Multiblock variable influence on orthogonal projections (MB-VIOP) for enhanced interpretation of total, global, local and unique variations in OnPLS models.

BACKGROUND: For multivariate data analysis involving only two input matrices (e.g., X and Y), the previously published methods for variable influence on projection (e.g., VIPOPLS or VIPO2PLS) are widely used for variable selection purposes, including (i) variable importance assessment, (ii) dimensionality reduction of big data and (iii) interpretation enhancement of PLS, OPLS and O2PLS models. For multiblock analysis, the OnPLS models find relationships among multiple data matrices (more than two blocks) by calculating latent variables; however, a method for improving the interpretation of these latent variables (model components) by assessing the importance of the input variables was not available up to now. RESULTS: A method for variable selection in multiblock analysis, called multiblock variable influence on orthogonal projections (MB-VIOP) is explained in this paper. MB-VIOP is a model based variable selection method that uses the data matrices, the scores and the normalized loadings of an OnPLS model in order to sort the input variables of more than two data matrices according to their importance for both simplification and interpretation of the total multiblock model, and also of the unique, local and global model components separately. MB-VIOP has been tested using three datasets: a synthetic four-block dataset, a real three-block omics dataset related to plant sciences, and a real six-block dataset related to the food industry. CONCLUSIONS: We provide evidence for the usefulness and reliability of MB-VIOP by means of three examples (one synthetic and two real-world cases). MB-VIOP assesses in a trustable and efficient way the importance of both isolated and ranges of variables in any type of data. MB-VIOP connects the input variables of different data matrices according to their relevance for the interpretation of each latent variable, yielding enhanced interpretability for each OnPLS model component. Besides, MB-VIOP can deal with strong overlapping of types of variation, as well as with many data blocks with very different dimensionality. The ability of MB-VIOP for generating dimensionality reduced models with high interpretability makes this method ideal for big data mining, multi-omics data integration and any study that requires exploration and interpretation of large streams of data.

Hyperspectral near infrared image calibration and regression.

Reference materials are used in diffuse reflectance imaging for transforming the digitized camera signal into reflectance and absorbance units for subsequent interpretation. Traditional white and dark reference signals are generally used for calculating reflectance or absorbance, but these can be supplemented with additional reflectance targets to improve the accuracy of reflectance transformations. In this work we provide an overview of hyperspectral image regression and assess the effects of reflectance calibration on image interpretation using partial least squares regression. Linear and quadratic reflectance transformations based on additional reflectance targets decrease average measurement errors and make it easier to estimate model pseudorank during image regression. The lowest measurement and prediction errors were obtained with the column and wavelength specific quadratic transformations which retained the spatial information provided by the line-scanning instrument and reduced errors in the predicted concentration maps.

Joint and Unique Multiblock Analysis for Integration and Calibration Transfer of NIR Instruments.

In the present paper, we introduce an end-to-end workflow called joint and unique multiblock analysis (JUMBA), which allows multiple sources of data to be analyzed simultaneously to better understand how they complement each other. In near-infrared (NIR) spectroscopy, calibration models between NIR spectra and responses are used to replace wet-chemistry methods, and the models tend to be instrument-specific. Calibration-transfer techniques are used for standardization of NIR-instrumentation, enabling the use of one model on several instruments. The current paper investigates both the similarities and differences among a variety of NIR instruments using JUMBA. We demonstrate JUMBA on both a previously unpublished data set in which five NIR instruments measured mushroom substrate and a publicly available data set measured on corn samples. We found that NIR spectra from different instrumentation largely shared the same underlying structures, an insight we took advantage of to perform calibration transfer. The proposed JUMBA transfer displayed excellent calibration-transfer performance across the two analyzed data sets and outperformed existing methods in terms of both prediction accuracy and stability. When applied to a multi-instrument environment, JUMBA transfer can integrate all instruments in the same model and will ensure higher consistency among them compared with existing calibration-transfer methods.

OnPLS-Based Multi-Block Data Integration: A Multivariate Approach to Interrogating Biological Interactions in Asthma.

Integration of multiomics data remains a key challenge in fulfilling the potential of comprehensive systems biology. Multiple-block orthogonal projections to latent structures (OnPLS) is a projection method that simultaneously models multiple data matrices, reducing feature space without relying on a priori biological knowledge. In order to improve the interpretability of OnPLS models, the associated multi-block variable influence on orthogonal projections (MB-VIOP) method is used to identify variables with the highest contribution to the model. This study combined OnPLS and MB-VIOP with interactive visualization methods to interrogate an exemplar multiomics study, using a subset of 22 individuals from an asthma cohort. Joint data structure in six data blocks was assessed: transcriptomics; metabolomics; targeted assays for sphingolipids, oxylipins, and fatty acids; and a clinical block including lung function, immune cell differentials, and cytokines. The model identified seven components, two of which had contributions from all blocks (globally joint structure) and five that had contributions from two to five blocks (locally joint structure). Components 1 and 2 were the most informative, identifying differences between healthy controls and asthmatics and a disease-sex interaction, respectively. The interactions between features selected by MB-VIOP were visualized using chord plots, yielding putative novel insights into asthma disease pathogenesis, the effects of asthma treatment, and biological roles of uncharacterized genes. For example, the gene ATP6 V1G1, which has been implicated in osteoporosis, correlated with metabolites that are dysregulated by inhaled corticoid steroids (ICS), providing insight into the mechanisms underlying bone density loss in asthma patients taking ICS. These results show the potential for OnPLS, combined with MB-VIOP variable selection and interaction visualization techniques, to generate hypotheses from multiomics studies and inform biology.

Hyperspectral near infrared imaging quantifies the heterogeneity of carbon materials.

For many applications heterogeneity is a direct indicator of material quality. Reliable determination of chemical heterogeneity is however not a trivial task. Spectral imaging can be used for determining the spatial distribution of an analyte in a sample, thus transforming each pixel of an image into a sampling cell. With a large amount of image pixels, the results can be evaluated using large population statistics. This enables robust determination of heterogeneity in biological samples. We show that hyperspectral imaging in the near infrared (NIR) region can be used to reliably determine the heterogeneity of renewable carbon materials, which are promising replacements for current fossil alternatives in energy and environmental applications. This method allows quantifying the variation in renewable carbon and other biological materials that absorb in the NIR region. Reliable determination of heterogeneity is also a valuable tool for a wide range of other chemical imaging applications.

Determination of methotrexate in spiked human blood serum using multi-frequency electrochemical immittance spectroscopy and multivariate data analysis.

This article describes an attempt to develop a sensor based on multi-frequency immittance spectroscopy for the determination of methotrexate (MTX) in blood serum using gold electrodes modified with antibodies. The attachment of antibodies was monitored with electrochemical immittance spectroscopy (EIS) and X-ray photoelectron spectroscopy (XPS). The EIS measurements of MTX resulted in a data matrix of size 39 × 55. The data were analysed using multivariate data analysis and showed a concentration dependence and time dependence that could be separated. This allowed the calculation of a multivariate calibration model. The model showed good linear behavior on a logarithmic scale offering a detection limit of 5 × 10-12 mol L-1.

Hyperspectral Imaging to Determine the Properties and Homogeneity of Renewable Carbon Materials.

Hyperspectral imaging within the near infrared (NIR) region offers a fast and reliable way for determining the properties of renewable carbon materials. The chemical information provided by a spectrum combined with the spatial information of an image allows mathematical operations that can be performed in both the spectral and spatial domains. Here, we show that hyperspectral NIR imaging can be successfully used to determine the properties of hydrothermally prepared carbon on the material and pixel levels. Materials produced from different feedstocks or prepared under different temperatures can also be distinguished, and their homogeneity can be evaluated. As hyperspectral imaging within the NIR region is non-destructive and requires very little sample preparation, it can be used for controlling the quality of renewable carbon materials destined for a wide range of different applications.

NIR hyperspectral imaging and multivariate image analysis to characterize spent mushroom substrate: a preliminary study.

Commercial mushroom growth on substrate material produces a heterogeneous waste that can be used for bioenergy purposes. Hyperspectral imaging in the near-infrared (NHI) was used to experimentally study a number of spent mushroom substrate (SMS) packed samples under different conditions (wet vs. dry, open vs. plastic covering, and round or cuboid) and to explore the possibilities of direct characterization of the fresh substrate within a plastic bag. Principal components analysis (PCA) was used to remove the background of images, explore the important studied factors, and identify SMS and mycelia (Myc) based on the pixel clusters within the score plot. Overview PCA modeling indicated high moisture content caused the most significant effects on spectra followed by the uneven distribution of Myc and the plastic cover. There were well-separated pixel clusters for SMS and Myc under different conditions: dry, wet, or wet and plastic covering. The loading peaks of the related component and the second derivative of the mean spectra of pixel clusters of SMS and Myc indicated that there are chemical differences between SMS and Myc. Partial least squares discriminant analysis (PLS-DA) models were calculated and classification of SMS and Myc was successful, whether the materials were dry or wet. Peak shifts because of high moisture content and unexpected peaks from the plastic covering were found. Although the best results were obtained for dried cylinders, it was shown that almost equally good results could be obtained for the wet material and for the wet material covered by plastic. Furthermore, PLS-DA prediction showed that a side face hyperspectral image could represent the information for the entire SMS cylinder when Myc was removed. Thus, the combination of NHI and multivariate image analysis has great potential to develop calibration models to directly predict the contents of water, carbohydrates, lignin, and protein in wet and plastic-covered SMS cylinders.

Detection of methotrexate in a flow system using electrochemical impedance spectroscopy and multivariate data analysis.

Methotrexate (MTX), a common pharmaceutical drug in cancer therapy and treatment of rheumatic diseases, is known to cause severe adverse side effects at high dose. As the side effect may be life threatening, there is an urgent need for a continuous, bed-side monitoring of the nominal MTX serum level in a patient while the chemical is being administered. This article describes a detection of MTX using a flow system that consists two modified gold electrodes. Interaction of MTX with the antibodies fixed on the electrode surface is detected by electrochemical impedance spectroscopy and evaluated using singular value decomposition (SVD). The key finding of this work is that the change in the electrode capacitance is found to be quantitative with respect to the concentration of MTX. Moreover a calibration curve constructed using the principal component regression method has a linear range of six orders of magnitude and a detection limit of 1.65 × 10(-10) M.

Noise Removal with Maintained Spatial Resolution in Raman Images of Cells Exposed to Submicron Polystyrene Particles.

The biodistribution of 300 nm polystyrene particles in A549 lung epithelial cells has been studied with confocal Raman spectroscopy. This is a label-free method in which particles and cells can be imaged without using dyes or fluorescent labels. The main drawback with Raman imaging is the comparatively low spatial resolution, which is aggravated in heterogeneous systems such as biological samples, which in addition often require long measurement times because of their weak Raman signal. Long measurement times may however induce laser-induced damage. In this study we use a super-resolution algorithm with Tikhonov regularization, intended to improve the image quality without demanding an increased number of collected pixels. Images of cells exposed to polystyrene particles have been acquired with two different step lengths, i.e., the distance between pixels, and compared to each other and to corresponding images treated with the super-resolution algorithm. It is shown that the resolution after application of super-resolution algorithms is not significantly improved compared to the theoretical limit for optical microscopy. However, to reduce noise and artefacts in the hyperspectral Raman images while maintaining the spatial resolution, we show that it is advantageous to use short mapping step lengths and super-resolution algorithms with appropriate regularization. The proposed methodology should be generally applicable for Raman imaging of biological samples and other photo-sensitive samples.

Multivariate modelling on biomass properties of cassava stems based on an experimental design.

Based on a factorial experimental design (three locations × three cultivars × five harvest times × four replicates) conducted with the objective of investigating variations in fuel characteristics of cassava stem, a multivariate data matrix was formed which was composed of 180 samples and 10 biomass properties for each sample. The properties included as responses were two different calorific values and ash, N, S, Cl, P, K, Ca, and Mg content. Overall principal component analysis (PCA) revealed a strong clustering for the growing locations, but overlapping clusters for the cultivar types and almost no useful information about harvest times. PCA using a partitioned data set (60 × 10) for each location revealed a clustering of cultivars. This was confirmed by soft independent modelling of class analogy (SIMCA) and partial-least-squares discriminant analysis (PLS-DA), and indicated that the locations gave meaningful information about the differences in cultivar, whereas harvest time was not found to be a differentiating factor. Using the PLS technique, it was revealed that ash, K, and Cl content were the most important responses for PLS-DA models. Furthermore, using PLS regression of fuel and soil variables it was also revealed that fuel K and ash content were correlated with the soil P, Si, Ca, and K content, whereas fuel Cl content was correlated with soil pH and content of organic carbon, N, S, and Mg in the soil. Thus, the multivariate modelling used in this study reveals the possibility of performing rigorous analysis of a complex data set when an analysis of variance may not be successful.

Application of Rapid Visco Analyser (RVA) viscograms and chemometrics for maize hardness characterisation.

It has been established in this study that the Rapid Visco Analyser (RVA) can describe maize hardness, irrespective of the RVA profile, when used in association with appropriate multivariate data analysis techniques. Therefore, the RVA can complement or replace current and/or conventional methods as a hardness descriptor. Hardness modelling based on RVA viscograms was carried out using seven conventional hardness methods (hectoliter mass (HLM), hundred kernel mass (HKM), particle size index (PSI), percentage vitreous endosperm (%VE), protein content, percentage chop (%chop) and near infrared (NIR) spectroscopy) as references and three different RVA profiles (hard, soft and standard) as predictors. An approach using locally weighted partial least squares (LW-PLS) was followed to build the regression models. The resulted prediction errors (root mean square error of cross-validation (RMSECV) and root mean square error of prediction (RMSEP)) for the quantification of hardness values were always lower or in the same order of the laboratory error of the reference method.

Assessment of PCDD/F source contributions in Baltic Sea sediment core records.

Spatial and temporal trends of sources of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) in the Baltic Sea were evaluated by positive matrix factorization (PMF) and principal component analysis (PCA). Sediment cores were sampled at eight coastal, one coastal reference, and six offshore sites covering the northern to the southern Baltic Sea. The cores, which covered the period 1919-2010, were sliced into 2-3 cm disks among which 8-11 disks per core (in total 141 disks) were analyzed for all tetra- through octa-CDD/Fs. Identification and apportionment of PCDD/F sources was carried out using PMF. Five stable model PCDD/F congener patterns were identified, which could be associated with six historically important source types: (i) atmospheric background deposition (ABD), (ii) use and production of penta-chlorophenol (PCP), (iii) use and production of tetra-chlorophenol (TeCP), (iv) high temperature processes (Thermal), (v) hexa-CDD-related sources (HxCDD), and (vi) chlorine-related sources (Chl), all of which were still represented in the surface layers. Overall, the last four decades of the period 1920-2010 have had a substantial influence on the Baltic Sea PCDD/F pollution, with 88 ± 7% of the total amount accumulated during this time. The 1990s was the peak decade for all source types except TeCP, which peaked in the 1980s in the northern Baltic Sea and has still not peaked in the southern part. The combined impact of atmospheric-related emissions (ABD and Thermal) was dominant in the open sea system throughout the study period (1919-2010) and showed a decreasing south to north trend (always >80% in the south and >50% in the north). Accordingly, to further reduce levels of PCDD/Fs in the open Baltic Sea ecosystem, future actions should focus on reducing atmospheric emissions.

Multivariate data analysis to characterize gas chromatography columns for dioxin analysis.

Principal component analysis (PCA) was applied for evaluating the selectivity of 22 GC columns for which complete retention data were available for the 136 tetra- to octa-chlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDFs). Because the hepta- and octa-homologues are easy to separate the PCA was focused on the 128 tetra- to hexa-CDD/Fs. The analysis showed that 21 of the 22 GC columns could be subdivided into four groups with different selectivity. Group I consists of columns with non-polar thermally stable phases (Restek 5Sil MS and Dioxin 2, SGE BPX-DXN, Supelco Equity-5, and Agilent DB-1, DB-5, DB-5ms, VF-5ms, VF-Xms and DB-XLB). Group II includes ionic liquid columns (Supelco SLB-IL61, SLB-IL111 and SLB-IL76) with very high polarity. Group III includes columns with high-percentage phenyl and cyanopropyl phases (Agilent DB-17 and DB-225, Quadrex CPS-1, Supelco SP-2331, and Agilent CP-Sil 88), and Group IV columns with shape selectivity (Dionex SB-Smectic and Restek LC-50, Supelco ßDEXcst, Agilent VF-Xms and DB-XLB). Thus, two columns appeared in both Group I and IV (Agilent VF-Xms and DB-XLB). The selectivity of the other column, Agilent DB-210, differs from those of these four groups. Partial least squares (PLS) regression was used to correlate the retention times of the tetra- to hexa-CDD/Fs on the 22 stationary phases with a set of physicochemical and structural descriptors to identify parameters that significantly influence the solute-stationary phase interactions. The most influential physicochemical parameters for the interaction were associated with molecular size (as reflects in the total energy, electron energy, core-core repulsion and standard entropy), solubility (aqueous solubility and n-octanol/water partition coefficient), charge distribution (molecular polarizability and dipolar moment), and reactivity (relative Gibbs free energy); and the most influential structural descriptors were related to these parameters, in particular, size and dipolar moment. Finally, the PCA and PLS analyses were complemented with linear regression analysis to identify the most orthogonal column combinations, which could be used in comprehensive two-dimensional gas chromatography (GC×GC) to enhance PCDD/F separation and congener profiling.

Large uptake of titania and iron oxide nanoparticles in the nucleus of lung epithelial cells as measured by Raman imaging and multivariate classification.

It is a challenging task to characterize the biodistribution of nanoparticles in cells and tissue on a subcellular level. Conventional methods to study the interaction of nanoparticles with living cells rely on labeling techniques that either selectively stain the particles or selectively tag them with tracer molecules. In this work, Raman imaging, a label-free technique that requires no extensive sample preparation, was combined with multivariate classification to quantify the spatial distribution of oxide nanoparticles inside living lung epithelial cells (A549). Cells were exposed to TiO2 (titania) and/or α-FeO(OH) (goethite) nanoparticles at various incubation times (4 or 48 h). Using multivariate classification of hyperspectral Raman data with partial least-squares discriminant analysis, we show that a surprisingly large fraction of spectra, classified as belonging to the cell nucleus, show Raman bands associated with nanoparticles. Up to 40% of spectra from the cell nucleus show Raman bands associated with nanoparticles. Complementary transmission electron microscopy data for thin cell sections qualitatively support the conclusions.

Non-invasive identification of melanoma with near-infrared and skin impedance spectroscopy.

BACKGROUND/PURPOSE: An early diagnosis of cutaneous malignant melanoma is of high importance for good prognosis. An objective, non-invasive instrument could improve the diagnostic accuracy of melanoma and decrease unnecessary biopsies. The aim of this study was to investigate the use of Near-infrared and skin impedance spectroscopy in combination as a tool to distinguish between malignant and benign skin tumours. METHODS: Near-infrared and skin impedance spectra were collected in vivo on 50 naevi or suspect melanomas prior to excision. Received data were analysed using multivariate techniques and the results were compared to histopathology analyses of the tumours. A total of 12 cutaneous malignant melanomas, 19 dysplastic naevi and 19 benign naevi were included in the study. RESULTS: The observed sensitivity and specificity of the proposed method were 83% and 95%, respectively, for malignant melanoma. CONCLUSION: The results indicate that the combination of near-infrared and skin impedance spectroscopy is a promising tool for non-invasive diagnosis of suspect cutaneous malignant melanomas.

Growth characteristics of three Fusarium species evaluated by near-infrared hyperspectral imaging and multivariate image analysis.

Colony growth of three Fusarium spp. on potato dextrose agar was followed by collecting near-infrared (NIR) hyperspectral images of the colonies at regular intervals after inoculation up to 55 h. After principal component analysis (PCA), two clusters were apparent in the score plot along principal component 1. Using the brushing technique, these clusters were divided into four groups of pixels with similar score values. These could be visualised as growth zones within the colonies in the corresponding score image. Three spectral bands, i.e. 1,166, 1,380 and 1,918 nm, were prominent in the multiplicative scatter corrected and Savitzky-Golay second derivative spectra. These indicated chemical changes, associated with carbohydrates (1,166 and 1,380 nm) and protein (1,918 nm), that occurred as the mycelium grew and matured. The protein band was more prominent in the mature fungal material while the carbohydrate band was less pronounced. The younger material and the agar were characterised by the carbohydrate spectral band. Integrating whole mycelium colonies as the sum of pixels over time made it possible to construct curves that resembled growth curves; this included the lag phase, active growth phase, deceleration phase and phase of constant growth. Growth profiles constructed from individual growth zones indicated more detailed growth characteristics. The use of NIR hyperspectral imaging and multivariate image analysis (MIA) allowed one to visualise radial growth rings in the PCA score images. This would not have been possible with bulk spectroscopy. Interpreting spectral data enabled better understanding of microbial growth characteristics on agar medium. NIR hyperspectral imaging combined with MIA is a powerful tool for the evaluation of growth characteristics of fungi.

Near-infrared (NIR) hyperspectral imaging and multivariate image analysis to study growth characteristics and differences between species and strains of members of the genus Fusarium.

Near-infrared (NIR) hyperspectral imaging was used to study three strains of each of three Fusarium spp. (Fusarium subglutinans, Fusarium proliferatum and Fusarium verticillioides) inoculated on potato dextrose agar in Petri dishes after either 72 or 96 h of incubation. Multivariate image analysis was used for cleaning the images and for making principal component analysis (PCA) score plots and score images and local partial least squares discriminant analysis (PLS-DA) models. The score images, including all strains, showed how different the strains were from each other. Using classification gradients, it was possible to show the change in mycelium growth over time. Loading line plots for principal component (PC) 1 and PC2 explained variation between the different Fusarium spp. as scattering and chemical differences (protein production), respectively. PLS-DA prediction results (including only the most important strain of each species) showed that it was possible to discriminate between species with F. verticillioides the least correctly predicted (between 16 and 47 % pixels correctly predicted). For F. subglutinans, 78-100 % pixels were correctly predicted depending on the training and test sets used. Similarly, the percentage correctly predicted values of F. proliferatum were 60-80 %. Visualisation of the mycelium radial growth in the PCA score images was made possible due to the use of NIR hyperspectral imaging. This is not possible with bulk spectroscopy in the visible or NIR regions.

Influence of grain topography on near infrared hyperspectral images.

Near infrared hyperspectral imaging (NIR-HSI) allows spatially resolved spectral information to be collected without sample destruction. Although NIR-HSI is suitable for a broad range of samples, sizes and shapes, topography of a sample affects the quality of near infrared (NIR) measurements. Single whole kernels of three cereals (barley, wheat and sorghum), with varying topographic complexity, were examined using NIR-HSI. The influence of topography (sample shape and texture) on spectral variation was examined using principal component analysis (PCA) and classification gradients. The greatest source of variation for all three grain types, despite spectral preprocessing with standard normal variate (SNV) transformation, was kernel curvature. Only 1.29% (PC5), 0.59% (PC6) and 1.36% (PC5) of the spectral variation within the respective barley, wheat and sorghum image datasets was explained within the principal component (PC) associated with the chemical change of interest (loss of kernel viability). The prior PCs explained an accumulated total of 91.18%, 89.43% and 84.39% of spectral variance, and all were influenced by kernel topography. Variation in sample shape and texture relative to the chemical change of interest is an important consideration prior to the analysis of NIR-HSI data for non-flat objects.

NIR and skin impedance spectroscopic measurements for studying the effect of coffee and alcohol on skin, and dysplastic naevi.

BACKGROUND/PURPOSE: Near infrared and impedance spectroscopy can be used for clinical skin measurements and need to be evaluated for possible confounding factors; (i) are skin conditions of the patient and the subsequent skin measurements influenced by alcohol and/or coffee consumption and (ii) are measurements of dysplastic naevi (DN) reproducible over time and significantly different compared to reference skin. METHODS: Near infrared and skin impedance spectroscopic data were analysed multivariately. In the first study, the skin characteristics of 15 healthy individuals were examined related to body location, gender, individual differences, and consumption of coffee or alcohol. The second study included five patients diagnosed with dysplastic naevi syndrome. Measurements were taken on DN and reference skin over time. RESULTS: In the first study, body location and gender had a major influence on measurement scores. Inter-individual skin characteristics and coffee or alcohol effects on skin characteristics were of minor importance. In the second study, it was shown that DN can be differentiated from reference skin and the measurements are stable over time. CONCLUSIONS: Moderate consumption of alcohol and coffee did not influence the results of the measurements. It is possible to follow, stable or changed, characteristics of DN over time.