Chemometric Data Analysis Research Articles

An untargeted lipidomic strategy combining comprehensive two-dimensional liquid chromatography and chemometric analysis

Untargeted lipidomic samples are extremely complex and often exceed the limits of peak capacity achievable by one-dimensional liquid chromatography (LC). Comprehensive two-dimensional liquid chromatography (LC × LC) appears as a promising alternative to overcome this drawback. Unfortunately, this approach generates highly complex datasets which untargeted analysis is challenging. In this work, a global methodological strategy combining LC × LC–MS/MS with chemometric data analysis is proposed for untargeted lipidomic studies. The feasibility of the proposed methodology is demonstrated by its application to assess the effects of arsenic exposure on the lipidome of growing rice samples. A two-dimensional chromatographic setup coupling reversed phase (RP) and hydrophilic interaction liquid chromatography (HILIC) modes together with a triple quadrupole mass detector (TQD) is proposed to analyze lipid extracts from rice samples at different experimental conditions. Chemometric tools were used for data compression, spectral and elution profiles resolution, feature detection and statistical analysis of the multidimensional LC × LC–MS/MS data. The obtained results revealed that the proposed methodology was useful to gather relevant information from untargeted lipidomic studies and detect potential biomarkers.

A preliminary study on the chemical characteristics of Kei apple (Dovyalis caffra), an undervalued South African fruit

Although a profusion of indigenous fruits occur in South Africa, knowledge of their uses lies mainly with the local people. This study was aimed at filling the void in scientific literature by investigating the nutritional value and characteristics of a lesser known wild fruit, the Kei apple (Dovyalis caffra). The colour was assessed, in addition to the chemical characteristics (total soluble phenolics, protein, volatiles, antioxidants) of the fruits. The results indicate that the fruit is a rich source of amino acids with good antioxidant properties. Although, the protein content was found to be lower than those of commercial fruits, it was similar to values reported for the wild plum. Chemometric analysis of chromatographic data on secondary metabolites produced by the fruit was used to determine if the collection of Kei apple trees established at the Agricultural Research Council's experimental farm in Mbombela is representative of the wild population of South Africa.

Effect of a grapevine-shoot waste extract on red wine aromatic properties.

The use of a grapevine-shoot extract (VIN) is being studied as an alternative to sulfur dioxide (SO2 ). VIN stabilizes anthocyanins and preserves polyphenolic compounds, and thus improves chromatic wine properties. In this study, selected aroma compounds (esters, C13 -norisoprenoids, oxidation and vine-shoot-related compounds), sensory analysis and the olfactometric profile were determined in the wines treated with VIN at two concentrations. Treatment with VIN hardly modified the content of esters and oxidation-related compounds in the wines. However, the high β-damascenone and isoeugenol contents and the increase in astringency at tasting in VIN wines were noteworthy, as were some odorant zones. All these were established as VIN markers after the chemometric data analysis. These data revealed that only the lowest dose tested may be recommended as a suitable alternative to SO2 . Although some aromatic properties of these wines may change, these changes are not considered to affect the quality of the wines negatively. These results are useful for wineries, which face having to discover the aroma-related processes in the challenge of producing SO2 -free wines without detriment to their sensory properties. © 2018 Society of Chemical Industry.

Classification of organic olives based on chemometric analysis of elemental data

The aim of this study was to discriminate organic from conventional olive samples based on the levels of macro and trace elements, combined with chemometric techniques. Ten elements (Na, K, Ca, Fe, Mg, Cu, Zn, Se, S and P) were determined in organic (n = 30) and conventional (n = 30) olive samples by inductively coupled plasma optical emission spectrometry analysis (ICP-OES). The classification of samples was performed by using a well-known chemometric techniques, linear discriminant analysis (LDA), partial least square-discriminant analysis (PLS-DA), support vector machine-discriminant analysis (SVM-DA), k-nearest neighbors (k-NN) and random forest (RF). The k-NN technique showed the best performance in discriminating organic from conventional samples (Accuracy: 94%) using all chemical variables. After variable reduction, an accuracy of 83% was found by using only the elements K and P. The use of a fingerprint based on multielemental levels associated with classification chemometric techniques may be used as a simple method to authenticate organic olive samples.

1H-NMR and UPLC-MS metabolomics: Functional tools for exploring chemotypic variation in Sceletium tortuosum from two provinces in South Africa

Sceletium tortuosum (Aizoaceae) is widely recognised for the treatment of stress, anxiety and depression, as well as for obsessive compulsive disorders. A comprehensive intraspecies chemotypic variation study, using samples harvested from two distinct regions of South Africa, was done using both proton nuclear magnetic resonance (1H-NMR) spectroscopy of methanol extracts (N = 145) and ultra performance liquid chromatography-mass spectrometry (UPLC-MS) of acid/base extracts (N = 289). Chemometric analysis of the 1H-NMR data indicated two main clusters that were region-specific (Northern Cape and Western Cape provinces). Two dimensional (2D) NMR was used to identify analytes that contributed to the clustering as revealed by the S-plot. The sceletium alkaloids, pinitol and two alkylamines, herein reported for the first time from S. tortuosum, were identified as markers that distinguished the two groups. Relative quantification of the marker analytes conducted by qNMR indicated that samples from the Northern Cape generally contained higher concentrations of all the markers than samples from the Western Cape. Quantitative analysis of the four mesembrine alkaloids using a validated UPLC-photo diode array (PDA) detection method confirmed the results of qNMR with regard to the total alkaloid concentrations. Samples from the Northern Cape Province were found to contain, on average, very high total alkaloids, ranging from 4938.0 to 9376.8 mg/kg dry w. Regarding the Western Cape samples, the total yields of the four mesembrine alkaloids were substantially lower (averages 16.4–4143.2 mg/kg). Hierarchical cluster analysis of the UPLC-MS data, based on the alkaloid chemistry, revealed three branches, with one branch comprising samples primarily from the Northern Cape that seemed somewhat chemically conserved, while the other two branches represented mainly samples from the Western Cape. The construction of an orthogonal projections to latent structures-discriminant analysis model and subsequent loadings plot, allowed alkaloid markers to be identified for each cluster. The diverse sceletium alkaloid chemistry of samples from the three clusters may facilitate the recognition of alkaloid profiles, rather than individual compounds, that exert targeted effects on various brain receptors involved in stress, anxiety or depression.

Discrimination of uranium ore concentrates by chemometric data analysis to support provenance assessment for nuclear forensics applications

This work describes a method for the discrimination of uranium ore concentrates (UOCs) to support provenance assessment for nuclear forensics applications using samples representing twenty producers from around the world. The concentrations were measured using inductively coupled plasma mass spectrometry. UOCs were classified using the support vector machine method relying on 61 down to only 18 element concentrations without affecting the accuracy. New features are calculated from combination of elements from the selected 18 elements, and added to the selected elements improve the classification results. Reducing the number leads to the optimization of laboratory measurements of element signatures in support of nuclear safeguards and forensics applications.

Chemometric Strategies for Peak Detection and Profiling from Multidimensional Chromatography

The increasing complexity of omics research has encouraged the development of new instrumental technologies able to deal with these challenging samples. In this way, the rise of multidimensional separations should be highlighted due to the massive amounts of information that provide with an enhanced analyte determination. Both proteomics and metabolomics benefit from this higher separation capacity achieved when different chromatographic dimensions are combined, either in LC or GC. However, this vast quantity of experimental information requires the application of chemometric data analysis strategies to retrieve this hidden knowledge, especially in the case of nontargeted studies. In this work, the most common chemometric tools and approaches for the analysis of this multidimensional chromatographic data are reviewed. First, different options for data preprocessing and enhancement of the instrumental signal are introduced. Next, the most used chemometric methods for the detection of chromatographic peaks and the resolution of chromatographic and spectral contributions (profiling) are presented. The description of these data analysis approaches is complemented with enlightening examples from omics fields that demonstrate the exceptional potential of the combination of multidimensional separation techniques and chemometric tools of data analysis.

Multivariate patent analysis—Using chemometrics to analyze collections of chemical and pharmaceutical patents

AbstractPatents are an important source of technological knowledge, but the amount of existing patents is vast and quickly growing. This makes development of tools and methodologies for quickly revealing patterns in patent collections important. In this paper, we describe how structured chemometric principles of multivariate data analysis can be applied in the context of text analysis in a novel combination with common machine learning preprocessing methodologies. We demonstrate our methodology in 2 case studies. Using principal component analysis (PCA) on a collection of 12338 patent abstracts from 25 companies in big pharma revealed sub‐fields which the companies are active in. Using PCA on a smaller collection of patents retrieved by searching for a specific term proved useful to quickly understand how patent classifications relate to the search term. By using orthogonal projections to latent structures (O‐PLS) on patent classification schemes, we were able to separate patents on a more detailed level than using PCA. Lastly, we performed multi‐block modeling using OnPLS on bag‐of‐words representations of abstracts, claims, and detailed descriptions, respectively, showing that semantic variation relating to patent classification is consistent across multiple text blocks, represented as globally joint variation. We conclude that using machine learning to transform unstructured data into structured data provide a good preprocessing tool for subsequent chemometric multivariate data analysis and provides an easily interpretable and novel workflow to understand large collections of patents. We demonstrate this on collections of chemical and pharmaceutical patents.

Study on sterol fraction of commercial brands of milk, yoghurt and butter sold in Iranian market and chemometric data analysis

A gas chromatographic procedure for analysis of sterols was developed to detect possible addition of vegetable oils to milk, yoghurt and butter samples. The method involved fat extraction, saponification, hexane extraction, silylation of sterols and then analysis by gas chromatography. The sterol content of 102 samples of milk, yoghurt and butter was determined. Phytosterol content higher than 5% of total sterol was observed in four samples of milk and yoghurt. Principal component analysis showed that there was a direct relationship between the level of phytosterols and fat percentage in yoghurt samples. The described method is accurate and reliable enough to be employed regularly in food quality laboratories.

Rapid quality control of Sutherlandia frutescens leaf material through the quantification of SU1 using vibrational spectroscopy in conjunction with chemometric data analysis

Sutherlandia frutescens, commonly known as cancer bush, is a popular herbal remedy which has been used traditionally for the treatment of internal cancers, stress, influenza, fever, diabetes, stomach ailments and as a tonic. SU1 (syn = sutherlandioside B) is a major cycloartane-type triterpene glycoside in the leaf material and it has been used as a biomarker for quality control. Vibrational spectroscopy in combination with chemometric data analysis has been successfully used to quantify target compounds in various materials and is considered a green alternative method to traditional analytical methods. In this study, the variation in SU1 content in S. frutescens wild-type samples was investigated using ultra high performance liquid chromatography coupled to mass spectrometery (UHPLC–MS). This UHPLC–MS reference dataset was used in conjunction with vibrational spectroscopy and chemometric modeling to develop a rapid alternative method to quantify SU1 in powdered leaf material. Partial least squares (PLS) regression models based on near- and mid-infrared spectroscopy data were developed using SIMCA-P+ 13.0 chemometric software and validated by inserting external prediction sets into the model. The concentration of SU1 showed high variation ranging from 0.04 to 10.10 μg/mg of dried leaf material as determined using UHPLC–MS. Calibration models with good model statistics were developed and used to predict the content of the major compound SU1. External datasets introduced into the model resulted in the accurate prediction of the SU1 content. NIR and MIR spectroscopy in combination with chemometric data analysis were shown to be viable rapid alternative quantification methods for SU1 in powdered leaf material.

Assessment of endocrine disruptors effects on zebrafish (Danio rerio) embryos by untargeted LC-HRMS metabolomic analysis

Bisphenol A (BPA), perfluorooctane sulfonate (PFOS), and tributyltin (TBT) are emerging endocrine disruptors (EDCs) with still poorly defined mechanisms of toxicity and metabolic effects in aquatic organisms. We used an untargeted liquid chromatography-high resolution mass spectrometry (LC-HRMS) metabolomic approach to study the effects of sub-lethal doses of these three EDCs on the metabolic profiles of zebrafish embryos exposed from 48 to 120hpf (hours post fertilization). Advanced chemometric data analysis methods were used to reveal effects on the subjacent regulatory pathways. EDC treatments induced changes in concentrations of about 50 metabolites for TBT and BPA, and of 25 metabolites for PFOS. The analysis of the corresponding metabolic changes suggested the presence of similar underlying zebrafish responses to BPA, TBT and PFOS affecting the metabolism of glycerophospholipids, amino acids, purines and 2-oxocarboxylic acids. We related the changes in glycerophospholipid metabolism to alterations in absorption of the yolk sack, the main source of nutrients (including lipids) for the developing embryo, linking the molecular markers with adverse phenotypic effects. We propose a general mode of action for all three chemical compounds, probably related to their already described interaction with the PPAR/RXR complex, combined with specific effects on different signaling pathways resulting in particular alterations in the zebrafish embryos metabolism.

A non-targeted metabolomic approach to identify food markers to support discrimination between organic and conventional tomato crops

In the last decade, the consumption trend of organic food has increased dramatically worldwide. However, the lack of reliable chemical markers to discriminate between organic and conventional products makes this market susceptible to food fraud in products labeled as “organic”. Metabolomic fingerprinting approach has been demonstrated as the best option for a full characterization of metabolome occurring in plants, since their pattern may reflect the impact of both endogenous and exogenous factors. In the present study, advanced technologies based on high performance liquid chromatography-high-resolution accurate mass spectrometry (HPLC-HRAMS) has been used for marker search in organic and conventional tomatoes grown in greenhouse under controlled agronomic conditions. The screening of unknown compounds comprised the retrospective analysis of all tomato samples throughout the studied period and data processing using databases (mzCloud, ChemSpider and PubChem). In addition, stable nitrogen isotope analysis (δ15N) was assessed as a possible indicator to support discrimination between both production systems using crop/fertilizer correlations. Pesticide residue analyses were also applied as a well-established way to evaluate the organic production. Finally, the evaluation by combined chemometric analysis of high-resolution accurate mass spectrometry (HRAMS) and δ15N data provided a robust classification model in accordance with the agricultural practices. Principal component analysis (PCA) showed a sample clustering according to farming systems and significant differences in the sample profile was observed for six bioactive components (L-tyrosyl-L-isoleucyl-L-threonyl-L-threonine, trilobatin, phloridzin, tomatine, phloretin and echinenone).

Application of SWIR hyperspectral imaging and chemometrics for identification of aflatoxin B1 contaminated maize kernels

A short-wave infrared (SWIR) hyperspectral imaging system (1000–2500 nm) combined with chemometric data analysis was used to detect aflatoxin B1 (AFB1) on surfaces of 600 kernels of four yellow maize varieties from different States of the USA (Georgia, Illinois, Indiana and Nebraska). For each variety, four AFB1 solutions (10, 20, 100 and 500 ppb) were artificially deposited on kernels and a control group was generated from kernels treated with methanol solution. Principal component analysis (PCA), partial least squares discriminant analysis (PLSDA) and factorial discriminant analysis (FDA) were applied to explore and classify maize kernels according to AFB1 contamination. PCA results revealed partial separation of control kernels from AFB1 contaminated kernels for each variety while no pattern of separation was observed among pooled samples. A combination of standard normal variate and first derivative pre-treatments produced the best PLSDA classification model with accuracy of 100% and 96% in calibration and validation, respectively, from Illinois variety. The best AFB1 classification results came from FDA on raw spectra with accuracy of 100% in calibration and validation for Illinois and Nebraska varieties. However, for both PLSDA and FDA models, poor AFB1 classification results were obtained for pooled samples relative to individual varieties. SWIR spectra combined with chemometrics and spectra pre-treatments showed the possibility of detecting maize kernels of different varieties coated with AFB1. The study further suggests that increase of maize kernel constituents like water, protein, starch and lipid in a pooled sample may have influence on detection accuracy of AFB1 contamination.

Study of characteristic ions in frequently-used traditional Chinese medicinal materials derived from animal bile and its application

A LC-MS method was established for study of characteristic ions in frequently-used traditional Chinese medicinal materials derived from animal bile. UPLC-Q-TOF was used in the data acquisition work, then, the software of MarkerLynxTM v.4.1 was performed in the chemometric analysis of data. Besides, selected ion chromatograms of these bile acid ions were comparative studied. Better results were gained in the specificity identification of pig bile, bear bile, cultivated cow-bezoar in vitro, artificial cow-bezoar and some compound preparations(Hugan tablet and Rengongniuhuang Jiaxiaozuo capsule) by this method. The method is suitable for the specificity identification of pig bile, bear bile, cultivated cow-bezoar in vitro, artificial cow-bezoar and compound preparations containing these medicinal materials.

Investigation of the photodegradation profile of tamoxifen using spectroscopic and chromatographic analysis and multivariate curve resolution

Photodegradation of tamoxifen (TAM) is investigated by chemometric analysis of the multiset data obtained by LC-DAD/MS and UV spectrophotometry. A hydroalcoholic solution of TAM was submitted to photodegradation by means of a dedicate irradiation cabinet able to simulate natural irradiation sources. The irradiance conditions were stressed by increasing the irradiation power to produce a rapid photodegradation of TAM. Drug photodegradation was monitored through UV spectrophotometry and the obtained photoproducts were investigated in detail by DAD/MS liquid chromatography. Data collected from the combination of both instrumental techniques were fused and processed jointly using the Multivariate Curve Resolution–Alternating Least Squares (MCR-ALS) method. A total number of five compounds were identified during the drug photodegradation and their kinetic evolution was described. The process included the isomerization of TAM to its (E)-form and the subsequent cyclization of these both compounds to give two phenanthrene derivatives. A photooxygenation process can also occur, giving a benzophenone derivative photoproduct as a result thereof. The multivariate resolution method proposed in this work allowed the resolution of this complex multicomponent system by the direct analysis of the experimental data.

Comparative studies on conventional and solvent-free synthesis toward hydrazones: application of PXRD and chemometric data analysis in mechanochemical reaction monitoring

We describe the implementation of chemometric analysis for mechanochemical synthesis monitoring.

A chemometric approach for characterization of serum transthyretin in familial amyloidotic polyneuropathy type I (FAP-I) by electrospray ionization-ion mobility mass spectrometry

In this study, we describe a chemometric data analysis approach to assist in the interpretation of the complex datasets from the analysis of high-molecular mass oligomeric proteins by ion mobility mass spectrometry (IM-MS). The homotetrameric protein transthyretin (TTR) is involved in familial amyloidotic polyneuropathy type I (FAP-I). FAP-I is associated with a specific TTR mutant variant (TTR(Met30)) that can be easily detected analyzing the monomeric forms of the mutant protein. However, the mechanism of protein misfolding and aggregation onset, which could be triggered by structural changes in the native tetrameric protein, remains under investigation. Serum TTR from healthy controls and FAP-I patients was purified under non-denaturing conditions by conventional immunoprecipitation in solution and analyzed by IM-MS. IM-MS allowed separation and characterization of several tetrameric, trimeric and dimeric TTR gas ions due to their differential drift time. After an appropriate data pre-processing, multivariate curve resolution alternating least squares (MCR-ALS) was applied to the complex datasets. A group of seven independent components being characterized by their ion mobility profiles and mass spectra were resolved to explain the observed data variance in control and patient samples. Then, principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were considered for exploration and classification. Only four out of the seven resolved components were enough for an accurate differentiation. Furthermore, the specific TTR ions identified in the mass spectra of these components and the resolved ion mobility profiles provided a straightforward insight into the most relevant oligomeric TTR proteoforms for the disease.

Protolytic Properties of Water-Soluble Magnesium 5,10,15,20-Tetra(1-methylpyridin-1-ium-4-yl)porphyrinate Tetraperchlorate MgTMPyP4

Protolytic properties of porphyrin complexes with Mg, analogs of the natural porphyrin chlorophyll, were investigated by the UV spectroscopy method with simultaneous potentiometric determination of pH of the medium with the aim to detect metalloporphyrin forms arising in a solution with changing its acidity. To consider interconversions in solutions, MCR-ALS soft modelling based on the chemometric analysis of spectrometric data was applied.

Accurate anisotropy recovery from fluorophore mixtures using Multivariate Curve Resolution (MCR)

Anisotropy resolved multidimensional emission spectroscopy (ARMES) provides valuable insights into multi-fluorophore systems like proteins that have complex overlapping emission bands. The method combines multidimensional fluorescence, anisotropy, and chemometrics to facilitate the differentiation of fluorophores with very similar emission properties. Here, we address the critical issue of standardizing the chemometric methods required to accurately extract spectral and anisotropy information from fluorophore mixtures using two standard sample sets: perylene in glycerol, and a mixture of Erythrosin B and Phloxine B with overlapping emission but different anisotropies. We show for the first time how to accurately model component anisotropy using Multivariate Curve Resolution (MCR) from data collected using total synchronous fluorescence scan (TSFS) and Excitation Emission Matrix (EEM) measurement methods.These datasets were selected to avoid the presence of inner filter effects (IFE) or Förster resonance energy transfer (FRET) that would depolarize fluorescence emission or reduce data tri-linearity. This allowed the non-trilinear TSFS data to yield accurate component anisotropy data once modelled using the correct data augmentation strategy, however, the EEM data proved to be more accurate once optimal constraints (non-negativity and correspondence among species) were employed. For perylene (S2) and Phloxine B which both have very weak anisotropy (<0.06), while the spectral recovery was excellent, the modelled anisotropy values were reasonably accurate (±20% of the real value) because of large relative noise contributions. However, for perylene (S1) and Erythrosin B which have large (>0.2) anisotropies, bilinear and trilinear EEM models built using a total tri-linearity constraint, yielded solutions without any rotational ambiguities and very accurate (±4% of real value) anisotropy values. These sample systems thus provide simple and robust test systems for validating the spectral measurement and chemometric data analysis elements of ARMES.

Discrimination of CRISPR/Cas9-induced mutants of rice seeds using near-infrared hyperspectral imaging

Identifying individuals with target mutant phenotypes is a significant procedure in mutant exploitation for implementing genome editing technology in a crop breeding programme. In the present study, a rapid and non-invasive method was proposed to identify CRISPR/Cas9-induced rice mutants from their acceptor lines (huaidao-1 and nanjing46) using hyperspectral imaging in the near-infrared (NIR) range (874.41–1733.91 nm) combined with chemometric analysis. The hyperspectral imaging data were analysed using principal component analysis (PCA) for exploratory purposes, and a support vector machine (SVM) and an extreme learning machine (ELM) were applied to build discrimination models for classification. Meanwhile, PCA loadings and a successive projections algorithm (SPA) were used for extracting optimal spectral wavelengths. The SVM-SPA model achieved best performance, with classification accuracies of 93% and 92.75% being observed for calibration and prediction sets for huaidao-1 and 91.25% and 89.50% for nanjing46, respectively. Furthermore, the classification of mutant seeds was visualized on prediction maps by predicting the features of each pixel on individual hyperspectral images based on the SPA-SVM model. The above results indicated that NIR hyperspectral imaging together with chemometric data analysis could be a reliable tool for identifying CRISPR/Cas9-induced rice mutants, which would help to accelerate selection and crop breeding processes.