Pubblicazioni Aperte DIgitali Sapienza >
CHIMICA ANALITICA E DEI SISTEMI REALI >
Please use this identifier to cite or link to this item:
|Title: ||FOOD QUALITY CONTROL AND AUTHENTICATION THROUGH COUPLING CHEMOMETRICS TO INSTRUMENTAL FINGERPRINTING TECHNIQUES|
|Authors: ||NESCATELLI, RICCARDO|
|Tutor: ||BUCCI, REMO|
|Issue Date: ||20-Dec-2013|
|Abstract: ||Given the emphasis that, especially in the last years, has been placed on food quality and safety, the research which constitutes the basis of the present doctoral thesis was focused on the development of analytical methods to verify the authenticity of foods and for their traceability.
In this context, attention was mainly posed on the verification of two closely related aspects:
i) the chemical characterization of foods, in terms of the assessment of their composition and the determination of their constituents;
ii) the identification of the origin of food.
To this purpose, on one hand, analytical methods for the determination of biologically, nutritionally and commercially relevant compounds present in different foods were optimized and validated. In particular, a spectroscopic method based on NIR spectroscopy for the determination of some of the indices (water, sugars and hydroxy methylfurfural) required by law to control the quality of honey was developed. An innovative method based on microwave-assisted extraction and subsequent HPLC chromatographic analysis for the determination of three analytes which are considered as quality indices (crocin, picrocrocin and safranal) of saffron was also developed. In collaboration with the University of Santiago de Compostela, where I spent three months for a research stay in the framework of my PhD, the simultaneous acetylation and dispersive liquid-liquid micro-extraction (DLLME) combined with gas chromatography-mass spectrometry (GC-MS) was proposed, for the first time, for the determination of benzotriazoles in water samples. Indeed, although not considered to be food, water, is clearly essential for human life and can lead to a primary contamination of foods. In all cases, the analytical methods represented an improvement with respect to the traditional ones, reducing costs and time of analysis and also better meeting the legal requirements; in parallel, great attention was posed to environmental issues, eliminating or minimizing the use of toxic and hazardous solvents, and sample pretreatment.
On the other hand, another relevant part of this thesis was dedicated to the development of analytical protocols for the authentication of the origin of foods. In this framework, a chromatographic method for tracing high quality extra virgin olive oil was optimized and validated, in particular in order to identify and discriminate Sabina PDO extra virgin olive oil from oils of other origins. Similarly, methods for checking the source of two other high value-added food products, honey and saffron, were developed. In the case of honey, both the geographical and botanical origins were considered, while for saffron together with the characterization of geographical origin, differences in the cultivation and production processes were also taken into account.
In the following, the main achievements of the PhD research will be illustrated: for the sake of a better understanding, the presentation of the results, both in the present abstract and in the thesis, has been organized in sections corresponding to the different food authentication problems addressed.
As already mentioned, the possibility of developing a model for tracing PDO Sabina extra virgin olive oils was investigated. To this purpose, an analytical strategy based on coupling chromatographic profiling of the phenolic fraction to chemometric classification tools was successfully adopted. At first, the extraction procedure from isolating phenols from the oil matrix was optimized in terms of efficiency, time and cost by means of an experimental design. Then, for each sample, chromatographic profiles at three different wavelengths were recorded (254, 280 and 340 nm), and each chromatogram was considered as the fingerprint of olive oil. Accordingly, the chromatographic profiles of the extracts, after baseline correction, alignment and normalization, were then used as data to build chemometric classification models aimed at authenticating the origin of the samples by Partial Least Square Discriminant Analysis (PLS-DA). In a first stage, whole profiles recorded at a particular wavelengths were individually processed but none of the resulting models allowed a good discrimination of the samples. Therefore, also to account for the fact that classification model can be affected by the presence of a high number of irrelevant (i.e. carrying information not linked to the problem at hand) predictors, variable selection through a sequential application of backward interval-Partial Least Squares (biPLS) and genetic algorithms (GA) was carried out to improve the discrimination ability of the models. Indeed, models built on selected regions of the chromatograms allowed the correct prediction of around 80% of validation samples, the better outcomes having been obtained using either the signals at 280 or at 340 nm.
Lastly, since each sample was characterized by three signals (the chromatographic profiles – or selected regions from those – at three different wavelengths), an attempt to further improving the classification ability by integrating the information from the three chromatograms into a single model through mid-level data fusion protocol was tempted. Indeed, the model built on the matrix obtained by concatenating the regions of the chromatograms at 280 and 340 nm, selected by biPLS-GA, allowed to correctlt predict the origin of about 90% of the validation samples: specifically, 6 out of 7 samples from Sabina PDO and 18 out of 20 samples from other geographical origins.
Interpretation of the chromatographic regions selected by biPLS-GA in terms of chemical compounds which could be used as traceability markers for the PDO Sabina was made on the basis of HPLC/ESI-MS analysis.
Another relevant part of this PhD research was devoted to the development of analytical strategies to characterize honey samples both for quality control and for traceability. In particular, the possibility of authenticating both the geographical and the botanical origin of honey samples was addressed by two different approaches: one based on HPLC-DAD analysis of the polyphenol fraction of honey, just as described in the case of olive oils, as the phenolic composition should be strongly linked to the geographical and floral origin, floral and to the climatic characteristics of local productions, and the other based on NIR spectroscopy, a fingerprinting technique which allows analyzing samples without any preparation, and which is rapid and relatively cheap. 70 honey samples from 7 different floral origins, all coming from the same geographical area, were considered in the botanical classification study. On the other hand, 204 wildflower samples coming from different regions and nations were investigated for the geographical classification study (all 204 were measured by NIR spectroscopy, while only 104 of those could be also analyzed by HPLC-DAD).
While, as mentioned, NIR spectroscopy didn’t require any sample pretreatment, extraction of the phenolic fraction from honey was required in the case of the chromatographic analysis. To this purpose, solid phase extraction (SPE) was used and the optimal experimental conditions were chosen on the basis of an experimental design.
As far as the HPLC analysis is concerned, chromatographic signals were recorded at 254, 280 and 340 nm and classification models were built both on the individual profiles, after baseline correction, alignment and normalization, and on matrices resulting from the integration of the three signals through a data fusion strategy. As far as the floral origin is concerned, the best model resulted to be the one based on fusing the profiles at all the three different wavelengths, which allowed correctly accepting, for each class, more than 80% of the samples and to correctly reject at least 93.3% of the samples from other origins. The same approach was used to develop the method for the geographical discrimination of honeys. The chromatographic profiles of the 140 honeys coming from different geographical areas have been pretreated and subsequently used as data set for the construction of the PLS-DA models. Specifically, since samples from two different production years were available, those from 2011 were used for model building while the 2012 ones constituted the validation set. Also in this case, the best model has been obtained by fusion of the chromatographic profiles recorded at the three wavelengths. This model was able, on average, to correctly classify 87% of the samples as belonging to their respective categories, at the same time rejecting as not belonging to the classes 97% of samples from other categories. Both for the floral and for the geographical traceability models, it was also possible to evaluate which compounds are mostly responsible of the discrimination of the categories of interest.
As anticipated, the same study has also been carried bout using NIR spectroscopy as fingerprinting technique: indeed, since NIR is a rapid, cheap, and non-destructive technique, which has the additional advantage of not requiring any sample pretreatment, the possibility of its use for honey authentication would perfectly fit in the framework of green analytical chemistry. To this purpose, 4 spectra have been collected for each sample in reflectance mode and the corresponding signals, after suitable preprocessing (baseline and scatter correction through the use of derivatives and SNV transform), have been used as dataset for PLS-DA analysis. Very good predictive ability, comparable to the one obtained by HPLC-DAD was achieved and chemical interpretation of the model by inspection of the VIP scores indicates that the spectral region which contribute the most are those which are attributable to the overtones of the stretching of CO, CH and OH.
Similarly, models for the geographical classification of honeys were built. In this case, the best model (which was the one pretreated by SNV only) allowed to discriminate very well the different sources with the exception of honeys from northern Italy and central France. In fact, excluding these two classes, sensitivity values ranging from 80.0% to 100.0% and specificity values ranging from 75.0% to 90.8% were obtained for the external validation samples (which corresponded, as already discussed in the case of HPLC-DAD to samples produced in a different year). Inspection of the VIP scores showed that the spectral regions carrying the highest discriminant information were those corresponding to CH, NH, OH, CO and NH stretching.
Apart from developing models for the geographical and botanical traceability of honey samples, the possibility of revising the methods for the determination of quality control parameters for this product exploiting the advantages of NIR spectroscopy was also investigated. Accordingly, NIR-based methods for the determination of water, reducing sugars and hydroxymethyl furfural (HMF) were designed and validated. Comparison of the outcomes with those obtained by reference methods showed that a comparable accuracy could be achieved.
Saffron is known to be the most expensive spice in the world, due to the limited cultivation and low production yields. Therefore, quality control of this fine product is essential in order to combat frauds. To date, ISO3632 norm of 2003 indicates the quality parameters for saffron and the procedures for their evaluation. In particular, three parameters that define the quality of saffron are crocin, picrocrocin and safranal, linked respectively to the color, taste and smell. As a part of this doctoral work, a method providing a fast, cheap and reliable analysis of crocin, picrocrocin and safranal in saffron has been developed. The method is based on microwave-assisted extraction of the analytes and subsequent HPLC analysis, using a limited volume of extracting solvent (ethanol/water) and a limited amount of sample. The developed method was validated in terms of linearity, limit of detection, limit of quantification, extraction efficiency, precision, and was compared with the ISO method and the newest methods proposed in the literature. The method developed was found to be very efficient and precise; in addition, it is characterized by a high concentration factor, allowing the analysis of the analytes even if present in minimal quantities. The limited quantity of sample and solvents required, as well as the reduced analysis times suggest that the method can be used routinely to ensure the quality of saffron samples. The proposed method was applied for the analysis of samples produced by different manufacturers and of different geographical origins, to determine their quality. In parallel, NIR spectroscopy was also used for the authentication of the origin of samples, with good results.
A part of PhD was carried out during a research stay in Santiago de Compostela, at the Instituto de Investigación y Análisis alimentarios. During this period, I contributed to the development of a method based on DLLME (dispersive liquid-liquid micro extraction) and subsequent GC-MS analysis, for the determination of benzotriazoles in different aqueous matrices. The importance of constructing a method for the analysis of these compounds derives from the fact that they are considered suspect human carcinogens, mutagens and toxic to some microorganisms and thus, new environmental pollutants. They were found in different aquatic systems, both indoors and even in human urine. The use of microextraction has numerous advantages: the use of minimal volumes of solvent allows to obtain high preconcentration of benzotriazoles and reduced time for the preparation of the sample. In parallel, gas chromatography, compared to liquid chromatography, allows to have an increase in the resolution of isomers. The purposes of this work were twofold: the development of a method for the preparation of the sample, based on the simultaneous derivatization-extraction DLLME, which could be simple, rapid, highly efficient, and with low environmental impact; its combination with a relatively inexpensive technique, such as GC-MS, which allows high resolution of the isomers and the determination of benzotriazoles in traces. The optimization of the extraction process was carried out following both uni-vand multi-variate procedures, based on the use of factorial experimental design. The optimized parameters were: the type and concentration of the base, the volume of acetic anhydride used for the acetylation, the timing of derivatization and extraction, the types and volumes of solvents used for the extraction (extractant and dispersant). The optimized method consists in the analysis of 10ml of water mixed with 1ml of Na2HPO4, a single step of derivatization and extraction with 100 uL of acetic anhydride, 1.5 ml of acetonitrile and 60μl of toluene. The times for the reaction and centrifugation (3000 rpm / min) are respectively 1 and 5 minutes. After the phase separation, 2μl are injected into the GC-MS system. The method was then validated in terms of linearity, extractive efficiency, accuracy, LOD and LOQ. Finally, it was applied for the analysis of wastewater samples (treated and untreated) and river waters. In summary, the procedure constitutes an interesting alternative to monitor the levels of various benzotriazoles during wastewater treatment and to investigate their fate in the aquatic environment.
In conclusion, new methods for analysis of foods were developed, based on the fruitful coupling of different instrumental profiling methods to chemometric data processing techniques, which allow reliable quality control and traceability of the origin of the product. In this respect, on one hand, chromatographic fingerprinting of the phenolic fraction proved to be a valid secondary traceability indicator for oil and honey samples.
On the other hand, thanks to the many benefits provided by NIR spectroscopy coupled with chemometric techniques, it was possible to build models of classification and regression which allowed to discriminate different samples, providing an additional tool to combat fraud. In this framework, particular attention was posed to the respect of the principles of "Green Chemistry", which has now become the focus of the chemistry of the new millennium. Indeed, the use of NIR spectroscopy allowed developing methods with less impact on the environment, humans and higher performance compared to existing methods for analysis of foods. But it is necessary to point out that, in all the examined cases, a key role is played by chemometrics. Indeed, the possibility of using a not selective fingerprinting technique such as NIR for calibration and classification, without needing any separation step or sample pretreatment is only made possible by the use of chemometric data processing which allow to mathematically manage the presence of interferents and other sources of unwanted variability in the signals. On the other hand, chemometric proved to be essential also for all the other studies presented in this PhD research, and ubiquitous in all the stages of the analytical process, starting from sampling strategies and experimental design to the final validation of the results obtained.|
|Research interests: ||analytical chemistry|
|Appears in PhD:||CHIMICA ANALITICA E DEI SISTEMI REALI|
Files in This Item:
|tesi dottorato Nescatelli.pdf||tesi di dottorato Nescatelli||5.29 MB||Adobe PDF|
File del Curriculum Vitae:
|CurriculumVitae.pdf|| ||213.29 kB||Adobe PDF|
Items in PADIS are protected by copyright, with all rights reserved, unless otherwise indicated.