Multiblock Partial Least Squares Research Articles

Identifying the active ingredients of carbonized Typhae Pollen by spectrum-effect relationship combined with MBPLS, PLS, and SVM algorithms

Typhae Pollen (TP) and its carbonized product (carbonized Typhae Pollen, CTP), as cut-and-dried herbal drugs, have been widely used in the form of slices in clinical settings. However, the two drugs exhibit a great difference in terms of their clinical efficacy, for TP boasts an effect of removing blood stasis and promoting blood circulation, while CTP typically presents a hemostatic function. Since the active ingredients of CTP, so far, still remain unclear, this study aimed at identifying the active ingredients of CTP by spectrum-effect relationship approach coupled with multi-block partial least squares (MBPLS), partial least squares (PLS), and support vector machine (SVM) algorithms. In this study, the chemical profiles of a series of CTP samples which were stir-fried for different duration (denoted as CTP0∼CTP9) were firstly characterized by UHPLC-QE-Orbitrap MS. Then the hemostatic effect of the CTP samples was evaluated from the perspective of multiple parameters—APTT, PT, TT, FIB, TXB2, 6-keto-PGF1α, PAI-1 and t-PA—using established rat models with functional uterine bleeding. Subsequently, MBPLS, PLS and SVM were combined to perform spectrum-effect relationship analysis to identify the active ingredients of CTP, followed by an in vitro hemostatic bioactivity test for verification. As a result, a total of 77 chemical ingredients were preliminarily identified from the CTP samples, and the variations occurred in these ingredients were also analyzed during the carbonizing process. The study revealed that all the CTP samples, to a varying degree, showed a hemostatic effect, among which CTP6 and CTP7 were superior to the others in terms of the hemostatic effect. The block importance in the projection (BIP) indexes of MBPLS model indicated that flavonoids and organic acids made more contributions to the hemostatic effect of CTP in comparison to other ingredients. Consequently, 9 bioactive ingredients, including quercetin-3-O-glucoside, kaempferol-3-O-rutinoside, quercetin, kaempferol, isorhamnetin, 2-methylenebutanedioic acid, pentanedioic acid, benzoic acid and 3-hydroxybenzoic acid, were further identified as the potential active ingredients based on PLS and SVM models as well as the in vitro verification. This study successfully revealed the bioactive ingredients of CTP associated with its hemostatic effect, and also provided a scientific basis for further understanding the mechanism of TP processing. In addition, it proposed a novel path to identify the active ingredients for Chinese herbal medicines.

Spectral sensor fusion for prediction of Li and Zr in rocks: Neural network and PLS methods

Integrating information from multiple sensors, known as sensor fusion, is particularly challenging for small datasets where selecting from the plethora of available methods poses a significant challenge in chemometric analysis. This study compares several sensor fusion methods (spanning data-level, feature-level, and decision-level fusion) based on partial least squares (PLS) and convolutional neural network (CNN) models. This study is the first to compare simple sensor fusion methods to the latest multiblock PLS models and parallel-input CNNs, to the best of our knowledge. We demonstrate sensor fusion using a small dataset of 177 rock samples on two prediction tasks: predicting lithium (Li) concentrations and zirconium (Zr) concentrations using three types of spectra, namely X-ray fluorescence (XRF), visible to short-wave infrared (Vis-NIR-SWIR), and laser-induced breakdown spectroscopy (LIBS). The best-performing Li model was PLS using XRF (mean RMSEP: 0.078%) and the best sensor fusion model was ROSA (mean RMSEP: 0.087%). The best Zr model was a high-level fusion of PLS models (mean RMSEP: 0.042%).

Structural and microstructural thalamocortical network disruption in sporadic behavioural variant frontotemporal dementia

BackgroundUsing multi-block methods we combined multimodal neuroimaging metrics of thalamic morphology, thalamic white matter tract diffusion metrics, and cortical thickness to examine changes in behavioural variant frontotemporal dementia. (bvFTD). MethodTwenty-three patients with sporadic bvFTD and 24 healthy controls underwent structural and diffusion MRI scans. Clinical severity was assessed using the Clinical Dementia Rating scale and behavioural severity using the Frontal Behaviour Inventory by patient caregivers. Thalamic volumes were manually segmented. Anterior and posterior thalamic radiation fractional anisotropy and mean diffusivity were extracted using Tract-Based Spatial Statistics. Finally, cortical thickness was assessed using Freesurfer. We used shape analyses, diffusion measures, and cortical thickness as features in sparse multi-block partial least squares (PLS) discriminatory analyses to classify participants within bvFTD or healthy control groups. Sparsity was tuned with five-fold cross-validation repeated 10 times. Final model fit was assessed using permutation testing. Additionally, sparse multi-block PLS was used to examine associations between imaging features and measures of dementia severity. ResultsBilateral anterior-dorsal thalamic atrophy, reduction in mean diffusivity of thalamic projections, and frontotemporal cortical thinning, were the main features predicting bvFTD group membership. The model had a sensitivity of 96%, specificity of 68%, and was statistically significant using permutation testing (p = 0.012). For measures of dementia severity, we found similar involvement of regional thalamic and cortical areas as in discrimination analyses, although more extensive thalamo-cortical white matter metric changes. ConclusionsUsing multimodal neuroimaging, we demonstrate combined structural network dysfunction of anterior cortical regions, cortical-thalamic projections, and anterior thalamic regions in sporadic bvFTD.

Evaluating the impact of NIR pre-processing methods via multiblock partial least-squares

Near-infrared (NIR) spectral data are used in many applications to predict physical and chemical properties. However, it can result in poor predictive models when untreated spectra are directly used to estimate these properties. Many pre-preprocessing techniques are available to reduce noise and variance unrelated to the studied property but choosing which one to apply can be tricky. Existing methods to select a pre-processing are time-consuming or do not allow for a meaningful comparison of the different techniques. Even though new methods focus on extracting complementary information from each pre-processing, an optimal combination is still required to obtain efficient predictive models and avoid extensive computational costs. Here, we propose an approach using multiblock partial least squares (MBPLS) to simultaneously compare the impact of the pre-processing techniques on spectral data and as a result on the regression models. Superloadings provide qualitative and quantitative information on pre-processed data. This tool helps compare and determine which pre-processing technique, or combinations thereof, that may be appropriate for a dataset, not just a single “best” one. Using this, the analyst is then better equipped to make a final choice when selecting which ones to include. This method is tested on artificial signals and NIR spectra from corn samples.

Sparse PLS-Based Method for Overlapping Metabolite Set Enrichment Analysis.

Metabolite set enrichment analysis (MSEA) has gained increasing research interest for identification of perturbed metabolic pathways in metabolomics. The method incorporates predefined metabolic pathways information in the analysis where metabolite sets are typically assumed to be mutually exclusive to each other. However, metabolic pathways are known to contain common metabolites and intermediates. This situation, along with limitations in metabolite detection or coverage leads to overlapping, incomplete metabolite sets in pathway analysis. For overlapping metabolite sets, MSEA tends to result in high false positives due to improper weights allocated to the overlapping metabolites. Here, we proposed an extended partial least squares (PLS) model with a new sparse scheme for overlapping metabolite set enrichment analysis, named overlapping group PLS (ogPLS) analysis. The weight vector of the ogPLS model was decomposed into pathway-specific subvectors, and then a group lasso penalty was imposed on these subvectors to achieve a proper weight allocation for the overlapping metabolites. Two strategies were adopted in the proposed ogPLS model to identify the perturbed metabolic pathways. The first strategy involves debiasing regularization, which was used to reduce inequalities amongst the predefined metabolic pathways. The second strategy is stable selection, which was used to rank pathways while avoiding the nuisance problems of model parameter optimization. Both simulated and real-world metabolomic datasets were used to evaluate the proposed method and compare with two other MSEA methods including Global-test and the multiblock PLS (MB-PLS)-based pathway importance in projection (PIP) methods. Using a simulated dataset with known perturbed pathways, the average true discovery rate for the ogPLS method was found to be higher than the Global-test and the MB-PLS-based PIP methods. Analysis with a real-world metabolomics dataset also indicated that the developed method was less prone to select pathways with highly overlapped detected metabolite sets. Compared with the two other methods, the proposed method features higher accuracy, lower false-positive rate, and is more robust when applied to overlapping metabolite set analysis. The developed ogPLS method may serve as an alternative MSEA method to facilitate biological interpretation of metabolomics data for overlapping metabolite sets.

Far and wide: Associations between childhood socio-economic status and brain connectomics

Previous studies have identified localized associations between childhood environment – namely their socio-economic status (SES) – and particular neural structures. The primary aim of the current study was to test whether associations between SES and brain structure are widespread or limited to specific neural pathways. We employed advances in whole-brain structural connectomics to address this. Diffusion tensor imaging was used to construct whole-brain connectomes in 113 6−12 year olds. We then applied an adapted multi-block partial-least squares (PLS) regression to explore how connectome organisation is associated with childhood SES (parental income, education levels, and neighbourhood deprivation). The Fractional Anisotropy (FA) connectome was significantly associated with childhood SES and this effect was widespread. We then pursued a secondary aim, and demonstrated that the connectome mediated the relationship between SES and cognitive ability (matrix reasoning and vocabulary). However, the connectome did not significantly mediate SES relationships with academic ability (maths and reading) or internalising and externalising behavior. This multivariate approach is important for advancing our theoretical understanding of how brain development may be shaped by childhood environment, and the role that it plays in predicting key outcomes. We also discuss the limitations with this new methodological approach.

Identifying Significant Metabolic Pathways Using Multi-Block Partial Least-Squares Analysis.

In metabolomics, identification of metabolic pathways altered by disease, genetics, or environmental perturbations is crucial to uncover the underlying biological mechanisms. A number of pathway analysis methods are currently available, which are generally based on equal-probability, topological-centrality, or model-separability methods. In brief, prior identification of significant metabolites is needed for the first two types of methods, while each pathway is modeled separately in the model-separability-based methods. In these methods, interactions between metabolic pathways are not taken into consideration. The current study aims to develop a novel metabolic pathway identification method based on multi-block partial least squares (MB-PLS) analysis by including all pathways into a global model to facilitate biological interpretation. The detected metabolites are first assigned to pathway blocks based on their roles in metabolism as defined by the KEGG pathway database. The metabolite intensity or concentration data matrix is then reconstructed as data blocks according to the metabolite subsets. Then, a MB-PLS model is built on these data blocks. A new metric, named the pathway importance in projection (PIP), is proposed for evaluation of the significance of each metabolic pathway for group separation. A simulated dataset was generated by imposing artificial perturbation on four pre-defined pathways of the healthy control group of a colorectal cancer study. Performance of the proposed method was evaluated and compared with seven other commonly used methods using both an actual metabolomics dataset and the simulated dataset. For the real metabolomics dataset, most of the significant pathways identified by the proposed method were found to be consistent with the published literature. For the simulated dataset, the significant pathways identified by the proposed method are highly consistent with the pre-defined pathways. The experimental results demonstrate that the proposed method is effective for identification of significant metabolic pathways, which may facilitate biological interpretation of metabolomics data.

Authentication of Grappa (Italian grape marc spirit) by Mid and Near Infrared spectroscopies coupled with chemometrics

The aim of the present study is to authenticate Grappa spirits and to develop a non-destructive methodology which would allow detecting possible adulteration (by less valuable spirits) on this product. Grappa is an Italian alcoholic drink obtained by distillation of grape marks which has recently received the Geographical Indication (GI) label. As a high added-value product, it is relevant to develop methodologies which allow its authentication and detecting possible frauds (e.g., adulterations); and, whether feasible, it would be suitable to achieve these goals through non-destructive approaches (in order to minimize the economic loss). Mid Infrared (MIR) and Near Infrared (NIR) spectroscopies have been used for the authentication and the characterization of the spirits under investigation. The present work is conceptually divided into two parts: the first one, centered on the authentication of grappa spirits, focused on distinguishing them from other Italian distillates, and a second one aimed at developing an analytical methodology suitable to discern between pure and adulterated grappas. Both classification problems have been investigated by PLS-DA and by three multi-block strategies, i.e., Multi-Block Partial Least Squares (MB-PLS), Sequential and Orthogonalized Partial Least Squares (SO-PLS) and Sequential and Orthogonalized Covariance Selection (SO-CovSel) in order to test whether a data-fusion approach would lead to an improvements of the classification rates. The best results (in terms of predictions) were provided by multi-block strategies; in particular, they provided 100 % of correct classification when applied to discriminate pure and adulterated samples, suggesting these methodologies are definitely suitable for the proposed purpose.

Monitoring microsphere coating processes using PAT tools in a bench scale fluid bed

Among the factors that influence adherence to medication within the pediatric population, taste/irritation has been identified as a critical barrier to patient compliance. With the goal of improving compliance, microspheres (matrix systems within which the drug is dispersed) can be coated with a reverse enteric polymer that will prevent the release of the drug in the oral cavity while maintaining an immediate release once the drug product reaches the stomach, thereby achieving a taste neutral profile. In this work, the in-line performance of three process analytical technology (PAT) tools is evaluated in order to monitor the microsphere coating process. These tools are Raman spectroscopy, near-infrared spectroscopy and focused beam reflectance measurements, together with process data and raw material attributes. The ability of these different sources of information to predict the coating's barrier performance is evaluated by using a combined-data-approach: multiblock partial least squares (MBPLS). Results show that Raman spectroscopy has a superior predictive performance and that it has the potential to monitor the coating process of the microspheres as well as to detect process discrepancies (such as spray rate changes), demonstrating its usefulness for the monitoring of fluid bed coating processes. It was also demonstrated that Raman can be used to clearly differentiate batches with significantly difference in-vitro dissolution performance. This monitoring is considered critical to ensure consistent coating performance for this thin film barrier membrane that is essential to patient compliance.

NIR and MIR spectral data fusion for rapid detection of Lonicera japonica and Artemisia annua by liquid extraction process

Lonicera japonica and Artemisia annua are two widely used plants that are typically combined together as botanical dietary supplements or herbal medicines. In this study, the liquid extraction process of various mixtures of these two plants was investigated with near-infrared (NIR) and mid-infrared (MIR) spectroscopy for rapid detection of important effective compounds. Eight main effective compounds of Lonicera japonica and Artemisia annua mixtures were modeled and selected as quality attributes: neochlorogenic acid (5-CQA), cryptochlorogenic acid (4-CQA), chlorogenic acid (3-CQA), caffeic acid (CA), isochlorogenic acid B (ICAB), isocholorogenic acid A (ICAA), isocholorogenic acid C (ICAC), and secoxyloganin (SL). First, partial least squares (PLS) was adopted to build quantitative models by combining spectra from MIR and NIR instruments. Two multiblock PLS methods were applied using NIR and MIR data matrix in which concatenated PLS (C-PLS) and sequential and orthogonalized PLS (SO-PLS) were utilized to optimize model performance. The results showed that models based on single NIR and MIR spectroscopy had close capacity for detection of the eight indexes. However, performance of the fusion methods C-PLS and SO-PLS both achieved better results than the results of single NIR or MIR PLS models, with SO-PLS being superior to C-PLS. This indicates that the multiblock methodology can be used to optimize rapid detection of indexes from complex industrial processes of plants using NIR and MIR spectroscopy.

Optimal Selection of Incoming Materials from the Inventory for Achieving the Target Drug Release Profile of High Drug Load Sustained-Release Matrix Tablet.

In the pharmaceutical process, raw material (including APIs and excipients) variability can be delivered to the final product, and lead to batch-to-batch and lot-to-lot variances in its quality, finally impacting the efficacy of the drug. In this paper, the Panax notoginseng saponins (PNS) sustained-release matrix tablet was taken as the model formulation. Hydroxypropyl methylcellulose with the viscosity of 4000mPa·s (HPMCK4M) from different vendors and batches were collected and their physical properties were characterized by the SeDeM methodology. The in-vitro dissolution profiles of active pharmaceutical ingredients (APIs) from matrix tablets made up of different batches HPMC K4M displayed significant variations. Multi-block partial least squares (MB-PLS) modeling results further demonstrated that physical properties of excipients played dominant roles in the drug release. In order to achieve the target drug release profile with respect to those far from the criteria, the optimal selection method of incoming materials from the available was established and validated. This study provided novel insights into the control of the input variability of the process and amplified the application of the SeDeM expert system, emphasizing the importance of the physical information of the raw materials in the drug manufacturing process.

Mid-infrared and near-infrared spectroscopy for rapid detection of Gardeniae Fructus by a liquid-liquid extraction process

Gardeniae Fructus is widely used in the pharmaceutical industry, and many studies have confirmed its medical and economic value. In this study, samples collected from different liquid-liquid extraction batches of Gardeniae Fructus were detected by mid-infrared (MIR) and near-infrared (NIR) spectroscopy. Seven analytes, neochlorogenic acid (5-CQA), cryptochlorogenic acid (4-CQA), chlorogenic acid (3-CQA), geniposidic acid (GEA), deacetyl-asperulosidic acid methyl ester (DAAME), genipin-gentiobioside (GGB), and gardenoside (GA), were chosen as quality property indexes of Gardeniae Fructus. The two kinds of spectra were each used to build models by single partial least squares (PLS). Additionally, both spectral data were combined and modeled by multiblock PLS. For single spectroscopy modeling results, NIR had a better prediction for high-concentration analytes (3-CQA, DAAME, GGB, and GA) whereas MIR performed better for low-concentration analytes (5-CQA, 4-CQA, and GEA). The multiblock methodology was found to be better compared to single spectroscopy models for all seven analytes. Specifically, the coefficients of determination (R2) of the NIR, MIR, and multiblock PLS calibration models of all seven components were higher than 0.95. Relative standard errors of prediction (RSEP) were all less than 7%, except for models of GGB, which were 10.36%, 13.24%, and 8.15% for the NIR-PLS, MIR-PLS, and multiblock models, respectively. These results indicate that MIR and NIR spectrographic techniques could provide a new choice for quality control in industrial production of Gardeniae Fructus.

Statistical modeling methods to analyze the impacts of multiunit process variability on critical quality attributes of Chinese herbal medicine tablets.

The quality of Chinese herbal medicine tablets suffers from batch-to-batch variability due to a lack of manufacturing process understanding. In this paper, the Panax notoginseng saponins (PNS) immediate release tablet was taken as the research subject. By defining the dissolution of five active pharmaceutical ingredients and the tablet tensile strength as critical quality attributes (CQAs), influences of both the manipulated process parameters introduced by an orthogonal experiment design and the intermediate granules’ properties on the CQAs were fully investigated by different chemometric methods, such as the partial least squares, the orthogonal projection to latent structures, and the multiblock partial least squares (MBPLS). By analyzing the loadings plots and variable importance in the projection indexes, the granule particle sizes and the minimal punch tip separation distance in tableting were identified as critical process parameters. Additionally, the MBPLS model suggested that the lubrication time in the final blending was also important in predicting tablet quality attributes. From the calculated block importance in the projection indexes, the tableting unit was confirmed to be the critical process unit of the manufacturing line. The results demonstrated that the combinatorial use of different multivariate modeling methods could help in understanding the complex process relationships as a whole. The output of this study can then be used to define a control strategy to improve the quality of the PNS immediate release tablet.

Latent variable modeling to analyze the effects of process parameters on the dissolution of paracetamol tablet

ABSTRACTThe dissolution is one of the critical quality attributes (CQAs) of oral solid dosage forms because it relates to the absorption of drug. In this paper, the influence of raw materials, granules and process parameters on the dissolution of paracetamol tablet was analyzed using latent variable modeling methods. The variability in raw materials and granules was understood based on the principle component analysis (PCA), respectively. A multi-block partial least squares (MBPLS) model was used to determine the critical factors affecting the dissolution. The results showed that the binder amount, the post granulation time, the API content in granule, the fill depth and the punch tip separation distance were the critical factors with variable importance in the projection (VIP) values larger than 1. The importance of each unit of the whole process was also ranked using the block importance in the projection (BIP) index. It was concluded that latent variable models (LVMs) were very useful tools to extract information from the available data and improve the understanding on dissolution behavior of paracetamol tablet. The obtained LVMs were also helpful to propose the process design space and to design control strategies in the further research.

Respective impact of bread structure and oral processing on dynamic texture perceptions through statistical multiblock analysis

Texture perception is a multidimensional and dynamic phenomenon resulting from both the initial structure of food and its breakdown during oral processing. The aim of this study is to identify the respective contribution of food and bolus properties to temporal changes in texture perceptions during bread consumption. For this purpose, the perception dynamics of three French baguettes with dif\\ferent structures were assessed through Temporal Dominance of Sensations and Progressive Profiling. Samples of crumb with and without crust were tasted by trained panelists. The intensities of nine texture attributes were evaluated at three key stages of oral processing (10%, 40% and 100% of individual swallowing time) using the Progressive Profiling method. Six of them were related with a Multiblock Partial Least Squares (MB-PLS) regression to the initial bread properties and to some bolus properties measured at these three stages. The evolution during oral processing of some attributes such as “soft”, “dry”, “doughy” and “sticky” was more influenced by modifications of bolus properties than by the initial characteristics of the breads. Among bolus properties, the MB-PLS highlighted that the hydration and texture properties of the bolus had a greater impact on texture perceptions than bolus structure. The “aerated” perception was more affected by the crumb structure, while the “heterogeneousness” and the “crispiness” were more affected by the presence of crust. This study thus contributes to improving our understanding of dynamic texture perceptions through a statistical model that takes the physical properties of bread and bolus during oral processing into account.

Variable selection in multi-block regression

The focus of the present paper is to propose and discuss different procedures for performing variable selection in a multi-block regression context. In particular, the focus is on two multi-block regression methods: Multi-Block Partial Least Squares (MB-PLS) and Sequential and Orthogonalized Partial Least Squares (SO-PLS) regression. A small simulation study for regular PLS regression was conducted in order to select the most promising methods to investigate further in the multi-block context. The combinations of three variable selection methods with MB-PLS and SO-PLS are examined in detail. These methods are Variable Importance in Projection (VIP) Selectivity Ratio (SR) and forward selection. In this paper we focus on both prediction ability and interpretation. The different approaches are tested on three types of data: one sensory data set, one spectroscopic (Raman) data set and a number of simulated multi-block data sets.

Combining DI-ESI-MS and NMR datasets for metabolic profiling.

Metabolomics datasets are commonly acquired by either mass spectrometry (MS) or nuclear magnetic resonance spectroscopy (NMR), despite their fundamental complementarity. In fact, combining MS and NMR datasets greatly improves the coverage of the metabolome and enhances the accuracy of metabolite identification, providing a detailed and high-throughput analysis of metabolic changes due to disease, drug treatment, or a variety of other environmental stimuli. Ideally, a single metabolomics sample would be simultaneously used for both MS and NMR analyses, minimizing the potential for variability between the two datasets. This necessitates the optimization of sample preparation, data collection and data handling protocols to effectively integrate direct-infusion MS data with one-dimensional (1D) 1H NMR spectra. To achieve this goal, we report for the first time the optimization of (i) metabolomics sample preparation for dual analysis by NMR and MS, (ii) high throughput, positive-ion direct infusion electrospray ionization mass spectrometry (DI-ESI-MS) for the analysis of complex metabolite mixtures, and (iii) data handling protocols to simultaneously analyze DI-ESI-MS and 1D 1H NMR spectral data using multiblock bilinear factorizations, namely multiblock principal component analysis (MB-PCA) and multiblock partial least squares (MB-PLS). Finally, we demonstrate the combined use of backscaled loadings, accurate mass measurements and tandem MS experiments to identify metabolites significantly contributing to class separation in MB-PLS-DA scores. We show that integration of NMR and DI-ESI-MS datasets yields a substantial improvement in the analysis of neurotoxin involvement in dopaminergic cell death.

Iterative weighting of multiblock data in the orthogonal partial least squares framework

The integration of multiple data sources has emerged as a pivotal aspect to assess complex systems comprehensively. This new paradigm requires the ability to separate common and redundant from specific and complementary information during the joint analysis of several data blocks. However, inherent problems encountered when analysing single tables are amplified with the generation of multiblock datasets. Finding the relationships between data layers of increasing complexity constitutes therefore a challenging task. In the present work, an algorithm is proposed for the supervised analysis of multiblock data structures. It associates the advantages of interpretability from the orthogonal partial least squares (OPLS) framework and the ability of common component and specific weights analysis (CCSWA) to weight each data table individually in order to grasp its specificities and handle efficiently the different sources of Y-orthogonal variation.Three applications are proposed for illustration purposes. A first example refers to a quantitative structure-activity relationship study aiming to predict the binding affinity of flavonoids toward the P-glycoprotein based on physicochemical properties. A second application concerns the integration of several groups of sensory attributes for overall quality assessment of a series of red wines. A third case study highlights the ability of the method to combine very large heterogeneous data blocks from Omics experiments in systems biology. Results were compared to the reference multiblock partial least squares (MBPLS) method to assess the performance of the proposed algorithm in terms of predictive ability and model interpretability. In all cases, ComDim-OPLS was demonstrated as a relevant data mining strategy for the simultaneous analysis of multiblock structures by accounting for specific variation sources in each dataset and providing a balance between predictive and descriptive purpose.

Integrating multiple analytical platforms and chemometrics for comprehensive metabolic profiling: application to meat spoilage detection

Untargeted metabolic profiling has become a common approach to attempt to understand biological systems. However, due to the large chemical diversity in the metabolites it is generally necessary to employ multiple analytical platforms so as to encompass a wide range of metabolites. Thus it is beneficial to find chemometrics approaches which can effectively integrate data generated from multiple platforms and ideally combine the strength of each platform and overcome their inherent weaknesses; most pertinent is with respect to limited chemistries. We have reported a few studies using untargeted metabolic profiling techniques to monitor the natural spoilage process in pork and also to detect specific metabolites associated with contaminations with the pathogen Salmonella typhimurium. One method used was to analyse the volatile organic compounds (VoCs) generated throughout the spoilage process while the other was to analyse the soluble small molecule metabolites (SMM) extracted from the microbial community, as well as from the surface of the spoiled/contaminated meat. In this study, we exploit multi-block principal component analysis (MB-PCA) and multi-block partial least squares (MB-PLS) to combine the VoCs and SMM data together and compare the results obtained by analysing each data set individually. We show that by combining the two data sets and applying appropriate chemometrics, a model with much better prediction and importantly with improved interpretability was obtained. The MB-PCA model was able to combine the strength of both platforms together and generated a model with high consistency with the biological expectations, despite its unsupervised nature. MB-PLS models also achieved the best over-all performance in modelling the spoilage progression and discriminating the naturally spoiled samples and the pathogen contaminated samples. Correlation analysis and Bayesian network analysis were also performed to elucidate which metabolites were correlated strongly in the two data sets and such information could add additional information in understanding the meat spoilage process.

Quantitative structure–activity relationship of organophosphate compounds based on molecular interaction fields descriptors

By using multi-block partial least-squares (MBPLS) method, quantitative structure–activity relationship (QSAR) between 35 organophosphate compounds (OP) and their 24h acute toxicities towards the housefly (Musca nebulo L.) was built on the molecular interaction fields (MIF) descriptors, which were obtained with O, N and DRY as probes, and then normalised with block unscaled weights (BUW) technique. The best QSAR model had 8 principal components, with the coefficient of determination R2=0.995 and that of leave-one-out cross-validation Q2=0.865, and the corresponding standard deviation of error 0.076 and 0.361, respectively. Block importance in the prediction (BIP) for O, N and DRY probe were 1.030, 0.962 and 1.007, respectively. Contour map of variable coefficients showed that hydrogen bonding between the O atom in PO and the NH groups in acetylcholinesterase (AChE) played an important role in the interaction between OP and AChE. Meanwhile, the hydrophobicity of OP also had significant contribution. QSAR based on the MIF descriptors could be a potential means to interpret the mechanisms of ligand–receptor interaction when the receptor was well known.