parts of the process as well as providing an overall effect of cold maceration applications.