Multivariate Statistical Data Analysis Research Articles

Metabolomics analysis of seminal plasma in patients with idiopathic Oligoasthenoteratozoospermia using high-resolution NMR spectroscopy.

Male infertility is a global health issue caused by a combination of different factors. Specialists generally rely on semen analysis to diagnose male infertility. However, it is known that diagnostic semen analysis fails to identify about 50% of male infertility disorders. Recently, metabolomics has been proven to be a powerful technique for the diagnosis of different diseases. To determine whether metabolites could be used as potential biomarkers for the diagnosis of male factor infertility through comparing seminal plasma samples from infertile men with oligoasthenoteratozospermia (OAT) and samples from normozoospermic controls. This study utilized high-resolution 1 H NMR spectroscopy to reveal whether the metabolomic changes of seminal plasma obtained from 31 patients with oligoasthenoteratozospermia (OAT) are different from the ones obtained from 28 normozoospermic controls. Multivariate statistical analysis of NMR data concluded that the metabolomic profile of samples from patients with OAT exhibits statistically significant differences when compared to the controls. The differences were based on the metabolites lactate, citrate, lysine, arginine, valine, glutamine, creatinine, α-ketoglutaric acid, spermine, putrescine, and tyrosine. Except the tyrosine, levels of the above metabolites were significantly decreased in patients with OAT compared to the controls. The levels of citrate, choline, spermine, putrescine, α-ketoglutaric acid, valine, and tyrosine were significantly different (p<5×10-4 ) between two groups. On the other hand, levels of lactate, creatinine, lysine, arginine, and glutamine were also statistically significant (0.001<p<0.05). However, considering the p-values, the physiological relevance of these metabolites may be lower when compared to the others. A PLS-DA model built on the NMR data achieved 89.29% sensitivity and 93.55% specificity results in a leave-one-out cross-validation process. 1 H NMR spectroscopy-based metabolomic analysis could be used as a diagnostic tool for the diagnosis of oligoasthenoteratozospermia.

Proteomic analysis of amniotic and allantoic fluid from buffaloes during foetal development.

The objective of this study was to describe the dynamic changes in protein composition and protein abundance in amniotic and allantoic fluids from buffaloes during gestation. Amniotic and allantoic fluids were collected during the first, second and third trimesters of gestation. The foetuses were measured and weighed. Fluid samples were centrifuged at 800g for 10min and then at 10,000g for 60min at 4°C. The supernatant was collected to determine the total protein concentration. Based on total protein concentration, an aliquot (50μg) was used for in-solution tryptic digestion, and mass spectrometry analysis (nano-LC-MS/MS) was performed. A multivariate statistical analysis of the proteomic data was conducted. Across the different stages of buffalo gestation, fifty-one proteins were found in the amniotic fluid, and twenty-one were found in the allantoic fluid. A total of twelve proteins were common among the stages, and four presented significant differences (VIP score α>1). Fibronectin and alpha-1-antiproteinase were more abundant in the amniotic fluid than in the allantoic fluid. Alpha-2-macroglobulin and alpha-2-HS-glycoprotein were more abundant in the allantoic fluid than in the amniotic fluid. Alpha-2-macroglobulin participates in remodelling and growth of the uterus at beginning of the gestation (first trimester), and these findings indicate that can serve as a potential tool for the early diagnosis of pregnancy in buffaloes.

Characterization and discrimination of selected chicken eggs in China's retail market based on multi-element and lipidomics analysis.

The present study focuses on the differentiation of Deqingyuan (n = 90), Taihe (n = 90) and crossbred (n = 90) black-boned silky fowl eggs based on chemometrics analysis of multi-element and lipidomics data. A total of 53 chemical elements were simultaneously quantified in the eggs using inductively coupled plasma mass spectrometry. Multivariate statistical analysis of element data screened 30, 13 and 12 candidate discriminants for whole egg, egg yolk and egg white samples, respectively. Moreover, we utilized an LC-MS/MS-based lipidomics approach to profile lipids in the three types of egg yolks. In total, 1633 lipid species (including 43.78% glycerophospholipids, 25.66% glycerolipids, 16.66% fatty acyls, 6.86% sphingolipids, 4.10% sterol lipids, 1.53% polyketides, 1.10% prenol lipids and 0.31% saccharolipids) were consistently detected in all samples. A partial least squares discriminant analysis model was established based on the lipidomics data, which enables a 100% correct classification of the egg yolks. Using a cutoff of variable importance in projection value >1 and p value <.05, a panel of 22 potential lipid markers was discovered for the discrimination of Deqingyuan, Taihe and crossbred egg yolks. These results could facilitate a better understanding of the nutritional qualities of the evaluated eggs. Our method of combining chemical analysis with chemometrics offers a convenient and powerful tool to distinguish commercially-available eggs.

An Integrated Approach for Studying the Hydrology of the Ljubljansko Polje Aquifer in Slovenia and Its Simulation

Groundwater and surface water are strongly connected. Therefore, understanding their interactions is important when studying the water balance of a complex aquatic system. This paper aims to present an integrated approach to study such processes, including a better understanding of the hydrological system behavior in the Ljubljansko polje (Slovenia). The study is based on multivariate statistical analyses of data collected over a long period, including the isotopic composition of groundwater, river water, and precipitation. The hydrology in the study domain was also simulated using a comprehensive modelling framework. Since boundary conditions are essential for simulating groundwater flow in a sensitive aquifer, a modelling system of rivers and channels (MIKE 11) and water flow and balance simulation model (WaSiM) were used to model river dynamics and the percolation of local precipitation, respectively. The results were then used as boundary conditions imposed on a transient state groundwater flow model performed in finite element subsurface flow simulation system (FEFLOW 6.2). Both the locations of recharge areas in the study domain and the calculated fluxes between the Sava River and the aquifer are graphically presented. The study revealed that a combination of the MIKE 11-FEFLOW-WaSiM tools offers a good solution for performing parallel simulations of groundwater and surface water dynamics.

Meat quality traits and canonical discriminant analysis to identify the use of illicit growth promoters in Charolais bulls

The administration of anabolic agents in farm animals to improve meat production has been prohibited in EU, due to the potential risks to human health. Meat quality was investigated to detect the effects of illegal administration of dexamethasone or prednisolone or 17β-estradiol on Charolais bulls. Three groups of 6 bulls were treated and 12 bulls were the control. Meat quality parameters were measured on live animals, carcasses and on samples of Longissimus thoracis and multivariate statistical data analysis was applied. In Charolais bulls, these parameters were affected by growth promoter administration and the multivariate canonical discriminant analysis was able to distinguish between treated and untreated animals mainly due to three electronic nose’s parameters, 24 h carcass temperature and drip loss. Therefore, meat quality control and the multivariate analysis could be useful as a first screening to address targeted controls on farms suspected of illicit use of growth promoters.

Interrelation of Volatile Organic Compounds and Sensory Properties of Alternative and Torrefied Wood Pellets

The increasing demand for wood pellets on the market, which is caused by their excellent combustion properties, inspires the production as well as the utilization of alternative biomass pellets as fuel. However, the emission of volatile organic compounds gives pellet materials a distinct odor or off-odor, which is directly perceived by the end user. Thus, there is an urgent need for knowledge about the emitted volatile organic compounds and their potential formation pathways as well as their contributions to odor properties of the pellets. In this study, pellets made of biomass energy crops (i.e., straw or miscanthus), byproducts from the food industry (i.e., rapeseed, grapevine, or DDGS (dried distillers grains with solubles from beer production)), or eucalyptus, as well as torrefied pinewood and torrefied sprucewood were investigated with respect to the emitted volatile compounds and their possible impact on the pellet odor. Headspace solid-phase microextraction in combination with gas chromatography–mass spectrometry was used to enrich, separate, and identify the compounds. Techniques used in sensory science were applied to obtain information about the odor properties of the samples. A total of 59 volatile compounds (acids, aldehydes and ketones, alcohols, terpenes, heterocyclic compounds, and phenolic compounds) were identified with different compound ratios in the investigated materials. The use of multivariate statistical data analysis provided deep insight into product–compound interrelation. For pellets produced from bioenergy crops, as well as from byproducts from the food industry, the sensory properties of the pellets reflected the odor properties of the raw material. With respect to the volatiles from torrefied pellets, those volatiles that are formed during the torrefaction procedure dominate the odor of the torrefied pellets covering the genuine odor of the utilized wood. The results of this work serve as a substantiated basis for future production of pellets from alternative raw materials.

An integrative investigation on the efficacy of Plantaginis semen based on UPLC-QTOF-MS metabolomics approach in hyperlipidemic mice

Plantaginis semen, the dried mature seed of Plantago asiatica L. or Plantago deprdssa Willd., has a prominent effect on the treatment of obesity, type 2 diabetes and lipid disorders, however, its clinical application is limited due to inadequate in-depth mechanism exploration and incomplete discussion of action targets of its in vivo. Therefore, an untargeted metabolomics approach was firstly applied to study the serum metabolic differences in mice. Metabolomics analysis was performed using ultra performance liquid chromatography quadrupole time of flight mass spectrometry (UPLC-QTOF-MS) together with multivariate statistical data analysis. The results showed that Plantaginis semen can mainly improve blood lipids, some degree in blood glucose and insulin levels in high-fat mice, in addition, the phenotype of liver and fat stained sections demonstrated remarkable results. A total of 22 metabolites involved in arachidonic acid, glycerophospholipid, glycosphingolipid, linoleate, Omega-3 fatty acid, phosphatidylinositol phosphate and tyrosine metabolisms were identified. In further, it was found that the possible mechanisms of Plantaginis semen on hyperlipidemic mice lied in the biosynthesis of thyroxine, biological effects of enzymes of phospholipase A2 activity, glucosylceramide synthase and inositol essential enzyme 1α, genes expressions of fatty acid metabolism and inflammation. Serum metabolomics revealed that Plantaginis semen could cure the organism disease via regulating multiple metabolic pathways which will be helpful for understanding the mechanism of this herb and providing references for better applications of it in clinic, even researches on other TCMs.

Data-Driven Analysis of Antimicrobial Resistance in Foodborne Pathogens from Six States within the US.

Foodborne pathogens cause thousands of illnesses across the US each year. However, these pathogens gain resistance to the antimicrobials that are commonly used to treat them. Typically, antimicrobial resistance is caused by mechanisms encoded by multiple antimicrobial-resistance genes. These are carried through pathogens found in foods such as meats. It is, thus, important to study the genes that are most related to antimicrobial resistance, the pathogens, and the meats carrying antimicrobial-resistance genes. This information can be further used to correlate the antimicrobial-resistance genes found in humans for improving human health. Therefore, we perform the first multivariate statistical analysis of the antimicrobial-resistance gene data provided in the NCBI Pathogen Detection Isolates Browser database, covering six states that are geographically either in close proximity to one another (i.e., Pennsylvania (PA), Maryland (MD), and New York (NY)) or far (i.e., New Mexico (NM), Minnesota (MN), and California (CA)). Hundreds of multidimensional data points were projected onto a two-dimensional space that was specified by the first and second principal components, which were then categorized with a hierarchical clustering approach. It turns out that aadA, aph(3’’), aph(3’’)-Ib, aph(6)-I, aph(6)-Id, bla, blaCMY, tet, tet(A), and sul2 constructed the assembly of ten genes that were most commonly involved in antimicrobial resistance in these six states. While geographically close states like PA, MD and NY share more similar antimicrobial-resistance genes, geographically far states like NM, MN, and CA also contain most of these common antimicrobial-resistance genes. One potential reason for this spread of antimicrobial-resistance genes beyond the geographic limitation is that animal meats like chicken and turkey act as the carriers for the nationwide spread of these genes.

Combined microbial and isotopic signature approach to identify nitrate sources and transformation processes in groundwater

Nitrate (NO3−) pollution is a serious problem worldwide. Identification of NO3− sources and transformation processes in aquifers is a key step in effectively controlling and mitigating NO3− contamination. In this study, hydrochemical, microbial, and dual isotopic approaches were integrated to elucidate the sources and processes influencing NO3− contamination in the Pearl River Delta, China. The results showed a severe NO3− contamination, with 75% of the samples having NO3−-N concentrations above the WHO standard of 10 mg L−1. The δ15NNO3- and δ18ONO3- values and a multivariate statistical analysis of hydrochemical data both revealed that manure and sewage were mainly responsible for NO3− contamination. Biological indicators further demonstrated that, manure and sewage had greater impacts on groundwater quality during the rainy season than during the dry season. Based on the significant relationships of δ15NNO3- and δ18ONO3- with the logarithmic NO3− concentration (Ln(NO3−)), denitrification was confirmed to occur in the discharge zone during the rainy season. Proteobacteria, Bacteroidetes, and Planctomycetes were identified as the dominant phyla, and Dechloromonas, Flavobacterium, and Nitrospira were dominant among the denitrifying bacteria in groundwater. The abundance of denitrifying bacteria had significant positive correlations with δ15NNO3- and NO2−-N during the rainy season, further confirming the occurrence of denitrification during the rainy season. This study showed that dual isotope techniques combined with microbial data can be a powerful tool for identifying the sources and microbial processes affecting NO3− in groundwater. Moreover, the results can provide useful insights for environmental managers to verify groundwater pollution and better apply remediation solutions.

GC-MS metabolomics comparison of yoghurts from sheep's and goats' milk

In this study, the polar metabolite profile of commercial yoghurt samples produced in Sardinia (Italy) from milk of local sheep and goats was studied by GC-MS and multivariate statistical data analysis (MVA). Milks underwent the same manufacturing procedures and yoghurts were analysed one day post-manufacture. Results of discriminant analysis indicated that the two yoghurt types had very different metabolite profiles, with different levels of health promoting compounds. Goats' milk yoghurt was richer in free amino acids, γ-aminobutyric acid, pyroglutamic acid and β-phenyllactic acid when compared with yoghurt produced with sheep's milk. Sheep's milk yoghurt was characterised by higher levels of myo-inositol, N-acetylgalactosamine and N-acetylglucosamine. Comparing yoghurt metabolites with those of the original milk, it was found that goats' milk underwent stronger metabolite changes after inoculum. The comparison between the two yoghurt types gave us a deeper insight on the effects of manufacture on different milks.

Metabolic Characteristics of WBC-deplted RBC suspension during Different Storage Stage in MAP Based on UPLC-MS/MS

To explore the metabolic characteristics and metabolic markers of WBC-depleted RBCs in MAP preservation solution and to analyzed the metabolic profile of RBC in MAP preservation solution by using metabolomics. The changes of metabolitcs in 10 U WBC-depleted RBC suspension at 3-different storage period (D 0, D 14 and D 35) were detected by using the UPLC-MS/MS, the charaeteristic ions and metabolic markers of RBC stored in preservation sblution for 0 d, 14 d and 35 days were analyzed by using the principal component analysis(PCA). The number of characteristic ions in RBC and supernatant extracts detected during the initial, middle and final storage could be clearly distinguiseed. The 5 metabolism-related substamces such as lact-c acid, nicotinamide, glucose, 5-htdroxyproline and malic acid showed statistically significant difference in 3 storage period. The UPLC-MS/MS method combined with statistical analysis of multivariate data can be used to study the metabolic characteristics of RBC, the different metabolites of RBC in different storage stages can be used as the potential markers for evaluation of guality of RBC in storag period. The results of this study provide a basis for studing the RBC guality changes in storage period.

Depositional and organofacies influences on the petroleum potential of an unusual marine source rock: Waipawa Formation (Paleocene) in southern East Coast Basin, New Zealand

Abstract The Paleocene Waipawa Formation is a widespread, ∼2–80 m-thick marine mudstone that occurs in several New Zealand basins. It is relatively enriched in organic matter (OM) and is the inferred source of a sub-commercial oil discovery and several correlated oil seeps. To determine the organofacies and depositional influences on the petroleum potential of the formation, we have undertaken a high-resolution, multidisciplinary study of the Taylor White section in the central East Coast Basin. This section is the thickest stratigraphically coherent surface exposure of the formation and includes contacts with the bounding Whangai and Wanstead formations. Bathymetric indicators within the foraminiferal assemblages indicate a slope setting for the section, deepening from middle to lower bathyal through the Waipawa–Wanstead succession. Significantly, the Whangai Formation in this section is barren of foraminifera. Dinoflagellate and calcareous nannofossil biostratigraphy provides somewhat conflicting age control but a resulting age model indicates a tenfold increase in compacted sediment accumulation rate during Waipawa deposition (from ∼1 to ∼10 cm/ky). Our multivariate statistical analysis of geochemical and paleontological data reveals systematic variation in paleoenvironmental and source rock parameters within the section, with samples forming four distinct clusters or organofacies: OM-rich and OM-poor Waipawa organofacies, Whangai, and Wanstead organofacies. The two Waipawa organofacies are distinguished by enrichments in total organic carbon (TOC) and organic 13C, dominance of phytoclasts (mainly degraded woody plant matter; 66–98%), and low amounts of amorphous organic matter ( Whangai and Wanstead organofacies are characterised by a greater proportion of marine OM but are distinguished from each other by differences in OM preservation: OM is well-preserved with indications of hypoxia in the Whangai organofacies but is poorly preserved in the Wanstead organofacies. Geochemical fingerprinting identifies an interval of Whangai organofacies that separates two phases of Waipawa organofacies deposition. The lower phase comprises four alternations between OM-rich and OM-poor Waipawa organofacies and the upper phase comprises a single pulse of OM-rich Waipawa facies bounded by OM-poor intervals. Overall, Waipawa Formation in this section has TOC and pyrolysable hydrocarbon (S2) values of 0.2–4.6 (mean 2.2) wt% and 0.1–12.4 (mean 3.9) mg HC/g rock, respectively. One-third of S2 values exceed 5 mg HC/g rock, indicating good‒very good bulk petroleum potential, with the remainder having poor‒fair potential. Petroleum potential (S2) increases with increasing TOC (R2 = 0.85). Hydrogen index (HI) values range from 21 to 295 mg HC/g TOC, averaging just 147 mg HC/g TOC. In line with these modest HI values, quantitative pyrolysis-gas chromatography confirms that the petroleum potential is primarily for gas or gas-condensate, with relatively little oil potential. Of the total paraffinic petroleum yield, gas (n-C1–5) contributes 80–100%, whilst total (n-C6+) and non-volatile (n-C15+) paraffinic oil contribute ≤20% and ≤6%, respectively. Despite the low paraffinic oil potentials, n-C6+ and n-C15+ correlate relatively strongly with HI, and less strongly with TOC. The overall gas-condensate-prone nature of the Waipawa organofacies at Taylor White reflects the predominance of woody phytoclasts, which are unlikely to produce paraffinic oil. The small paraffinic oil component is instead more likely to be associated with the marine algal component. Although this component is volumetrically subordinate, the higher paraffinic oil potential appears to result from the combination of high productivity of specific marine algae (i.e., pelagophytes), hypoxic depositional conditions and, most importantly, increased preservation of OM, including through sulfurisation. This raises the possibility of more algal-rich, oil-prone facies existing within the Waipawa Formation in areas with higher TOC or, alternatively, in areas beyond the direct reach of the terrestrial influxes, yet still sufficiently influenced by low water-column oxygen levels and high sulfur contents that promote enhanced OM preservation.

Utility of multivariate statistical analysis to identify factors contributing river water quality in two different seasons in cold-arid high-altitude region of Leh-Ladakh, India

Monitoring water quality of surface water resources is the key concern in determining the potable water quality in high-altitude region. Therefore, there is a need to evaluate different parameters affecting water quality of river and identify the most important variables and factors significantly affecting water quality. In the present study, multivariate statistical methods including cluster analysis and principal component analysis/factor analysis were applied to analyze the Indus River water quality in the Trans-Himalayan region of India. For this total 25 no. of physicochemical parameters were analyzed in water samples taken from seven different monitoring sites in summer and winter season. All the physical, microbial, chemical, and mineral parameters were analyzed by using the standard methods of American Public Health Association, whereas minerals were determined with the inductively coupled plasma optical emission of spectroscopy method. Thereafter, experimental two-season (28 samples × 25 parameters) matrices of both the seasons were run through the multivariate statistical data analysis. The varifactors obtained from the FA of both the seasons and results indicate that the parameters responsible for water quality variations are mainly related to discharge and temperature (natural), organic pollution (point source: domestic sanitary waste), and nutrients (non-point sources: agriculture) in the summer season. However, in the winter seasons, results showed that the river water was less affected by anthropogenic activities and natural weathering process. Therefore, it is concluded that quality of Indus River water is affected by agricultural, domestic, and hydrogeochemical sources in the summer season. These findings corroborate suitability of multivariate statistical techniques in the elucidation of various parameters for water quality monitoring and determination of different contamination sources.

Exploration and Sources of Bauxite Deposit in the Boloven Plateau, Southern Laos

The Boloven bauxite deposit occurs either in the weathered basalt (alkali basalt and tholeiite), or in the Cretaceous sandstone. It is generally agreed that the bauxite deposits/laterites overlying the alkali basalt and tholeiite were derived from weathering of underlying basalt, however, the origin of bauxite deposit overlying the sandstone remains controversial. Chondrite-normalized REE patterns show that the bauxite ores/laterites overlying the sandstone exhibit quite similar chondrite-normalized REE patterns to those overlying the alkali basalt. Diagram of Al-Ti-Zr shows that the bauxite ores/laterites overlying the sandstone, tholeiite and alkali basalt are close to each other and to the calc-alkaline suite, however, significantly different from the sandstone and shales. Binary diagram of log Nb/Y vs log Zr/Ti further indicates that the parent rocks of bauxite ores/laterites overlying the sandstone belong to the suit of ultra-alkali to alkali basalt. Multivariate statistical analysis of geochemical data exhibit that the geochemical characteristics of HREE, Y, LREE and Al2O3 for bauxite ores/laterites overlying the sandstone are similar to those overlying the alkali basalt (15.7 Ma), obviously different from those overlying the tholeiite (1.2 to 0.5 Ma). Consequently, it can be inferred that the bauxite deposits/laterites overlying the sandstone were derived from the alkali basalt.

EXPECTATIONS AND PERFORMANCE IN THE BRAZILIAN NORTHEAST CREDIT UNION

Credit cooperativism is characterized by providing access to financial services and intermediation of funds, taking risks and generating benefits. In more mature economies works as a booster of strategic economic sectors, including occupying spaces not filled by banking institutions, in response to the worldwide phenomenon of concentration in the industry. This study aims to evaluate the quality of services provided by a cooperative credit northeast from the expectations of its members. From multivariate techniques and univariate statistical analysis of data, the search resulted in the identification of negative gaps between expectation and performance on the dimensions of the quality of services provided by the cooperative from the proposed model SERVQUAL scale.

Geochemical Soil Survey for Base and Precious Metals in Dagbala-Atte District, Southwestern Nigeria

Geochemical survey of residual soils was undertaken in Dagbala-Atte District of Igarra Schist Belt, southwestern Nigeria to explore for base and precious metals (Cu, Pb, Zn, Au and Ag). It involved sampling residual soil from 49 sites in a grid pattern; analyzing the soil samples for the base and precious metals and other commonly associated elements including As, Cd, Fe, Hg and Sb; subjecting the geochemical data generated to statistical analysis; and preparing the geochemical distribution map of the district for each element. On the basis of the multivariate statistical analysis (correlation matrix and factor analysis) of the soil geochemical data, three mineralization types were inferred to be present in the district. The suspected types of mineralization consist of: 1) Cu-Pb-Zn-Ag mineralization with associated As, Hg, Sb and Fe; 2) Au-Cu mineralization with associated As; and 3) minor Hg-bearing Au-Pb mineralization. With the aid of the elemental geochemical distribution maps, the suspected mineralization was linked to rock types from which the sampled residual soils were apparently derived. Hence, the suspected Cu-Pb-Zn-Ag mineralization is linked to the quartz-biotite schist at the southwestern part of the district while the suspected Au-Cu mineralization is located on the silicified sheared rock at the northern and southern parts and the minor Hg-bearing Au-Pb mineralization within the granitic gneiss at the northeastern and eastern areas of the district. Based on the foregoing findings, it is recommended that lithogeochemical survey for the base and precious metals should be done in the area.

Classification of Profit-Based Operating Regions for the Tennessee Eastman Process using Deep Learning Methods

Abstract The focus of this work is on the classification of an input-design space into different regions of input conditions that result in different corresponding ranges of productivity costs in the Tennessee Eastman Process (TEP). Although similar classification tasks had been previously carried out using linear multivariate statistical data analysis methods, these are of limited efficacy when dealing with highly non-linear dynamics. In this work, we present two Deep Learning Tools for classification: using either supervised learning or un-supervised learning. For classification with supervised learning a Recurrent Neural Network (RNN) known as Long Short-Term Memory (LSTM) is trained on normalized training data. Since deep learning networks generally involve a large number of nodes and parameters an algorithm named Sequential Layer-wise Relevance Propagation (SLRPFP) is proposed for selecting the relevant inputs and for pruning the LSTM network such that the test accuracy at each step is maintained or even improved. For classification with unsupervised learning, main features from the input dataset are first extracted using an Autoencoder. Then a Multi-dimensional Support Vector Machines (MSVM) model is applied to the features identified by the autoencoder. The performance of the proposed supervised and unsupervised deep learning approaches are compared to an approach that combines linear Dynamic Principal Component Analysis (DPCA) and a MSVM based classification and conclusions are drawn on the relative advantages of the deep learning methods.

Spatial partial least squares autoregression: Algorithm and applications

Partial least squares (PLS) is an efficient multivariate statistical data analysis method and has shown to be particularly useful to deal with knotty problems in practical applications, e.g., the ill-conditioned regression issues where the predictor variables are highly collinear or the sample size is much smaller than the number of covariates. Various works on PLS and its extension have been developed under the assumption that all the observations are mutually independent. However, this independence assumption may be violated in practice, especially when we collect data with network dependence structure among observations from some scientific disciplines, such as sociology and spatial economics. Yet relatively few works are available for PLS with network dependence structure. Here, we propose an spatial partial least squares autoregression (SPLSAR) to solve this problem, incorporating a spatial autoregressive parameter and a spatial weight matrix into PLS to accommodate dependence structure between individuals. Thus, the proposed method is more flexible as it can not only inherit the advantages of PLS, but also have the ability to deal with the network dependence. Moreover, we propose an efficient algorithm for the implementation of parameter estimation in SPLSAR regression. The finite sample performance of the proposed method is evaluated via extensive simulation study and a real data analysis, where results show that SPLSAR can achieve a better and more robust performance in terms of parameter estimation precision and out-of-sample prediction accuracy, when compared to classical PLS and SAR.

Dispersive liquid-liquid microextraction combined with gas chromatography–mass spectrometry for in situ determination of biogenic amines in meat: Estimation of meat's freshness

A dispersive liquid-liquid microextraction (DLLME) gas chromatography–mass spectrometry (GC–MS) technique was developed for the determination of selected biogenic amines (BAs) in samples of poultry, pork and beef. Prior to the extraction process, an appropriate volume of sodium hydroxide solution was added to each of the portioned samples. Next, samples were homogenized, centrifuged and finally sonicated at an increased temperature. After another centrifugation, the supernatant was made up to 50 mL in a calibrated flask. Subsequently, 5 mL of supernatant was separately subjected to a derivatization and extraction procedure. A mixture of methanol (dispersive solvent; 210 μL), chloroform (extractive solvent; 300 μL), and isobutyl chloroformate (derivatizing reagent; 100 μL) was used in the extraction process together with an admixture of pyridine and HCl in order to eliminate the by-products. The application of the method enables fast derivatization and extraction of the BAs and a straightforward and rapid sample enrichment. It displayed good linearity, intra- and inter-day precision and good recoveries. The proposed methodology is characterized by low limits of detection and quantification (0.003–0.009 μg/g and 0.009–0.029 μg/g, respectively). The green character of the method was established based on the results of two tools, namely the Analytical Eco-Scale and GAPI. It was successfully used to analyse samples of poultry, porcine and bovine meat. Multivariate statistical data analysis was applied in order to evaluate the potential use of the determined BAs as spoilage markers of particular meat types.

Geographical discrimination of red garlic (Allium sativum L.) produced in Italy by means of multivariate statistical analysis of ICP-OES data

Sixty-five samples of red garlic (Allium sativum L.) coming from four different production territories of Italy were analysed by means of inductively coupled plasma optical emission spectrometry. The garlic samples were discriminated according to the geographical origin using the content of seven elements (Ba, Ca, Fe, Mg, Mn, Na and Sr). Both classification and class modelling methods by using linear discriminant analysis (LDA) and soft independent model class analogy (SIMCA), respectively, were applied. Classification ability and modelling efficiency were evaluated on an external prediction set (21 garlic samples) designed by application of duplex Kennard-Stone algorithm. All the calibration and prediction samples were correctly classified by means of LDA. The class models developed using SIMCA exhibited high sensitivity (almost all the calibration and external samples were accepted by the respective classes) and good specificity (the majority of extraneous samples were refused by each class model).