You are on page 1of 23

View Online / Journal Homepage / Table of Contents for this issue

Chemical Society Reviews
www.rsc.org/chemsocrev
Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.rsc.org | doi:10.1039/C2CS35138B

Volume 41 | Number 17 | 7 September 2012 | Pages 5569–5868

ISSN 0306-0012

CRITICAL REVIEW David I. Ellis, Victoria L. Brewster, Warwick B. Dunn, J. William Allwood, Alexander P. Golovanov and Royston Goodacre Fingerprinting food: current technologies for the detection of food adulteration and contamination

Chem Soc Rev
Cite this: Chem. Soc. Rev., 2012, 41, 5706–5727 www.rsc.org/csr

Dynamic Article Links
View Online

CRITICAL REVIEW

Fingerprinting food: current technologies for the detection of food adulteration and contamination
David I. Ellis,*a Victoria L. Brewster,a Warwick B. Dunn,ab J. William Allwood,a Alexander P. Golovanovc and Royston Goodacreab
Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.rsc.org | doi:10.1039/C2CS35138B

Received 9th June 2011 DOI: 10.1039/c2cs35138b Major food adulteration and contamination events seem to occur with some regularity, such as the widely publicised adulteration of milk products with melamine and the recent microbial contamination of vegetables across Europe for example. With globalisation and rapid distribution systems, these can have international impacts with far-reaching and sometimes lethal consequences. These events, though potentially global in the modern era, are in fact far from contemporary, and deliberate adulteration of food products is probably as old as the food processing and production systems themselves. This review first introduces some background into these practices, both historically and contemporary, before introducing a range of the technologies currently available for the detection of food adulteration and contamination. These methods include the vibrational spectroscopies: near-infrared, mid-infrared, Raman; NMR spectroscopy, as well as a range of mass spectrometry (MS) techniques, amongst others. This subject area is particularly relevant at this time, as it not only concerns the continuous engagement with food adulterers, but also more recent issues such as food security, bioterrorism and climate change. It is hoped that this introductory overview acts as a springboard for researchers in science, technology, engineering, and industry, in this era of systems-level thinking and interdisciplinary approaches to new and contemporary problems.

Introduction
Major food adulteration/contamination events with international impacts and consequences, such as the milk scandal in China and the Irish pork crisis (both in 2008) and the resultant media attention, throw into sharp focus the requirement for continued vigilance, research and development of rapid analytical and detection techniques for food, feedstuffs,1 and food-related products2–4 and systems.5 The problem of food adulteration is by no means a contemporary phenomenon and is likely to be as old as the food processing and production systems themselves.6 In the ‘modern’ scientific era, the first to address this issue was the German analytical chemist Frederick Accum who completed a treatise on adulteration of food and culinary poisons published in 1820.7 This was the first serious attempt to expose both the extent and dangers of food adulteration.8 Table 1 lists some of the adulterants identified by Accum
a

School of Chemistry, Manchester Institute of Biotechnology, University of Manchester, 131 Princess Street, Manchester M1 7ND, UK. E-mail: D.Ellis@manchester.ac.uk b Manchester Centre for Integrative Systems Biology, Manchester Institute of Biotechnology, University of Manchester, 131 Princess Street, Manchester M1 7ND, UK c Faculty of Life Sciences, Manchester Institute of Biotechnology, University of Manchester, 131 Princess Street, Manchester M1 7ND, UK

and Hassall at this time, and some of these ‘additives’ are seriously toxic to humans, one of the more worrying ones was the practice of using red lead (Pb3O4) to give a vibrant colour to cayenne pepper. This seminal work was followed three decades later by further in-depth analyses of food adulterants in thousands of food and drink samples, initiated by the then editor of The Lancet, Thomas Wakley, undertaken by the physician Arthur Hill Hassall and published in The Lancet in the 1850s.8,9 This was subsequently followed by food regulation becoming part of UK law, such as the Adulteration Act of 1860 and the Sale of Food and Drugs Act 1875 (the basis for many modern food regulation Laws globally). By the end of that century however, it was still apparent that the practice of adulteration was rife and knew no international boundaries. In an article from the New York Times in 1900, the then Consul General to France, James T. Dubois, reported to the US State Department that: ‘‘in spite of the activity of the authorities in endeavouring to suppress the debasement of the food supply; the adulterators continue to grow even bolder, and extend their depredations with discouraging rapidity and success’’.10 The report went on to describe the adulteration of a wide range of products, such as chocolate, wine, butter, tea, cheese, coffee, and milk; whereby milk was diluted and then ‘certain kinds of soap’ used to restore its colour.
This journal is
c

5706

Chem. Soc. Rev., 2012, 41, 5706–5727

The Royal Society of Chemistry 2012

University of Manchester. and went on to obtain her MPhil in the application of Raman spectroscopic methods to forensic analysis. with a blatant disregard for the potential health consequences16 and resultant impacts on international trade.14 The contamination widened and spread to several other countries. William Allwood in the School of Chemistry at The University of Manchester. and secondly performing clinical metabolic profiling applications with Cancer Research UK.15 In addition to melamine adulteration of milk powder within China. a nitrogen-rich organic base normally used in plastic manufacture. Health and Outreach. much research effort has been focussed on the detection of melamine in a wide range of products using a wide variety of analytical techniques. biscuits in Switzerland.12 followed in December of the same year by the first reports in China of sick babies with discoloured urine (who had ingested deliberately adulterated milk). Warwick is an International Board member of the Metabolomics Society and is a member of the editorial board of the journal Metabolomics.net) and Douglas Kell’s labs (dbkgroup. chicken feed. baking powder in Malaysia.1039/C2CS35138B The World Health Organisation (WHO) called China’s international milk crisis: ‘‘one of the largest food safety events the UN health agency has had to deal with in recent years’’. with a further 860 infants hospitalised. Aberystwyth obtaining a BSc in Environmental Science and a PhD in Analytical Biotechnology/ Microbiology. including six child fatalities due to kidney damage. to raise nitrogen-based tests spuriously for protein content.20 in muscle tissue by LC–MS.org) in the School of Chemistry. It was reported that by the end of the year there were an estimated 300 000 victims in China.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. with the first child fatality occurring on 1 May 2008. through a BBSRC CASE PhD studentship in collaboration with Avacta Plc. Chem. His research involving the rapid and quantitative detection of foodborne bacteria using FT–IR and machine learning has been widely publicised. batches of eggs in Hong Kong. Warwick Dunn obtained degrees in Analytical Chemistry at The University of Hull and moved to The University of Manchester in 2003 to develop and apply bioanalytical tools in systems biology research. Victoria L. School of Chemistry. Manchester Institute of Biotechnology (MIB). During and following the scandal. UK. and many countries around the world recalled a range of potentially contaminated products.18 in animal feed by chromatography and tandem MS.. featuring on BBC TV and radio. 41. Will has since worked as a postdoctoral research associate for the Wheat Pathogenesis team at Rothamsted Research and for the last five years withJ. Victoria is developing spectroscopic methods for the characterisation of proteins at the Manchester Institute of Biotechnology. as well as in the David I. animal feedstuffs and other products had also regularly been adulterated for short-term economic gain. 5706–5727 5707 The Royal Society of Chemistry 2012 .org | doi:10. 2012. firstly on an EU funded project. where a number of food products tested positive for melamine. Rev. cakes and noodles by surface enhanced Raman David Ellis was educated on the Welsh coast at the University of Wales. This journal is c After having been schooled and worked in the picturesque South Shropshire countryside for 20 years. confectionary in New Zealand. Dunn investigate molecular processes related to mammalian diseases and in discovery and development of therapeutics. Warwick has the research objectives to Warwick B. He is employed at The School of Biomedicine as Lecturer in Applied Metabolomics and as the Metabolomics Lead in The Centre for Advanced Discovery and Experimental Therapeutics. UK. Currently. The University of Manchester.19 in cereal flours by HPLC. This involved melamine (Fig.rsc.23 in gluten. These include: in milk and milk products by chromatography17 and low temperature plasma probe and MS.11 The sequence of events started in May 2007 when a number of cats and dogs died in the USA after ingesting melamine-contaminated pet food imported from China. He is now Senior Experimental Officer/Laboratory Manager for Roy Goodacre (biospec. These included egg powder in Japan and South Korea. which was added to food products.21 in pet food by ELISA22 and UPLC and tandem MS. 1). including milk. META–PHOR Metabolomics for Plants. Soc. Ellis national and international press. at the National Science Museum in London.13. Brewster Victoria Brewster studied for her BSc in Chemical and Forensic sciences at The University of Bradford. Will Allwood moved to Aberystwyth (Ceredigion) where he studied for a BSc in Microbiology and completed a PhD in the areas of metabolic profiling and molecular plant pathology.

. FeSO4). sawdust. red ochre (hydrated iron(III) oxide).29 200 times the recommended amount in meat. copperas (Iron(II) sulfate. also known as poison nut.View Online Some food adulterants identified by Accum and Hassall in the 1820s and 1850s Adulterant Food product Cayenne pepper Cocoa and chocolate Coffee Confectionery For bulking/weight Bulked with ground rice.rsc. (KC4H5O6). This journal is c The Royal Society of Chemistry 2012 . In 1998 he moved to the UK. cream of tartar (Potassium bitartrate. turmeric Venetian red. Nux vomica seeds. liquorice. The University of Manchester. maize. ginger. orpiment (arsenic sulfide. when it was revealed that animal feed contaminated with industrial oil had been provided to over 50 beef and pork farms and one dairy farm across the Republic of Ireland and Northern Ireland.e. green sweets were often found coloured with copper salts (i. KAl(SO4)2Á12H2O). salt of tartar (potassium carbonate. and vermilion (mercury sulfide. and turmeric to enhance the yellow colour Cayenne. other iron compounds Burnt sugar (known as black jack) as a darkener White comfits coloured with Cornish clay (kaolinite. as well as advanced chemometrics and machine learning. 5706–5727 Roy Goodacre is Professor of Biological Chemistry at the School of Chemistry. wormwood (Artemisia absinthium).org | doi:10. dried leaves from various plants other than tea. and from 2001 was appointed Senior Experimental Officer and NMR Manager at The University of Manchester. alum (hydrated potassium aluminium sulfate. potato.26 and finally by ultrasound assisted HPLC in pet food27 as well as ultrasound assisted electrospray ionization MS in untreated milk and wheat gluten. vermilion. mustard seed husks. Prussian blue (Fe7(CN)18Á14H2O) for black tea. which is 1–3 ng WHO-TEQ/kg fat (a WHO-TEQ is the World Health Organisation’s definition of a toxic equivalency Alexander Golovanov was born in Russia. where he graduated from the Moscow Institute of Physics and Technology in 1988. Fe2O3). Pb3O4). honey. He obtained his PhD in Chemistry in 1994 studying the structure of cobra neurotoxin using NMR spectroscopy. His expertise involves mass spectrometry. salt Arrowroot (Maranta arundinacea). venetian red (derived from nearly pure ferric oxide. HgS) Plumbago. During routine monitoring of Irish pork at this time. tapioca flour. Cu(CH3COO)2) and Scheele’s green (copper arsenite.bios pec. Soc. starch. roasted wheat. salt. Al2Si2O5(OH)4).28 The so-called ‘Irish pork crisis’ became public in December 2008. 5708 Chem. a founding director of the Metabolomics Society and a director of the Metabolic Profiling Forum. and containing dissolved tin and lead when boiled in pewter vessels Table 1 Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. sago. sugar. FT–IR and Raman spectroscopy. where he leads a research group and continues to work in the area of NMR applications to Structural Biology. rye flour. chicory Chicory. The University of Manchester. His group’s main areas of research (http://www. red sweets coloured with red lead and vermilion. cinnamon. In 2006 he joined the academic staff in Alexander P. capsicum.25 in meat and pet food using LC–MS. Cervus elaphus). turmeric. roasted beans.net/) are broadly within analytical biotechnology. potato flour.1039/C2CS35138B Custard powders Wheat. Rev. K2CO3) Often contained lead from olive presses Coloured green with copper (Cu2+) salts Brown sugar. metabolomics and systems biology. 41. As2S3). UK. wheat. acorns For appearance/taste/aroma Coloured with red lead. 2012.24 in rice concentrates by LC. treacle Coloured with red lead (lead tetroxide. sand. coriander and caraway seeds. a poisonous crystalline alkaloid. rice flour Gin Water Olive oil Pickles Porter and stout Water Red cheese Tea Vinegar Previously used tea leaves. Golovanov the Faculty of Life Sciences. china clay scattering (SERS) and HPLC. C30H34O13). from the strychnine tree (a major source of the highly poisonous alkaloids strychnine (C21H22N2O2) and brucine (C23H26N2O4). indigofera. He is Editor-in-Chief of the Royston Goodacre journal Metabolomics. CuHAsO3) Lead chromate (PbCrO4). hartshorn shavings (horns of the male Red Deer. Cocculus indicus (commonly known as fishberry or Levant nut and a primary source of picrotoxin. gum. copper salts for green tea So-called ‘sharpened’ with sulfuric acid (H2SO4). potato flour. copper acetate. elevated levels of polychlorinated biphenyls (PCBs) were found in pork and further investigation revealed the presence of dioxins and dioxin-like PCBs up to 200 pg WHO-TEQ/g fat. cassia (Cinnamomum aromaticum).

foodstandards. then milk.42 In addition to these cases.org | doi:10.37 Spanish Toxic Oil Syndrome in 1981 from the consumption of denatured rapeseed oil intended for industrial use.1039/C2CS35138B Fig.38–40 and polyethylene glycol (anti-freeze) and raw methanol adulteration of wines. 41. where food adulteration was a means for fraudulent economic gain (or attempts at cost savings in terms of using cheaper ingredients). factor to the most toxic dioxin. Chem. Rev. http://www. that of deliberate adulteration/contamination of the food supply as an act of bioterrorism..46 indeed. This led to a mass recall of pork products. (C) ESI+/LTQ-Orbitrap XL MS collision induced dissociation (CID) fragmentation mass spectra.62 min.uk/multimedia/faq/dioxins_qanda/).View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. As a consequence of the Belgian incident alone. This journal is c with the subsequent knock on effects and product recalls being equally swift.43–45 It is apparent that the subject of detection science in foods. Soc. Later reports confirmed that dioxins were also detected in beef.36. from two dairy farms in Northern Ireland. when feed contaminated by transformer oil was distributed to farms in Belgium and to a lesser extent the Netherlands.30 This was reminiscent of an earlier incident in Belgium in January 1999. 1 Reverse phase C18 ultra high performance liquid chromatography (UHPLC) mass spectrometry system analysis of melamine: (A) UHPLC chromatogram showing melamine eluting at 0. France and Germany.rsc.34. 5706–5727 5709 The Royal Society of Chemistry 2012 . it has been stated that between 10–15 kg of PCBs and 200–300 mg of dioxins were ingested by 10 million Belgians and that neurotoxic and behavioural effects in neonates were expected but could not be quantified. and within several days thousands in the industry had lost their jobs and in excess of 20 countries had banned the import of Irish pork. Some of the more prominent of those publicised in recent history (apart from those already mentioned) include: carcinogenic red dye Sudan-1 found in some chilli powder and tomato based products. during the last decade this area has become open to a perhaps more sinister and destructive potential. they would very easily fill several volumes. (B) ESI+/LTQ-Orbitrap XL MS full scan mass spectra of melamine in positive ion mode.35 BSE in the late 1980s and early 1990s (a result of scrapie infected meat and bone meal (MBM) being fed to ‘ordinarily’ herbivorous cows). events with international impacts such as these are not unique or infrequent. is by its very nature huge in scope and that all of the detection methods cannot be covered with any depth in a single review of this size. gov. With electronic mass media these are rapidly published throughout the global community.31–33 Unfortunately. reported to be the result of dioxin residues in the fat of animals which had previously eaten contaminated feed.41. 2012.

fatty acids. This technique is based on the principle that when a sample is interrogated with an infrared beam (usually in the mid-infrared from 4000–400 cmÀ1 or near-infrared 14 000–4000 cmÀ1). its rapidity cannot be overstressed as 1000s of spectra can be collected within a 24 h period. 41.. and P–O vibrations from nucleic acids (1245–1220 and 1090–1085 cmÀ1). a range of MS-based. The rapidity and reproducibility of FT–IR spectroscopy is demonstrable through the large body of research published using this technology. FT–IR has been recognized as a valuable tool for metabolic fingerprinting.58–61 and Fig. 2. validated and robust chemometrics must be used in order to translate data into information. Raman spectroscopy (vide infra). amino acids. 5710 Chem. as well as raising awareness of the range of possibilities available in this exciting and rapidly developing era of interdisciplinary detection science. nuclear magnetic resonance (NMR) spectroscopy. and within the biosciences this was assisted in the 1990s by the technique being demonstrated to be particularly useful for the discrimination of axenically cultured bacteria.52–55 Alternatively. every chemical or biochemical substance will have its own unique infrared profile. as this part of the spectrum is where the fundamental vibration is seen (as opposed to overtones or harmonics in near-IR) and is particularly information rich. These include vibrational spectroscopies (such as infrared and Raman techniques). having the potential for high-throughput analyses) and finally. Fourier transform infrared (FT–IR) spectroscopy FT–IR spectroscopy allows for the extremely rapid. 5706–5727 This journal is c The Royal Society of Chemistry 2012 . 2012. one may use the related vibrational technique. With relevance to biological applications the mid-IR can be further broken down into specific regions.49. The resultant infrared absorbance spectrum can therefore be described as a chemical ‘fingerprint’. reconstituting in D2O. Soc.1039/C2CS35138B Therefore.org | doi:10.48 Moreover. highthroughput and non-destructive analysis of a wide range of sample types in both research and industrial environments. 2 Schematic showing collection of typical Raman and FT–IR spectra collected from whole milk. notable for both their rapidity and portability.49 A significant proportion of the work undertaken in recent years has concentrated on the mid-IR part of the electromagnetic (EM) spectrum. or through the use of attenuated total reflectance (ATR) as a sampling method.56. N–H and C–N from proteins and peptides (1750–1500 cmÀ1).View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Applications of FT–IR spectroscopy have been broad in scope. or windows. as it is characteristic of the particular sample under analysis and hence. nucleic acids and polysaccharides rapidly and simultaneously with a minimum amount of sample preparation or background training.57 the various approaches for chemometrics will be detailed below. C–O and C–O–C from polysaccharides (1200–900 cmÀ1). It is hoped that this will encourage the general reader to explore this area in greater depth. of biological interest such as: C–Hx stretching vibrations from fatty acids (3050–2800 cmÀ1).50. again notable for its rapidity (and. this problem can be overcome either by dehydration of samples. functional groups within the sample will absorb this radiation and vibrate in one of a number of recognized ways (such as stretching or bending vibrations)47 and these vibrations/absorptions can then be correlated directly with (bio)chemical species. lipids. and indeed.48. Typical FT–IR and Raman spectra from milk can be seen in Fig. A further perceived disadvantage is that as a holistic measurement is made containing biochemical information from across the whole of the IR (or Raman) spectrum. like vibrational techniques. proteins.49–51 as it is able to analyse carbohydrates. Rev. The molecular vibrational states along with a simplified Jablonski diagram are also shown. chromatographic and other technologies. reduction in path length. CQO. the authors wish to give a ‘flavour’ of some of the methods used by discussing a selection of the rapid detection technologies currently available. due to its holistic nature. and a number of their applications to a variety of food and feedstuff authentication/ adulteration problems.rsc. Whilst one of the potential limitations of FT–IR spectroscopy is that the absorption of water is very intense in the mid-IR. subtraction of the water signal.

77. 64 scans were co-added and averaged.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. 2012. the most prominent of which is the highlighted band at 1730 cmÀ1.74 It is consumption of a diet high in these artificial fats which can elevate levels of serum low-density lipoprotein (LDL)-cholesterol and lower levels of high-density lipoprotein (HDL)-cholesterol. studies into the characterization and determination of adulteration of edible oils for example. Studies by Mossoba74.66 and retail industries. which can be assigned to the CQO stretching in the flavonoid component of the grapefruit juice. the major source for concern in terms of human health are the commercially hydrogenated trans fats found in some food products. including peanut.51. although this limitation has been overcome with a more recent methodology which uses the height of the negative second derivative of the 966 cmÀ1 band. Soc. This journal is c The Royal Society of Chemistry 2012 Chem. employing a motorised HTS module. in addition to the regulatory bodies responsible for monitoring the food chain from farm to fork. This newer methodology has been shown to determine trans fat levels in unknown samples as low as approximately 1% of total fat. highlighting the vibrational modes on which the PLSR model was based. 5706–5727 5711 .org | doi:10.78 Thus the detection of this modification is important. 3 illustrates this with respect to fruit juice adulteration). 3 (A) FT–IR spectra of pure orange juice (red) and pure grapefruit juice (blue).74 Whilst trans fatty acids are known to be present in low levels in various foodstuffs. this had a limit of detection of 5% trans fat..80 and others have clearly demonstrated the efficacy of ATR as a rapid and accurate sampling method for trans fat analysis in foods.82 Other applications have included the direct measurement of trans fats in French fries acquired from Fig.67–70 Such applications have included those involved with product authenticity and adulteration (Fig. This demonstrates that the two juices can be discriminated between based on spectral features. 41. This includes those sectors involved in the production. Samples were spotted onto a 96 well silicon plate. These bands correlate well with those highlighted in (A). FT–IR and particularly near-infrared (NIR) spectroscopy have also been viewed as exceptionally useful spectroscopic tools for the analysis of foodstuffs.63 more specifically referred to as metabolic fingerprinting.49.48. like dairy75 and meat produce76 for example. Each point represents one spectrum (data were pre-processed using extended multiplicative scatter correction (EMSC)344 (polynomial order 7) and column mean centred). Many of these have utilised what they termed the trans absorption band at 966 cmÀ1.79.62 The work undertaken during the 90s spawned a considerable amount of interest which has led to other areas of research.64 Concomitant with these multidisciplinary applications.1039/C2CS35138B work on food-related bacteria continues.81 One recent study used partial least squares regression (PLSR) to build a calibration model using the height of the second derivative at 966 cmÀ1 and applied a handheld diamond ATR spectrometer for the measurement of trans fats in several edible oils.rsc. safflower. During the last decade it has become increasingly evident that mid-infrared (MIR) spectroscopy has considerable potential for industrial applications in the food. packaging65. corn and coconut oils. (C) PLS loadings from the 1st latent variable (w*c[1]). feed. such as disease diagnostics50 and metabolomics. Previously. processing.71–73 and rapid determination of total trans fat content in foods. (B) PLS predictions from FT–IR data of orange juice adulterated with grapefruit juice (0–100% in 10% steps).74 leading to increased risk of coronary artery disease. Rev. and related industries. Spectra were collected on a Bruker Equinox 55 infrared spectrometer. Spectra were acquired over 4000–600 wavenumbers.

.98 has been undertaken on both poultry and beef using FT–IR52.52 These studies focussed on acquiring a biochemical fingerprint of metabolites formed as a result of the growth and enzymatic activity of bacteria in situ on the meat surface. Soc.109–111 The rapidity of this method also allows for frequent measurements at all stages.86 Other applications of FT–IR have included determination of the quality. its highly precise and efficient nature and its portability. as already mentioned.rsc.87. The authors found that the most significant process detected by both infrared spectroscopies and highlighted by the OPA approach was proteolysis. though these are not yet as mature as NIR. one very recent study used both NIR and MIR in an attempt to extend current knowledge of the ability of both these methods for the rapid detection of microbial spoilage in poultry. and several more concerned with bacterial spoilage/contamination and food poisoning.101 and by a very similar study using short-wavelength NIR detection of the microbial spoilage of chicken.120.97 Rapid detection of microbial spoilage51. they stated that this was not due to the decrease in protein content of the muscle food per se but rather as a function of an increase in absorption in both NIR and MIR wavelengths corresponding to free amides and amines.52.123 This latest study by Alexandrakis and co-workers further demonstrates the potential of infrared spectroscopy for the rapid non-destructive detection of microbial spoilage in foods. Food science-related NIR embodies a large and vibrant research community and.83 determination of trans fats from cereal-based South Asian foods84 and a wide variety of cereal-based snack products85 as well as deep fat frying oils. in regards to the maturity of this technique.99–101 and the first publication of a food metabolomics study was undertaken using FT–IR spectroscopy. One other study using this approach and the same platforms employed OPA of NIR–MIR data for the authentication of the protected designation of origin (PDO) of Spanish cheese. These retrospective measurements are not useful as the food has usually left the retailer and entered the food chain and may cause illness.96 as a first step in product authentication. 5706–5727 simultaneously.122 This particular study by Alexandrakis and co-workers involved the collection of NIR and MIR spectra from 65 poultry samples over 14 days at 4 1C. Multiplicative scatter correctiontransformed NIR spectra and standard normal variate-transformed MIR spectra were first analysed using PCA and partial least squares-discriminant analysis (PLS-DA) prior to OPA.88 bacterial spoilage of milk53 and bacterial growth interactions in milk leading to dominating metabolic effects. it has been stated that on/in-line NIR instrumentation has a proven record of 40 years108 in the food processing industry. Therefore rapid detection and quantification within minutes is highly desirable in order to allow adequate control within HACCP (hazard analysis critical control point) regulated systems.124 This journal is c Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Rather than having first to separate physically the microorganisms from the substrate (which is generally the case) and undertaking the time-consuming analysis and enumeration of the bacteria themselves. through process monitoring to quality control of finished products.g. as they were said to contain information highlighted in OPA revealing a relationship from both methods. Results from OPA were reported from the wavenumber ranges 1550 nm–1996 nm (NIR) and 1000 cmÀ1–1188 cmÀ1 (FT–IR). structural characteristics and identity of dairy products. perhaps due to the robustness and simplicity of instrumentation. such as trans fat analysis117.114 (2) the water absorbance in the near-IR is not as strong as in the mid-IR so analyses are more routine and no additional accessories are needed (e. a long history of applications in industry and both its inclusion here and its continuing contribution to food science are not to be underestimated. though two main advantages that near-IR has over mid-IR analysis is that: (1) near-infrared light penetrates much deeper into an intact food sample (>10 mm). 41. particularly with respect to continuous monitoring and control of process and product quality.93 A number of studies have also been undertaken on meat and meat products.102. These correlations validate the results and conclusions made by two earlier MIR studies and mentioned above.118 and food spoilage for example.. Further. 2012.120 For this study the authors utilised outer-product analysis (OPA).89 identification of modified starches90 and rapid identification of foodborne fungi have also been undertaken.92 amongst others. from production and collection of raw ingredients.115. Near infrared (NIR) spectroscopy For many decades much of the food-related infrared research was almost exclusively reserved to NIR spectroscopy. ATR). Indeed. as well as through various packaging materials that are transparent to NIR light.109 These instruments usually have a very high signal-to-noise ratio112 and.View Online local food outlets with no sample preparation and pressed directly against a ZnSe ATR crystal.103 The recurring theme in these studies is the need to replace traditional methods which can take days to obtain results on the level of bacterial spoilage.95 and at the intra-species level.91.org | doi:10. This has enhanced and accelerated the detection of microbial spoilage from hours to seconds and could theoretically be used to detect other forms of contaminating substances or bacteria.113 For food and feedstuff analyses this method clearly has many advantages similar to those of MIR (both being infrared methods).119 Indeed. in addition to having remote sampling capability allowing for the collection of real-time information in a process stream. like MIR. can measure multiple components in a food sample 5712 Chem. with many focussing on adulteration and authenticity issues.121 a method said to enable the emphasis of co-evolutions in spectral regions from data acquired from two different domains (heterospectral).112 though this can also depend on the chemical species being detected and the complexity of the food matrix under analysis (vide infra). In terms of limitations. as well as attempts to detect beef adulterated with offal. Though in terms of portability. more recently handheld portable FT–IR82 and Raman104–107 instruments are also becoming more widely available (see both the FT–IR and Raman sections). These have included studies discriminating meat at the species level.94.1039/C2CS35138B The Royal Society of Chemistry 2012 .116 It is perhaps not surprising that many of the research applications of NIR to food and feedstuff analysis cover similar areas to FT–IR in the MIR. Rev. one of the main perceived disadvantages has been stated to be the low sensitivity to some minor constituents. or from a single domain (homospectral) following spectral perturbation.

an electromagnetic enhancement effect (thought to dominate).133 faecal contamination of fruit.147 Whilst Raman and infrared are both vibrational spectroscopies. Soc.144–146 Like infrared spectroscopy. there are a number of ways to enhance the Raman effect. The study by Mauer and co-workers for example was able to differentiate unadulterated infant formula powder from samples containing 1 ppm melamine with an assay time of 1–2 min to detection. This radiation is at a specific wavelength. Excitation of this type is in resonance with the electronic transition. v = 2) (See Table 2 for a summary of different types of Raman scattering and Fig. largely due to the diversity of analytes which can be probed. a colloidal solution or roughened electrode (usually Ag or Au). The Raman spectrum obtained is a measure of how much energy a photon has lost (Stokes Raman scattering) or gained (anti-Stokes Raman scattering) which is expressed as a wavenumber shift. Two types of light produced: Stokes (energy loss) and anti-Stokes (energy gain) scattering. 41.76 Æ 0. proteins141 and carbohydrates. Raman spectroscopy is able to detect analytes in solutions with minimal interference from water. provides a unique spectral fingerprint of many analytes. v = 1.108 Cen and He. known as chemical enhancement.1039/C2CS35138B Raman scattering Conventional Resonance Raman (RR) UV resonance Raman (UVRR) Surface enhanced Raman scattering (SERS) Scattering probabilities 1 in 106–8 photons Typically 104 enhancement Typically 104 enhancement 103–1014 enhancement Single molecule detection possible Sample is excited by a monochromatic light source and inelastic light scattering occurs. The enhancement explained by two processes.125 as well as the assessment of meat adulteration with processed animal proteins (PAPs). these include resonance Raman (RR) spectroscopy.143 The utility of Raman spectroscopic instrumentation is greatly increased by the ability to couple it to microscopes resulting in 1 mm resolution which is ideal for single microbe identification or trace analyses and when interfaced with fibre optic probes for in situ or online analysis. Has a fluorescence quenching effect. Used to excite specific chromophores such as nucleic acids and aromatic amino acids.org | doi:10.View Online A significant number of studies have been undertaken with NIR for the authentication of a wide variety of food products.132 In addition to studies concerning milk adulteration with less expensive components such as water or whey. infant formula and milk powder just below 1 ppm (0. other recent studies have also utilised both MIR and NIR in combination with statistical methods to develop rapid and sensitive detection methods for melamine in complex dairy matrices. allowing analysis of food products through plastic or glass packaging (as does NIR. However. pigments and flavourings.137 As well as NIR investigation of the levels of specific chemical species in foods and beverages such as the typical complications regarding the NIR measurement of water112 for example. 2012. Rev. Raman spectroscopy Raman spectroscopy is a versatile..127 provenance and classification of edible oils and fats with a high degree of accuracy72.139 For those interested in NIR analysis. The technique is seemingly ideally suited to the analysis of foodstuffs. infrared measures the absorption of energy. whereas Raman spectroscopy measures the exchange of energy with EM radiation (Fig. pork and chicken). or detection of phenols in wines.128 and a recent study estimating the concentrations of polyunsaturated fatty acids and their biohydrogenation products in the subcutaneous fat of cattle beef cows. and a charge transfer mechanism. to the minor components such as dyes.130. Moreover. Deep UVRR measurements below 260 nm are not plagued by fluorescence interference. Requires close proximity adsorption onto a roughened metal surface.130 Other NIR studies include deliberate adulteration followed by spectral analysis of meat and poultry (beef. The collected scattered photons will be either Rayleigh (elastically) scattered or Raman (in-elastically) scattered. due to the low probability of the molecule being in a excited vibrational state (viz. many reviews and several books have been published on this key tool in food science and the reader is directed to more recent reviews by Cozzolino.126 detection and identification of transgenic corn.142 Also of interest are the microorganisms which play a role in food spoilage. surface enhanced Raman scattering (SERS) and tip enhanced Raman scattering (TERS).138. Sample is excited with a frequency of light that is within the molecular absorption bands of the sample. In RR spectroscopy the sample is excited with a frequency of light which is within the Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. already mentioned above).rsc. 5706–5727 5713 . Can be plagued by fluorescence. Unfortunately the Raman effect is typically very weak as only 1 in every 106–108 photons are in-elastically scattered.11 ppm). However. and especially so in the case of NIR.106. Raman spectroscopy offers a distinct advantage over FT–IR spectroscopy in that the confocal nature of the technique allows samples to be profiled through a variety of transparent materials. usually provided by a monochromatic light source in the visible to near-IR part of the EM spectrum. lamb. 2 for the different vibrational energy levels). to identify whether they had been fed flaxseed. Stokes Raman scattering is most commonly measured as this is of higher intensity than anti-Stokes scattering.109 amongst others. The combination of the above properties with the possibility of portable instrumentation makes Raman spectroscopy a promising technique for food science. like IR spectroscopies. The more recent study by Balabin and co-workers used a variety of non-linear regression methods for pre-processing and spectral analysis of NIR and MIR data to detect triazine derivative products in liquid milk. These include meat and poultry. 2). ranging from macro-food Table 2 Summary of different types of Raman scattering Properties components.129 In regards to detection of deliberate adulteration or accidental contamination. In Raman spectroscopy a sample is illuminated with light from a laser and the interaction of photons from this EM radiation with molecules in the sample induces a series of molecular vibrations.113 and Huang et al.131 These were shown to be truly rapid methods. minimal or no sample preparation is required.140 Woodcock et al. such as lipids. non-destructive analytical technique which. Can be coupled with RR to ‘tune’ for a specific chromophore for additional resonance. with FT–IR requiring up to 5 min. This journal is c The Royal Society of Chemistry 2012 Chem.125.134 mycotoxin infected135 or adulterated forage/feed products136 and spirulina powder.

pure (hazel)nut oil. Chromatographic methods of edible oil authentication have proved unable to detect adulterations at low concentrations (5–20%) as the major chemical compositions are very similar. allowing the Raman scattering to be enhanced by a factor of 104 over conventional Raman scattering. 4 (A) Raman difference spectrum of olive oil subtracted from olive oil adulterated with 10% unrefined hazelnut oil for the region 800–1800 cmÀ1 (obtained at 785 nm). usually silver or gold. in particular extra-virgin olive oil. such as electrospray and pyrolysis MS72. An example of this is naringin. This excitation is in resonance with the electronic transitions. a major flavonoid glycoside found naturally in grapefruit (metabolized to the flavonone naringenin in humans. 2012. 5714 Chem. 4).166 have been employed with some success. and fraudulently increasing profits (Fig. Work by Baeten and co-workers167 has shown that FT-Raman and FT–IR spectroscopy used in conjunction with stepwise linear discriminant analysis (SLDA) can distinguish between olive oil and hazelnut oil. Spectra were collected over 200–2000 wavenumbers with 30 s exposure time. as the chromophores that are measured are the aromatic amino acids and the nucleic acids. This discrimination was mainly based on a shift Fig. Whilst the closely related TERS combines atomic force microscopy (AFM) and Raman spectroscopy together.156 As already noted above. thereby reducing manufacturing costs.148 By contrast. and very often purposefully so.1039/C2CS35138B molecular absorption bands of the sample.149 Deep UV resonance Raman (UVRR) spectroscopy is ideally suited to the study of biological materials. as consumers may be hypersensitive to this adulterant. this form of adulteration can also have serious health implications resulting from pharmacological interactions.155 thereby allowing single molecule detection. SERS involves the analyte being in close proximity to a roughened metal surface or a colloidal solution. with claims of 1014 reported for specific analytes154. Once a specific test has been designed to identify an adulterant. adulteration is a major concern for both the food industry and consumers for many reasons.org | doi:10. whereby cheaper oil of a similar chemical makeup is often added in order to mimic the more refined and expensive oil. such as the clinical modulation of drug transport. (C) Band assignments from oils. but none have the simplicity and rapidity of Raman spectroscopy which has proved exceptionally useful in this area. due of course to the fact that the adulterant components are usually very similar to the authentic product. 5706–5727 This journal is c The Royal Society of Chemistry 2012 . PC2) of the Raman spectra from pure olive oil. and olive oil adulterated with 10% nut oil.148. 41.96 is far from contemporary and yet it is often still difficult to detect. Rev. Aside from being blatantly dishonest. potential fraudsters can become aware of this and then add or remove that component from the adulterated foodstuff.150–153 This gives an enhancement of the Raman signal typically 103–6 over conventional Raman.160–162 One notable food adulteration example is the production of high-value edible oil. Soc. This can be particularly problematic for example when nut oils (such as unrefined hazelnut oil) are used.165 and NMR spectroscopy. operating with a 532 nm excitation wavelength.. major bands correlate well with those reported in the literature.rsc.164. whose pattern was developed as a rapid method for the detection of orange juice adulteration with grapefruit juice157–159). (B) PCA scores plot (PC1 vs.168.170 Raman data were collected using an Ocean Optics portable Raman probe.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.163 Other methods. altering their bioavailability and hence effectiveness with potentially serious consequences.

41.187 and as chemometric and spectroscopic techniques have become more sophisticated both single cell analysis and strain level identification has been possible.180 Raman spectroscopy has a rich history as a tool for studying protein conformation both in solution and in solid state.173 More recently Korifi and co-workers combined confocal Raman spectroscopy with PLS and PLS-DA to authenticate the PDO labels of six French virgin olive oils. were used to quantify the amount of hazelnut oil in a variety of olive oils at levels of commercial interest (0–20%) successfully. the detection of food spoilage bacteria by Raman spectroscopy has been reported by many groups62. it can be used to identify functional groups and provides valuable structural information and determines the physical and chemical properties of atoms within a molecule. Multivariate analysis methods. with a prediction error of 5–7. Olive oils may also be adulterated with sunflower or vegetable oils and their similar triacylglycerol and fatty acid composition has meant that detection by conventional techniques can be problematic. SERS has been used for the analysis of pesticides present not only in the juice but also on the surface of the fruit. including solids and solutions.176 Key examples include the detection of yeast in apple juices. El-Abassy and co-workers170 used Raman spectroscopy with a 514.181 UVRR spectroscopy has been used to monitor the conversion of oxy-myoglobin to deoxy-myoglobin during the pressure treatment of meat products182 and the decrease in a-helix and increase in b-sheet content in electron beam irradiated fish products. The vast majority of elements possess a magnetic spin number greater than zero which is necessary for the NMR effect and common nuclei observed by NMR spectroscopy have a spin quantum number of 1 2. such as pair wise minima and maxima selection (PMMA).196. This study also differentiated between extra virgin olive oil and olive oil adulterated with hazelnut oil.1039/C2CS35138B of SERS to evaluate the quality of tomato juice. volume. Marini and colleagues used artificial neural networks (ANNs) in conjunction with Raman data to identify the geographical origin of European oil samples from a variety of regions171 and were able to distinguish between binary blends of oils from different cultivars. The authors discovered that signal intensity was increased by use of a higher energy laser and that some components (mainly carotenoids) gave resonant enchancement at this wavelength allowing analysis at lower concentrations. NMR spectroscopy can be used across a wide range of sample types. although this can be overcome using SERS as this is fluorescence quenching whilst also enhancing the Raman cross section. as well as complex mixtures195 (Fig. Using this method it was possible to differentiate between 13 different oils using PCA and PLSR was able to quantify the level of adulteration with a detection limit of 0. and portable SERS systems are beginning to be developed for the express purpose of on-line bacterial identification in food products. by employing a 785 nm excitation source and PCA. designed to simulate the way brain neurons communicate and have been previously used to detect the adulteration of extra virgin olive oils.5%. The relatively recent in situ investigation of meat quality through packaging using a portable.View Online in the CQC band from 1670–1660 cmÀ1. SERS has been used to identify Salmonella spp. resulting in two spin states..183 Raman spectroscopy has also successfully been combined with PCA and PLSR in this area to monitor tertiary changes occurring in protein isolates from kidney beans184 and soya products.194 The radiation is able to move through a variety of materials with ease with minimal heating effects and is sensitive to different chemical environments. In this study the authors focussed on determining the fatty acid and triacylglycerol compositions of oils in order to achieve a prediction accuracy of 92. Soc. such as changes in myofibrillar and connective tissue proteins in that occur during freezing and after frozen storage. such as PLSR and genetic programming (GP). has been used to detect vegetable and sunflower oil adulteration of olive oil at concentrations as low as 1% v/v.188 Finally.174 Similar approaches have previously been used to confirm botanical and geographical origins of European honey. More recently.org | doi:10.. Escherichia coli and many other foodborne pathogens with unambiguous results.189–193 Nuclear magnetic resonance (NMR) spectroscopy NMR spectroscopy is a fingerprinting technique which takes advantage of the magnetic properties of certain nuclei.194 As with the vibrational approaches already mentioned above. Previously only NIR excitation wavelengths had been used to avoid fluorescence contributions. foods can be readily described as complex systems in terms of both their chemistry (often containing multiple compounds varying in terms of their nature.179 In addition. there is little literature on the use of Raman spectroscopy to monitor food spoilage directly. as well as investigating various strategies for wavelength selection. Many such studies take advantage of the sensitivity of the Amide I band (1650–1680 cmÀ1) and Amide III band (1200–1300 cmÀ1) to conformational changes in proteins.197 Chem.05%. The transitions between these two spin states are detected when the nucleus is placed in a magnetic field and induced by radiation in the radiofrequency region of the electromagnetic spectrum. with a limit of detection for hazelnut oil in Turkish olive oil of 8% v/v.3%. work has also been undertaken on the geographical origin and provenance of oil. in conjunction with PLSR. In addition to concerns over adulteration. Rev. This approach has been applied in many cases to the study of food proteins and perhaps most notably in changes which occur in muscle foods during preservation treatments. 5706–5727 5715 The Royal Society of Chemistry 2012 . 2012. an example of such an approach is for the analysis of Sudan I in chilli powder. In earlier studies ´ pez-Dı´ ez and colleagues168 also demonstrated that Raman Lo spectroscopy could distinguish between pure olive oil and oil adulterated with hazelnut oil. hand-held Raman probe186 is among these few.169 In this study PLSR was performed on the full spectrum. That being said. and physically (frequently in multiphase systems). applications of Raman spectroscopy in this area are limited and this is due to the fluorescent nature of the pigments in the juice.175 Another problem area for product authentication is fruit juices and whilst there are many examples of this using FT–IR and NIR spectroscopy. FT-Raman spectroscopy.172. and abundance). Of course. 5). ANNs are a supervised non-linear multivariate methods consisting of groups of interconnecting artificial nodes.rsc.185 In comparison to FT–IR and NIR spectroscopies.177 monitoring the production of clarified apple and apricot juices178 and the use This journal is c Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.5 nm laser excitation wavelength.

NMR does suffer from relatively low sensitivity.1039/C2CS35138B The Royal Society of Chemistry 2012 . grape juice and wines). their hydrogen atom environments). providing information about a food sample regarding chemical pathway formation for example.198 However. These have included combined studies (1H NMR and LC–MS) for the direct analysis of the aromatic composition of some beverages (such as beer. the reader is directed to a review by Wishart.209 Indeed.208 due to the different metabolic pathways used by plants to generate sugars. whilst sugar cane uses the C4 pathway. other diverse food-based applications of NMR have included several studies of muscle foods. there is momentum within the NMR research community to move towards ever higher magnetic fields in order to improve sensitivity and to explore hyperpolarizability effects. 5716 Chem. care must be taken to control that environment as pH can affect the robustness of these metabolite fingerprints.199 Moreover.217 its provenance and geographical origin.org | doi:10. This is because quantifying the level of microbial or enzyme related metabolites produced during the fermentation and aging process (such as betaine for example).201 Whilst this is the case at the moment. such as olive oils222.219 quality.197 Other more recent combined studies (1H NMR and GC–MS) have involved quantifying polar extracts and identifying the major metabolites of interest (i. could be used as an assessment of quality216 (and conversely.218.211 and indeed. Spectra were recorded on a 600 MHz Bruker Avance spectrometer equipped with a Cryoprobe. The spectra illustrate the somewhat limited dynamic range available for identification of trace contaminants: very strong signals present in each spectrum may mask weaker signals from minor components. 5 (A) Generalised schematic of NMR spectroscopy. The strongest signals shown are clipped. and amino acids) in melon fruit.210.213 As with many of the other fingerprinting techniques discussed above. organic acids. sugar beet plants fix CO2 via the C3 pathway.. and/or its provenance or geographical origin. 2012.200 Additional metabolomic investigations have included global metabolic profiling and characterization of the complex metabolome of aged soy sauce. NMR spectroscopy has been said to be exquisitely sensitive to different chemical environments.215 and more appropriate to this article. an NMR spectrum can contain resonances which provide significant quantities of structural information. using 0. as well as being a quantitative and highly reproducible technique. food metabolomics.211 This technique determines the relative deuterium and specific deuterium-site locations in a molecule.rsc. Rev. Fortunately. this further enables the identification of individual constituents through the interpretation of chemical shifts for example. a number of studies have again been undertaken with edible oils. which allows for distinction of a wide range of chemicals in a mixture. as with all the approaches mentioned thus far. as an assessment of fraudulent labelling).220 as well as the provenance and classification of fish. more comprehensive reviews have been composed on this specific area. composition.5 mL samples.204 beer. especially so if compared directly to mass spectrometry based techniques (vide infra).205 spirits206 and fruit juices. Typical one-dimensional solution 1H NMR spectra of (B) coffee. As well as investigating the spatial localization of metabolites within the fruit mesocarp. Only signals from soluble components are visible in these spectra. or SNIF-NMR.42. several of these and other studies have employed a technique known as site-specific natural isotope fractionation. or This journal is c Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. sugars. and may need to be assigned to allow their identification.224 NMR has also been applied to the analysis of commercial culinary spices adulterated with Sudan I–IV dyes. (C) wine and (D) milk. such as the composition and authenticity of meat. Soc. Individual signals in these spectra report the presence of various chemical components (namely. complicating analysis.210.229.203 Applications of NMR spectroscopy have included studies based on the adulteration of a wide range of beverages including wines.226–228 and fish oils. Minor contaminating species. For a far more comprehensive overview of food metabolomics studies and in particular those related to NMR (amongst some other techniques).221.230 these have been concerned with provenance.198 However.e.View Online Fig.222 the quality and identification of dairy products88. 41.225 the carcinogenic red dye already mentioned above. with concentrations below 50–100 mM.46.212 A considerable body of work has been undertaken using SNIF-NMR related primarily to authenticity of fluidic food samples such as maple syrup and fruit juices amongst many others.200. it has a substantial record of food-based applications over the last three decades. As with other fingerprinting approaches. 5706–5727 and indeed. which has previously been analysed in foods using various mass spectrometry techniques.202 NMR spectroscopy therefore presents itself as another potential tool of choice for the analysis of foods and feedstuffs.223 and the adulteration of honeys with commercial sugar syrups. may not be readily identified.63 Finally.207 A number of these studies have clearly demonstrated that stable isotope ratios are particularly important in authenticity analysis of food and food ingredients. a significant number of studies have used NMR within the framework of systems biology and in the field of metabolomics/metabonomics214.

For example. B by ‘010’ and the third country C as ‘001’. so that country A is represented by ‘100’. level of adulteration.232 Due to the large number of variables collected in infrared.236 The output of these algorithms is an ordination plot or dendrogram and these graphical representations (or projections) of the original data allow the investigator to cluster objects (samples) together on the basis of their perceived closeness in the high dimensional spectral space (Table 4). country or species of origin. After clustering ordination (2D or 3D) plots or dendrograms learning (tree-like diagrams) are used to visualise the clusters. as it combines linear algorithms as principal components and simple non-linear techniques such as ANNs.) and stored securely if possible within a database. Rev. (3) Data can rarely be analysed directly and need some pre-processing and pre-treatment (Table 3). this information needs to be extracted using chemometric methods. These methods thus allow one to distinguish objects or populations. this being a result of all chemicals being detected simultaneously.235. or spoilage. This information is used in classification or quantification modelling. harvest time. often referred to as the Y-data or the output data. For example. diacylglycerols.1039/C2CS35138B adulteration and/or authentication of these foodstuffs. these include for example baseline corrections. The multivariate algorithms used for data analysis fall into three categories. multivariate models cannot extrapolate and so exemplar spectra must be included within reasonable ranges and include many possible outcomes else the model will not generalise.327 These are the traits one is interested in.235 It is therefore essential that the Y-data must be unequivocally known during model construction. Multivariate data This journal is c The Royal Society of Chemistry 2012 Chem. This process is often achieved by reducing the error between the models output prediction and the actual known target trait.226 Different admixtures of these two oils were analysed by 1H NMR and 31P NMR.rsc.. The focus of these chemometric methods is to build a multivariate model which best describes the system under analysis. contamination. and free fatty acids. Often referred to as the X-data or input data. (5) Following this stage the data are then interpreted and the results collated for dissemination.231. (2) The data are then collected along with any associated metadata (data about the data. ripeness. A popular algorithm used for such a purpose is PCA and this results in the generation of a scores plot (a scatter plot which summarise variation in samples). If groups do appear then these can then be interpreted chemically Spectroscopic data comprising many variables (e..235 This process is often referred to as dimensionality reduction or simplification. sterols. One of the most recent of these involved the detection of the adulteration of olive oils with hazelnut oil.g. Within food analysis explanatory data analysis typically uses unsupervised learning and the multivariate algorithms seek to discover natural groupings in the data. Any grouping that arises within the PCA scores could be due to some specific chemical difference between samples. The chemometrics pipeline has several data handling stages:234. 2012.235 (1) The first step is the design of the experiment and this should be as rigorous as possible and aim to capture adequately the domain under analysis. Raman and NMR spectroscopic investigations of food adulteration. Whilst both multivariate methods were said to provide good discrimination between the two oils. if there are three modelling countries then the Y-data are encoded such that a ‘1’ is placed in the correct country and ‘0s’ elsewhere. Pre-treatment Transforming the pre-processed data to make them suitable for analysis. absorbances at specific wavenumbers for FT–IR. Whilst these ‘composite’ spectral fingerprints contain relevant information on adulterants in the foodstuff. classification or quantification. This is usually the level of adulterant/contaminant and the Y-data is simply modelling the level of the trait of interest. This marriage of chemometrics with spectroscopy is the key to the success of fingerprinting methods in the area of food forensics.org | doi:10.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. This process uses some mathematical transformation to associate the X-data with the target trait (Y-data). This may be for different countries of origin. 41.235 This may include removing baseline artefacts.g. merely the construction of standard curves for absolute quantification. fatty acids for example were determined by 1H NMR and 31P NMR being used for the determination of minor compounds including phenolic compounds. Soc.327 Metadata Data about the data. This is because the retention time of naringin under explicit LC separation conditions is known and there is no need for complex data processing. these address particular problems and are used for exploratory analysis.233 Table 3 Definition of common chemometric terms There are many algorithms within the ‘chemometrics zoo’ and a recurring theme within spectroscopic analyses is the development of a range of different chemometric approaches for detection and quantification of food adulteration. transformations and removing any outliers (unusual atypical samples) in the data. Classification The goal is to classify samples into groups. the aim of combining these high dimensional data with multivariate analysis (MVA) is to simplify the data matrix. Supervised learning Analysis conducted on both the spectroscopic X-data and Y-data.226 Data processing For targeted analyses of specific adulterants in food there is little need for data processing. (4) The data are then ready for analysis by the chosen chemometrics algorithm and include adequate statistical validation (vide infra). Pre-processing Methods used to go from raw data to clean data ready for analysis. Classification of the oils based on the fatty acids and the concentration of minor compounds was undertaken using canonical discriminant analysis (CDA) and classification binary trees (CBTs). level of adulterant. if one were interested in detecting grapefruit adulteration in orange juice then naringin can be readily detected and levels easily quantified with high performance liquid chromatography (HPLC). peak-picking or alignment.. scaling. e. etc.328 This typically includes normalisation. Quantitative The goal here is to quantify the trait of interest. or ion counts at specific m/z for MS) collected on the samples. Unsupervised Analysis conducted on only the X-data with the goal to generate clusters from these input data. it was stated that CBT classification was superior. By contrast. 5706–5727 5717 . normalisation routines and scaling. binning. etc. spectroscopy is typically performed without recourse to prior chromatographic separation and therefore the spectra are highly complex. such as plant species of fruit juice analysed. country of origin.

rsc.View Online Chemometric strategies and methods employed for the analysis of spectroscopic data Model type Exploratory Unsupervised or supervised Unsupervised Comments Very old method used to explore the data to look for any obvious clusters and for the detection of outliers. For supervised learning to be effective primary reference data must be used. The third and final category of multivariate analysis is based on quantification. or (b) using re-sampling methods so as to select subsets from the training data. There are many different types of supervised algorithms that are used for the analysis of infrared. % adulteration. A particularly popular algorithm used is PLSR which is based on multivariate regression. but rather than extracting the maximum variance from the spectra.. Similar to PCA this is a projection method.56 During supervised learning it is essential to test the validity of the model being constructed. This is especially the case for supervised learning algorithms which have no say over the initial data that they are trained with. Validation is a vital step in any multivariate data analysis and must not be omitted else the results are questionable. Generates a dendrogram which is a tree-like structure where the leaves are the samples and the branches shows their relationships Kohonen neural network based on unsupervised learning. One may therefore wish to discriminate between oils produced in Tuscany from those ‘lower grade’ oils from other Italian regions. details of these and their main characteristics are summarised in Table 4. An additional use for these exploratory multivariate approaches is to identify outliers which are atypical samples and these can arise either from the samples themselves or from analytical artefacts. Raman and NMR data and these have been highlighted in the text above. Series of algorithms to perform variable selection. Model validation is needed to make sure that conclusions about the model are statistically valid. This primary reference data refers to the sample origin (region. After which validation of this model occurs. Finally. production method) and is the desired response (output) that one wants from the algorithm. or on. botanical.g. As the name suggests the aim here is for the algorithm to generate predictions on the level of something of interest. discriminant function analysis (DFA). Common methods include genetic algorithms (GAs) and genetic programming (GP) Results in a tree where the leaves are categories of samples and the branches are decision boundaries An ensemble classification algorithm that uses many CART solutions to allow more robust predictions. In this calibration phase a training set is employed which contains training pairs of spectroscopic input data (X-data) with known target outputs (Y-data). These are viewed in ordination plots of the PC scores. The output from PLSR is a real numerical value (e. Relevant chemical information can be extracted from the loadings matrix. Includes linear discriminant analysis (LDA). Rev. this could be the percentage level of adulteration or 5718 Chem. 2012.237 There are several re-sampling strategies. Output nodes cluster similar samples together Multivariate linear regression method Series of algorithms that perform classification resulting in scores plots where the goal is to separate clusters according to group classifications. This is achieved by re-sampling the training set either by: (a) separating it into two sets a training set and a monitoring set. factorial discriminant analysis and the very popular PLS-discriminant analysis (PLS-DA) Multilayer perceptrons (MLPs) are powerful approaches that can map non-linear functions from X-data to Y-data. These algorithms also involve supervised learning but the primary reference data is the level of the target determinand(s). Based on Darwinian concepts. This could for example be with respect to the country or region of origin of a particular foodstuff. Soc. 336 Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. 41. The goal of classification is to identify an unknown sample into the correct group. a food for example.. In supervised learning these primary reference metadata (target(s) or output(s)) are paired with the input data (vibrational frequencies or chemical shifts). There are different approaches to validation of the multivariate model and all use some resampling of the initial data that are collected during this calibration phase. This journal is c The Royal Society of Chemistry 2012 . The goal of supervised learning is to find a model or mapping that will correctly associate the inputs with the targets. the PLSR loadings plot can also be used to extract those features which are related to the determinand being quantified. ng mLÀ1 contaminant) and can be compared with known targets to assess the accuracy of the multivariate model. in PLSR the projection is with respect to the concentration of the target determinand(s). 338 339–341 342 343 Classification and Classification regression trees (CART) Random forests Classification by inspecting the corresponding loadings vectors as these summarise variation in the variables and can highlight which inputs (vibrational frequencies or chemical shifts) are important for such a separation. for example with respect to geographical location of extra virgin olive oil production Tuscany is considered as the most desirable oil as it has perceived higher quality and hence is a more valuable oil. 335. These classification algorithms are categorical in nature and are based on supervised learning (Table 4). while keeping the training pairs together. As detailed above the first stage in supervised learning is model construction.56 The golden rule is ‘‘garbage in–garbage out’’.org | doi:10. 5706–5727 the bacterial load within. References 329. 330 Table 4 Modelling method Principal components analysis (PCA) Hierarchical cluster analysis (HCA) Self-organising (feature) maps (SO(F)M) Partial least squares regression (PLSR) Discriminant analysis Exploratory/ Summative Exploratory Quantitative Classification Unsupervised Unsupervised Supervised Supervised 331.1039/C2CS35138B Artificial neural networks (ANNs) Evolutionary-based analysis Quantitative or Supervised classification Quantitative or Supervised classification Supervised Supervised 337. 332 333 334 332.

However. the model is constructed with the remaining samples in the training set.g. Further reviews are also available specific to this area of course. Raman and NMR.239 samples are selected using a replacement method. NIR.244 Mass spectrometers operate through sample introduction.241 is sometimes used to evaluate whether the specific classification of samples in two designed groups (e. classification or quantification. The remainder (36. This random sample picking process is repeated until all (100) samples are in the training set. To supplement this approach permutation testing240. One can think of this as having all samples analysed (say n = 100) in a bag. Quantification of analytes is typically applied with liquid chromatography coupled to triple quadrupole mass spectrometers (LC–MS/MS) which provides sufficiently high levels of sensitivity and specificity. For the ultimate test this final independent test set should be given to the investigators double-blind encoded so that no one can influence the choice of the best model. 2012. In bootstrapping238.252 The coupling of chromatography and mass spectrometry platforms also enables the detection of analytes in complex samples at sub-mg mlÀ1 or sub-mg gÀ1 concentrations. 6). this will include the same sample multiple times. liquid chromatography. Mass analysers apply electrical or magnetic fields to separate ions of different m/z ratios and include quadrupole. 41. Mass spectrometry has been applied in a range of studies in the food industry.. A single sample is taken out randomly and its number noted. the newly calculated classification model should not be able to predict the classes very well. Distributions can be generated for model accuracy (or error) and compared with those from bootstrapping (tested with the correct Y-data labels). more comprehensive model validation involves bootstrapping and permutation testing. Sample introduction can involve chromatographic or electrophoretic separation of analytes prior to detection (gas chromatography. capillary electrophoresis) or direct introduction to the mass spectrometer which includes direct infusion (or injection) mass spectrometry (DIMS). Finally.org | doi:10. More complex methods include splitting the data into several slices and one slice is used to monitor the model and the remainder used as the training set. Contamination by pathogenic bacteria which cause food poisoning in humans is an important application area with the majority of cases being associated with Salmonella contaminated food. 5706–5727 5719 Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.247 The formation of charged analytes in the ionisation source is a necessary requirement as subsequent mass analysis applies electrical and magnetic fields which will only create separation of ions of different m/z ratios if the analytes are charged (either with a positive or negative charge).172 In summary. Soc.g. time-of-flight (ToF). This modelling set allows the metaparameters (e. Ionisation sources include MALDI.g.243 Mass spectrometry Another powerful analytical technique is mass spectrometry which operates by the formation of positively or negatively This journal is c charged analytes (ions) and subsequent measurement and detection of the mass-to-charge (m/z) ratio of these ions. electron impact ionisation (EI) in combination with gas chromatography (GC). In adults. This can be significantly longer (minutes to hours) than for other techniques including FT–IR.245 desorption electrospray ionisation (DESI)246 and matrix assisted laser desorption ionisation (MALDI). these methods require extensive sample preparation and therefore time from sampling to acquisition of results. number of latent variables to use) of the model to be estimated accurately.251. country 1 versus country 2 of origin) is significantly better than any other random classification in two arbitrary groups. and the sample is placed back into the bag. This process is repeated until all samples have been left out once. though detection of food contamination is by far the most frequent application. and the interested reader is directed to a review of supervised pattern methods in food analysis for example by Berrueta and colleagues. triple-quadrupole. often combined with liquid chromatography (LC). DESI) sample preparation and allow rapid analysis times. the chemometrics pipeline has many stages and there are many chemometric approaches that can be used for data analysis.rsc. and on average 63.242 This null distribution is formed by performing many permutation tests (typically 1000 times). LC–MS/MS platforms are the gold standard for performing quantification in the food industry and many other industries including the pharmaceutical industry for regulatory purposes. Indeed.248–250 GC–MS platforms are also applied for analyte quantification. the model is then tested by comparing the predictions of the sample left out with the known real target.g. ion formation in the ionisation source. A definition of common chemometric terms can be found in Table 3. DESI. Other mass spectrometry methods can be applied with minimal (e.View Online In the simplistic method leave-one-out validation is used where a single training pair is selected.1039/C2CS35138B The Royal Society of Chemistry 2012 . ion separation based on their m/z ratio in a mass analyser and finally ion detection. MALDI) or no (e.2% of all of the samples will have been selected for training.8%) are used as the monitoring or test data.. After the model is constructed using the approaches above it is tested for generalizability using an independent test set not used in model construction. though generally with lower sensitivity and specificity compared to the targeted methods for analyte quantification.. Targeted analytical methods of high specificity and sensitivity are available to quantify analytes of interest accurately. ion trap. Orbitrap and Fourier transform ion cyclotron resonance (FTICR) mass analysers. The rationale behind the permutation test is that with the wrong class labels. magnetic sector. large doses are required Chem. using the same spectroscopic data but with the Y-data (‘1s’ or ‘0s’) randomly assigned (permuted) to different countries of origin. Rev. This bootstrapping process is repeated many times (1000 times is typical) and all accuracy and precision statistics are generated on the distributions from the test data sets only. The code can subsequently be broken by a third party and used to check the model’s performance. When combined with chromatographic separation this provides the ability to analyse complex samples containing analytes of similar or identical masses and physicochemical properties with specificity provided by mass and chromatographic separation (Fig. and a summary of the main chemometric themes and methods are detailed in Table 4. These are chosen depending on the analysis type be it exploratory analysis. as well as atmospheric pressure ionisation (API) sources including electrospray ionisation (ESI) and atmospheric pressure chemical ionisation (APCI). this sample now forms part of the training data. Mass spectrometers offer a number of advantages and limitations for application in the food industry.

DESI has been introduced and its potential for the detection of pesticides. the elderly. 5720 Chem. The study of pyrrolizidine alkaloids in honey252 and how these can lead to downstream contamination in the food chain (candy. natural toxins. As mentioned previously. low doses can result in serious illness. baby food and even jelly babies) have been undertaken. and packaging migrants contamination in foods have been presented. 346.org | doi:10. Ultra high performance liquid chromatography – electrospray ionisation – time of flight mass spectrometry (UHPLC-ESI-ToF/MS) negative ion mode total ion chromatograms for (C) broccoli and (D) melon generated according to the method of ref. or the immunocompromised.258 Fumonsins have been studied in cereal products248 and dried vine fruit.257 and their presence in Hong Kong foodstuffs have been surveyed.103 Fig. Therefore. B and C and fumonisins. Rev. produced by S.250 ochratoxins have been quantified in Belgian beers.263 Recently. where it was found that 32% of all samples contained detectable levels. 5706–5727 Mycotoxins are toxic secondary metabolic products synthesised by fungi which colonise food products and can cause health problems ranging from irritation to a weakened immune system to death. Although sample collection and preparation is simple. 345. further research from the same group led to the development of a rapid protocol applying the collection of volatile organic compounds (VOCs) and analysis by thermal desorption into GC–MS. though at concentrations significantly lower than legal limits set in the EU. Xu and colleagues have developed metabolic profiling methods for the detection of Salmonella typhimurium in pork and its discrimination from the natural spoilage microflora. a specific and sensitive analytical platform is needed for robust analysis of the volatilome. 2012. the detection of bacterial contamination at low levels is essential. all of which have been studied in dried fruits255 and chemical structures are shown in Fig.. 41. This journal is c The Royal Society of Chemistry 2012 . Mycotoxins include the aflatoxins.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Aflatoxins have been detected in brazil nuts. 6 2-Dimensional gas chromatography – electron impact – time of flight mass spectrometry (GC Â GC-EI-ToF/MS) total ion chromatograms for (A) broccoli and (B) melon generated according to the method of ref.102 whilst this was not a particularly rapid method. adulteration.rsc. typhimurium contamination of fresh and aged beef strip loins253 and contamination by other natural products which present health risks have also been investigated.254 as well as the study and detection of specific mycotoxins. typhimurium and this VOC was able to differentiate between natural spoilage and contamination by this pathogen. food additives. One method developed the extraction of metabolites from meat samples followed by analysis by GC–MS.256 as well as barley grains. soft drinks.259 Other food contaminants investigated include endocrine disrupting chemicals.246. ochratoxin A.261 epoxidised soybean oil in baby food.264 The power of DESI comes from its ability to be coupled with hand-held or on-site mass spectrometers for screening of contamination during sampling before further detailed studies in food analysis laboratories. Bhattacharjee and co-workers have also applied the analysis of VOCs to detect S. Authentication of food products have included the use of isotope ratio mass spectrometry (IRMS). 7 shows the mass spectrum for dimethlytrisulfide.251 polycyclic aromatic hydrocarbons (PAHs) in milk262 and drugs including furazolidone residues in pigs. Soc. 8 for these three types of mycotoxins. veterinary drugs. power bars and cereals.260 Sudan I dye. to cause illness but in children.1039/C2CS35138B Fig.

267 Inductively coupled plasma-mass spectrometry (ICP–MS) has also been applied to authentication and traceability studies for high value hazelnut products.278.287 A related Chem.274.265 This is important in authentication and adulteration studies to provide accuracy. 7 Volatile organic compound-derived MS spectrum of dimethlytrisulfide from Salmonella typhimurium contaminated pork from ref. 5706–5727 5721 The Royal Society of Chemistry 2012 .281 and forensics282–284 for example.org | doi:10. C3 and C4 plants use different photosynthetic pathways which produce different levels of isotopic fractionation and small but different 13C/12C ratios. Rev.270 milk powder. In terms of food analysis. DART–MS is said to be a rapidly developing technique that can be used for the target and non-target analysis of complex food matrices It is seen as a simple and high-throughput system for the qualitative identification of chemical species. noncontact analysis of materials at ambient pressure and at ground potential. This journal is c Additionally. and ultraviolet light stabilizers. and in addition. These structures are taken from the PubChem database.266 Adulteration of natural vanilla flavour with synthetic vanillin can be determined using IRMS for determination of changes in hydrogen (2H/1H) and carbon isotope ratios. metabolomic fingerprinting/profiling285 and quantification of low-molecular weight food components such as trace organic contaminants. IRMS studies have been performed to assess fossil fuel-based changes to the carbon isotope ratio of atmospheric CO2 and how this is affecting the carbon isotope ratio in plants. These differences in the carbon isotope ratio are employed to differentiate between foods sourced from C3 or C4 plants.277 When installed in a time-of-flight MS. 103. Fig.268 Authentication of oysters according to geographic location has been assessed using GC–MS and pyrolysis-MS269 and adulteration of milk. similar to that reported above for SNIF-NMR. 41.271 coffee272 and buffalo mozzarella270.rsc. including the ‘Tonda Gentile delle Langhe’ (TGL) variety from Piedmont in Italy. antioxidants. Temporal changes in atmospheric CO2 have been shown to correlate with temporal changes in plants. grease-proofers.1039/C2CS35138B Fig. the source has been said to provide improved selectivity and the assignment of accurate elemental composition through exact mass measurements. 8 The chemical structures of three common mycotoxins found to contaminate food after fungal growth. colorants. is able to detect directly chemicals on surfaces without the need for any sample preparation.279 fragrance industry. Soc.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. pyrolysis-MS has been applied to adulteration studies of milk and orange juice.276. including plasticizers.280.273 have all been reported. Adulteration of royal jelly has been detected using changes in carbon and nitrogen (15N/14N) isotope ratios. 2012.286 Some of the food-related applications of this technique include the rapid screening of food packaging for the presence of additives. such as solvent extraction.275 Another relatively recent development in mass spectrometry has been the introduction of the direct analysis in real time (DART) ion source by Laramee and Cody for the rapid. related to adulteration with high fructose corn sugar (using carbon isotope ratio) and yeast powder as a protein source (using nitrogen isotope ratio)..276 DART–MS has been applied across a wide range of areas such as the pharmaceutical industry.

145..297 Fig. 9 The interdisciplinary approaches that are typically combined within a multidisciplinary framework for the detection of food authentication.320. 2007. P. D.308.2-cyclohexane-dicarboxylate. Wareing.300 It is also inevitable that some potential food supplies will be diverted towards biofuel production. This journal is c The Royal Society of Chemistry 2012 . 9). Williams. In addition to the technologies mentioned here. Frost.309 which not only involve the impact of climate change on food production systems (and food security).1039/C2CS35138B study analysed the PVC gaskets of lids to glass food jars. 41. Jorgensen. 151–164. Bolton. A. R. the amount of food wasted globally would potentially be enough to sustain the undernourished human population on the planet of almost 1 billion people. many others of course exist and deserve to be mentioned.307 All these are in addition to continued concerns regarding quality. and polyadipates.22.319 thin film sensors.293 identification of contaminants in soft drinks. R. for funding her PhD CASE studentship.. These concerns readily illustrate the requirement for continued development and refinement of current analytical methods. such as electrochemical detection. will affect the world’s ability to produce enough food. Kenawy. to assess the capability of DART–MS for real-time analysis of compounds in natural raw materials and compare the efficiency of this technique to the traditional HPLC method.290 quantitative analysis of multiple mycotoxins in cereals.. 9).299 leading to very real concerns over future food security. Food Microbiol. S. monitor.View Online Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. VLB would like to thank BBSRC and Avacta Plc. and analyse food systems at all stages from farm to fork.298 These demands and competition for resources. References 1 J. concerns which will doubtless increase concurrent with population pressures. and that fossil fuels are likely to be depleted in the next 70 years. fatty acid amides. and that the technique could be applied as a chemical profiling and semi-quantitative method for the quality control of a range of medicinal plants as well as food ingredients.301–305 Other potential issues involving food production systems include the threat of deliberate contamination of the food or water supply by bioterrorism43–45. J. JWA would like to thank the EU for experimental funding as part of the Framework VI initiative within the META-PHOR project (FOOD-CT-2006-036220). and physical sciences (Fig. With a limit of detection of r1% DART–MS was able to identify a range of chemicals in PVC plastisols.318 nanosensors.312 If addressed and distributed properly.306 and biotracing of harmful pathogens in food chains. J.316. from which additives could migrate into foodstuffs. 3 E. authenticity and adulteration.294 olive oil authenticity295 and dairy produce. An often overlooked fact for example is the energy budget used in food production and the huge amounts of energy lost through food waste. A significant proportion of 5722 Chem. F. 2011. On top of these there are of course the omnipresent potential future impacts of climate change. Broughton.324. an area of research covered within this review. L. Bailey. 2 F. Ward and T. 76. A. Food Microbiol. 2012.291 authentication of animal fats and minced meat adulteration.296 including the detection of the adulterants melamine and cyanuric acid in milk products. Henderson.288 One study analysed antioxidant curcumins directly from the turmeric plant (Curcuma longa) and turmeric containing beverages and curry powder. epoxidised soybean oil.322. 1359–1384.rsc. Zentek. 5706–5727 this waste is food spoiled prior to it ever reaching consumers. Acknowledgements DIE would like to thank Chemical Society Reviews for their kind invitation to write this article.325 and protein quantification via ELISA. the referees for their invaluable comments and suggestions for improvements.292 beer brand profiling. J.27.311 equivalent to 350 million barrels of oil per annum. tributol O-acetylcitrate dibutyl sebacate.323 and already well-established techniques such as DNA detection233. D. Results showed that DART–MS could be used for semi-quantitative determinations of curcumin from 5–100 mg mlÀ1. it is clear that research and novel developments will continue a pace. J. 8.298 In addition to the fact that potable and irrigation water are also in decline in certain areas.289 Other studies include the rapid and direct screening of wheat for target fungicides (strobilurin residues) in less than 1 min. S. bis(2-ethylhexyl) adipate. Biomacromolecules. 1. Worley and R. Humphrey.68. Int.326 In this exciting era of systems-level thinking and the resultant interdisciplinary collaborations between the life. but also the way food production systems and consumer consumption itself can have an environmental impact310 and have the potential to further enhance climate change. RG and WD would like to thank the UK BBSRC and EPSRC (BB/C008219/1) for financial support of the MCISB (Manchester Centre for Integrative Systems Biology). Rev. R.org | doi:10. Soc. J. S5–S6.321 nanoparticle detection systems. to the detriment of those who wish to interfere negatively with our food systems for whatever end.317 ultrasound. engineering.313–315 electronic noses and tongues. as well as the development of new technologies (Fig.and diacylglycerides. whereby global population estimates are in the region of 9–13 billion. and Steve O’Hagan for assistance with graphics and software. contamination and spoilage. acetylated mono.2-diisononyl 1. Conclusion The continuing rapid increase in human population means that global demand for food resources will continue unabated for at least another four decades. Int.264. 2002. in the US alone between one-quarter and one-third of all the food produced is wasted. including phthalates. to allow us to measure.

W. 2008. Kramer. Burton. 8076–8088. 2008. 2001. 1855. Z. K. D. S. H. N. Naumann. R. Wiser and R. 5706–5727 5723 . 69 M. E. J. Vennard. 2006. Andre and E. O. D. G. B.. Yu. M. I. 42. Gounadaki. Food Sci. 58 R. N. M. Bioanal. Accum. Hassal. 68 J. Commun. 2.. 1319–1325. K. A. 81. 63 D. 2011. 43. 1562–1569. J. 23–41. S. Springer. L. H. Yarwood. 68. A. Pharmacogenomics. Nesbit. O. 24 M.. Cambron. P. Fix. Hens. 35.. 77–92. Guardian News and Media. H. 42. Analyst. Ledoux. S. M. R. Rooney. 2005. Qiao. Gossner.. 38 J. B. Feigenbaum. 690–700. Rapid Commun. Woodward. J. Hocquette. 137. Food Microbiol. Schimmel. 1994. TLC. in The Guardian. Contam.. Skandamis. E. D. 1157–1180. Motlagh. 20. 84–95. Berdal. Vittoria. 4716–4723. 36 S. 59 R. Hollywood. Careri. R. Spraul and E. 2000. Clin. 1243–1266. Chem. Z. 523–531.. R. 1997. N. 76 N. 2008. E. 60 D. J. Y. Zeidler. Lai. 2012. 1998. 46 R. 2009. London. Belessi. Analyst. N.. Baker. J. Mustapha. Appl. J. S. L. Yang. J. H. R. Nuernberg. I. 1118–1122. Schvartzman. Kaderbhai. Covaci. M. Contam. L. K. K. Koopmans and E. Ellis. M. I. H.. N. W. 1820. Dunn and D. Varelis and R. Morales. 2757–2765. 24. J. M. Sci. Health Perspect. G. 196–196.. 1657–1667. 72 P. Mura. 29 Anon. 4141–4152. Am. S. Kemsley and R. 1045–1056. 152. 1991. Food Chem. Awika. P. Tang. Brand-Miller. Food Chem. 2005. Infect.. Environ. B. 1998. 2011.. Clin. J. 2008. P. H. Appl. K. 285–294. Vlietinck and G. Health Perspectives.. Caccamo. A. Nicolaou and R. M. Wagner. P. G. 21. Environ. 2003.. Skiby. Wall.. 47 B. 2004. 140. Puschner. Natl. 15 C. 323. Helm. C. Zagnoni. J. A. Stuart. Food Protection. Yang. J. Chen. P. G. Evergreen Review. 921–925. A. 8 N. 42 A. 43 J. 1986. Toxicol. E. Dehareng. K. Chem. Coley. Pestana and E. Hodge. R. New York. AOAC Int. 30 Anon. 8. 2007. Ellis.. 859–865. L. 351. Dunn. 129. Lebensm. London. 95. R. 2004. I. Reimschuessel. G. Miraglia. Poser. Environ. F. G. D. Chichester. M. J. Zenilman.. J. J. Harrigan. Int. Microbiol. Ellis. J. Richardson and A. M. Swart and D. 48 W. S. M. 919–935. R. A. Technol. I. 24. Allwood and R. 22 E. L. 42. Planar Chromatogr. Tefera and I. 2007. E.. De Pauw. Cubedo-FernandezTrapiella and R. 22. Acad. Goodacre. 12 R. Q. Pullen. 14 J. Jordan and P.. 2006. Ben Embarek. Angulo. H. 2001. D. Xiao.org | doi:10. J. 6. G. Chromatogr. R.. M. G. 20 S. J.. Microbiology. R. 51 D. G. 23 B. C. S. P. Huang. 61 D. Wang and X. 2002.. Dobson. Munoz. 109. 66 A. 101–103. Vergnaud and A. Brison and R. S1–S4. Quijano. S. Bernheim. Lo-Fo-Wong. pp. Focant. J. 87. FEMS Microbiol. Kok. L. I. 16 R. Rentsch. Muniz-Valencia. P. 9 A. 1208–1215. McParland. W. Poppenga. T. J.1039/C2CS35138B This journal is c The Royal Society of Chemistry 2012 Chem. van Rie and J. Scollan. 17 M. J. 57 K. 2007. Tritscher. J. Beltran. M. J. Wishart. Rapid Commun. A. P. J. A. M. Food Microbiol. 2002. E. 373. 69–79. 1991. 116. 70 M. Timmins. S. 1157–1170. 71 M. T. X.. L. Griffin. 52 D. 117. Rev. C. 57. Wilson.. in New York Times. R. 540–544. M. J. Forsch. L. Nature. S. S. C. Hewinson and R. A Treatise On Adulterations Of Food and Culinary Poisons.. A. 19 D. A. Yurawecz. Mann. D.. Zhu. Ellis and R.. 911. Goodacre. 2009. 71. 1424–1431. 25. J. Borda and S. Gordon. Naumann. 265–273. 75 H. Tor. Bernert. Rapp.rsc. Metabolomics. Goodacre. Rasco. R.. Int. Kim. T. A. Heller and C. 40 A.. 90. and Methods Of Detecting Them. Covaci. 28 L. J.. 4. E. 3–14. B. M. Zeng and R. Agric. 25 R. A. 559–561. 2011. Sebastian. 171–196. 41. L. J. 1998. P.. TrAC. D. S. Garber. 9984–9989. 65 C. T129–T134. Chemosphere. Surf. Laoubi. 590–594. Food Chem. Ed. Ho. Rowland and D. Analyst. M. 1987. 1174–1175. Food Agric. Lin and B. R. Daston. 45 L. Sheridain and V. J.. Acad. V. Mangia. 32 P. 145. A. 167–179. Biodeterior.. Pocchiari. 469–470. 354. Santos-Montes. 2008. Hourant. A. Broadhurst and D. 123–130. Microbiol. J. 74 M. Dardenne. Rooney. 75. 2008. Food Addit. Guo. Goodacre. Lancet. X. 50 D. Cooks. Proc. J. D. Akins. S. Regg. in The Times. pp. S.. Goodacre. 54. Delmonte and J. Food Control. P. 1900. Trends Food Sci. Sorrentino. J. 2009. 233–239. 33 N. Labischinski. Technol. N. 2001. 74. 2004. 102. Alperovitch. A. 19. L. Mass Spectrom. Agric. T. Helm and H. Jordan. Prusiner. 2008. Kell. Chingin and R. Natl. Interface Anal. G. 31 J. Will. J. 2822–2828. M. Timmins. Lancet. Schlundt. Shan and G. S. M. Proteomics. Wages. K. N. 1986. M. 2004. Rosales-Martinez. M. 205–208. Y. 41 F. D. 3624–3632. Y. J. H. V. 2009. H. Chem. J. Hu. Chem.. Chromatogr. N. Guillen and N. Filigenzi. Estibeiro. ed. McGorrin. M. 1996. Gorrasi and V. G. S. C. Environ. Dale. Labischinski. 556–558. 13 T. Hird. W. 109. Trends Food Sci. 1996. 81–82. Anal. Corbisier. H. Schallehn and D. Xiao. Lu. R. Boston. T.. Lin. O’Keefe and J. Perkins. 35 H. Jeskelis. S. 62 X. 2009. Fan. Pistara and I. 2008. Zagon. 2007. Food Sci. 482–493. 13363–13383. Soc. 49 D. Aston and B. Food Additives Contaminants Part A. 2005. Penninks and W. Baeten. 131. M. 18 G. Trends Anal. Polymer. M.-Unters. B.. S. Mossoba. Humpfer. Schepens. Gen. 1–11. 94. I. Luo. 1803–1808. Rowland and R. Q. Bigalow-Kern. Riquet. W. Lancet. Kluwer Academic. Meurens and R. Gengler. 2006. 2008. J. 1998. M. 10 Anon. Goodacre and G. B. 2008. J. in Metabolic Profiling: Its Role in Biomarker Discovery and Gene Function Analysis. Broszat. J.. Dordrecht. L. Massart. C. 2011. H. Blatny. B. 341–354. Lett. Balicer. M. Int. 145. G. E. Q. Cordain. Dicorcia and R. 347. Winder. Wang. 2008. 1984. 7 F. Brera. 313–316. B. 1154–1159. Calbiani. D. D. Meat Sci. Pirard. 298. S. P. J. 5 C. Coffey and P. H. Nochetto. B. van Larebeke. Aparicio. S. Schmitt and H. AOAC Int. Teoh and A. 2006.. R. R. M. J. A. 39 M.View Online 4 A. K. 73. 41. H. N. 2004. in Risk Assessment and Risk Communication Strategies in Bioterrorism Preparedness. Greene. Harrigan and R. 478–481.. 55 C. S. J. Schepens. S. Dis. 11 Editorial. B. 1–11. Biological Applications of Infrared Spectroscopy. Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. 2006. B. I. Izquierdo-Hornillos. Crump. Eaton. U. 15. L. J. 392. 52. 2009. 4027–4032. Eppe. 182. Commun. I. Juranek. J. 2008. Inc. Mass Spectrom. 26 P. ed. 2002. P. 6 L. N. Duizer. Soyeurt. Swerdlow and D. Cousens. 875–885. M. De Poorter. Winder and R. D. G. Kell. The EFSA Journal. Green. 2009.. Hofman and P. S. 2001. Int. H. J. U. Seinen. 27 C. John Wiley & Sons. Flemming. L. 111–124. 608–608. Zheng and R. L. Wolff. Toxicol. Goodacre. Hoorfar. D. Food Addit. Spectrosc. 73 Y. Holst-Jensen. He. Zhang. F. Biodegrad. T. Zheng. 44 A. 419–421.. Liu. A. L.. Hens. B. A. L. A. Ironside. Watkins. J. Nature. E. Goodacre. Dannenberger. Cohen. Proc. Public Health Reports. Ehling. G. 21. K. 91. W. A. Kammuller. 251–262. Mass Spectrom. Marvin. G. Khan. M. Kell and P. Baeyens. H. Chen. K. J. Zeng and X. 42. Elviri.. Times Newspapers Limited. 2007. R. 1168–1174. Wein and Y. S53–S60. Food Microbiol. Bouquin and J. Overmann. Sci. A. Food Bioprocess Technol. Everaert. Moloney. 353–353. 54 J. A. B. 18. C. 1997. Samperi. 106. 48. B. F. F. A. 164. Branigan. Lancet. K.. J. Lindeberg. D. Sci. 133.–Mod. 144. 2007. Nutr. 53 N. Ceballos-Magana. D. Zenobi.. Ellis. Goodacre. Sodhi. Scharpe and N. G. 42. 67 J. L. M. Berry. A. T. Dairy Sci. Veterinary Record. Goodacre. 2011. Broadhurst. E. Delapaz. Goodacre. Al-Qadiri. Kell. Chem. Zhu.. 1042. van Larebeke. 56 D. J. Kilbourne. J. E. Jorens. Sci. 1982. Bramer and B. G. Health Perspect. A. 2008. 408–413. P. Li and A. J. Bushway. E. J. 2005.. S. 37 R. 46–49. Gamez. P. 17–33. B. Microbiology. 34 F. 21 M. G. Chen. S. 64 C. MacCartney. Griffin and F. Smith. Cabo. A. A. J. Meredith. A. Gonzalo-Lumbreras. Spangenberg.

M. Rutledge. 2009. VoDinh. Scherer and H. 139 T. von Holst. J. V. 447–450. Tan. Song. J. U. Polesello and R. L. Academic Press. 70. Dunn and R. 120 D.. Wu. 2006. Goodacre. Azizian and J. Alonso and M. G. Boubellouta and E. H. 7851. 8. 2012. Goodacre. I. 2001. A. Clarke and R. O’Donnell. 129 N. 82.. 2011. A. Goodacre. Near Infrared Spectrosc. Mossoba. W. G. W. Cheung. Anal. Boix. Seiler. Daniels. Soc. 55. Schwagele and H. Panagou. Sedman. M. Xu and Y. A. Y. 2439–2449. Kiatsoonthon. 193–201. Mauer. Chem.. Zhou.. De Baerdemaeker. V. Deering and R. 84 S. B. Reid. 75. I. 18. 37. Food Eng. G. Yu. 56. M. 1997. G. 2008. Near Infrared Spectrosc. 2002. 2004. F. Winston and J. 81 M. Kramer. Soc. Krska. 14. 1998. R. Biotechnol. pp. H. 2004. J. Garde-Cerdan. Q. Cozzolino. 5706–5727 This journal is c The Royal Society of Chemistry 2012 . S. S. A. B. P. Chem. Cynkar. F.View Online 77 A. 98 D. Cattaneo and S. P. 115 L. 2011. S203–S209. S. Fumiere. 122 B. 137 D. 88. Han. Kos. Mossoba. J. He. Lieberman and G. Da Costa. Z. Hiatt. H. E. 2009. Trends Food Sci. J. Robinson. Chem. Anal. Wang. B. J. 448. M. Spect. Winder. Li. 148–155. 135–140. 39. 11520–11525. Appel. 2011. Cozzolino and I. Q. Food Chem. Soc. J. Trans. A.. A. J.. 107–115. McAllister. P. 5. 2004. Ismail. Clin. F. 1983. 91 G. Zock. Food Bioprocess Technol. 30. Oil Chem.. 447–452. Spectrosc. L. 2009. Lo. M. Technol. D. Barzaghi. Commun. G. Yu and W. ed. K. Appl. Zhang. 1997. 1998. Z. Am. S. 128 T. Soc. C. Food Chem. A. 113 T. A. O’Donnell. 2009. 34. 8024. A. M. Appl. Wylie-Rosett. 2010. R. 92 M. Jarvis. Near Infrared Spectrosc. L. J. J. Woodcock. Chan. Food Chem. 2008. 273–284. Banwart. Lorenzo. Agric. Food Chem. Dugan. Spectrosc. N. Food Chem. 18. F. 2006. G. Delgadillo. Nie. 2011. 338–347. Goodacre. 138 D. W. W. vol. Chao. Bergman and C. Franch. Cen and Y. E. Kwiatkowski. Zhou. 125 D. Spect. Xu and R. 102. N. R. A. J. E. H. Maiwald.. W. Wilson. 2010. 87 T. Agric. Chemical. 45. Huang.. J. 123 M. 143 D. Gishen and M. Li and S. Technol. 2012. Thanapase and A. 9. Alexandrakis. Rev. Katan. Anal. 2011. J. Newhouse and L.. J. M. M. Barros and D. LWT–Food Sci. Sinnaeve and P. 133 S. S. P. 303–313. J.-W. 117 H.. Chem.1039/C2CS35138B 5724 Chem. 414–424.. Kasemsumran. N. T. 1642–1651. P. Aalhus and B. 107 B. R. Wilson. Wang and L. 63. Zalacain. A. H. Murray. Food Chem. 116 D. 16. 104 A. Uttaro. Sci. Kummerle. 17. 2010.. Appl. A. Ruan and W. ASAE. Downey and C. D. J. Parker. O’Donnell and G. G. Technol. M. B. M. E. Food Eng. Milosevic. He. Bansal. Ellis. Technol. vol. H. G. M. 2007. Rui. Gu. Steinhart. Food Chem. 195–201.. P. Y. K. Birkel and L. G. Anal. Am.. Veys. 1477–1483. K. Scannell. D.. Am. M. Yang. 118 Y. Cambridge. 329–338. D. Li.. Lopez-Campos. Xu.. 100 A. Blecker. H. D. J. Dambergs. 2001. Downey and A. 108 D. 88. Olsen. A. 86. Broadhurst. 2012.. Rasco. Pike. London. 1121–1125. A. X. J. in Information Optics and Optical Data Storage. Cavinato and M. 59. 12. P. 97 O. Xu. Food Chem. Lett. I. K. Environ. Chim. J. Spie-Int Soc Optical Engineering. M. Nutr.. B. D. S. vol. Z. 1st edn. Huang. Brosseau. 85 Y. 2012. Downey and C. 2011. S. New Yok. C. J. 135 Q. Carnethon. M. He. G. T. J. C. F. H. C. S. Heia. 59–70. Cho. 13. Wang and H. 2008. Baeten and G. Cuello. in Magnetic Resonance in Food Science: A View to the Future. 2003. Huang. J. Kramer. T. Belton. P. M. 132 O. A. H. Fearn. Spectrosc. 52. R. 96 D. N. Am. LWT–Food Sci. 3974–3980.. A. Yu. E. E. 39. 401–404. p. 9843–9847. Spectrosc. 2005. A. X. 2207–2214. V. 73–80.. P. S. 50.. Kim.. 131 L. in Food Spoilage Microorganisms. Bhanger. I. Howard. 1037–1045. and Biological Sensing Technologies VIII... E. 40. Kemsley and R. M.. 88 R. 401–405. Rev. Huang. Jutamulia. Wu. R. J. K. 102 Y. Talanta. Chen. Harroun. 3.. 127 Y. Agric. P. Food Sci.. H. Nychas. P. B. 2008. Actuators. 2000.. Polysiou and G. 46. H. 82 E. 508–514. 130 R. Harris. Rutledge. J. 41.. Cerretani. Dolmatova. 82. 2005. Oil Chem. Analyst. 2012. Tao. L. 73. Cambridge UK. Mossoba. 2011. K. M. Azizian and H. Spectrosc. W. A. Rudel. Sumpf and H. Oil Chem. Sacks. R. Lefcourt. Bioanal. 114. Griffith. Zhong and Q. A. 85. Legrand. 1581–1583. E. L. Anal. Food Chem. P. Steinhart. Food Chem. 17.. Lefevre. Nutr. Downey. 2003. 513. 1325–1329. N. Ellis and R. Prieto. 24. 83. 855–867. Dufour. A. 114 Y. K. 2005. Meat Sci. Y. B. Murray. Gu. Barros and R. M. Dupuy. 5. Food Bioprocess Technol. Lin.. M. L. pp. Microbiol. Cheung. K. Y.. G. Dardenne. 112 H. Rasco. Y. Z. Nychas. G. Y.. R. Seiler. Y. F. 6144–6168. Ruckebusch. Luo. Kronfeldt. 87. ed. 2009.. Ellis. 56. Spie-Int Soc Optical Engineering. Karoui. Zhang. 2011. T. Am. Goodacre.. Microbiol. Mousavi-Hesary. 136 I. 2010. Near-infrared spectroscopy in food analysis. P. Circulation. Jin. M. J. 562–568. 145. 2003. 262. Pierceall. ed. Anal. Kim. L. L. Cozzolino. Anal. 1243–1247. Berben. 125. I. 103 Y. Food Chem. H. Jaillais. 57. 109 H. J.. L. Near Infrared Spectrosc. J.. van de Voort. 121 D. Al-Holy. Environ. Elsevier Academic Press Inc. S. 90. Lichtenstein. Y. R. 43–51. M. Huvenne and P. L. 134 M. L. Q. R. Spectrosc. 621–640. Cavinato and B. Davis. 1648–1654. Food Chem. Nicolaou. Lipids. J. 86 G. 2008. Technol. 83 I. pp. in Advanced Environmental. 2012. Goodacre. P. Immink and K. L. Osbourne and T. 144 C. B. Yue. Oil Chem. 76. Kris-Etherton.. S. 2011. L. 2010. 89 N. Food Sci. ed.. L. Kronfeldt. 2007. 130. J. Schmidt. 179–192. Spect. Li. A. 77. 99 M. F. X. A. Anal. Milosevic. 140 H. A. Ammor. Gauglitz. Downey and C. G. Webb.rsc. A.. 125–131. M. Ying. 1146–1155. 907–910. Food Bioprocess Technol.. C. Khurana. Anal. Al-Holy. L. T. M. Vib. 2009. S. Giangiacomo. J. 2009. Rodrigues-Saona.. Agric. 2010. 153–186. G. 57. Bellingham. G. 2005. Kim and D. A. Hu and M. Tarantilis. Franklin. A. D. 25. C. Sens.. Meat Sci. H. Blackburn. Neo and H. E. R. Sun. X.. D. O. W. 107. C. Kays. J. B. Fumiere. Mahesar. Trends Food Sci. Ag. K. Rodriguez-Saona. H. 397. T. Sowoidnich. M. S. S. Downey.org | doi:10. A.. I.. C. Anal. Karanja. Broadhurst and R. Ellis and R. M. 878–882. Gil and I. 126 O. 80 M. M. Kim and S. 1289–1293. Kim. Chim. 3–27. Trends Food Sci. W. Hu. Yurawecz and R. 2002. 514. Agric. M. 344–353. Kim. 2006. 4327–4333. 84. Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs.. Soc. J. 1760–1764. 50–58. Winder and R. 27–34. J. 78 R. Lin. Buning-Pfaue. ed. Bendini. 94 O. Woodcock. Royal Soc Chemistry. Brands. 1996.. Jun. 146–154. Kandhro. 111 A. 2007. Kemsley and R. A. E. Woodhead Publishing. in Advances in Applied Microbiology. N. San Diego. 79 M. A. Sowoidnich. 101 D. Comb. W. 2027–2037. Wang. Shono. A. 278–286. H. CRC Crit. Seiler. A. 116. R. C. 141 A. 2001. Acta. J. T. Goodacre and S. Pinto. 2010. H. 102. 2010. Zhang. Acta. in Infrared Spectroscopy for Food Quality Analysis and Control. 5681–5687. 142 L. 1003–1009. H. 297–311. 2009. 39–46. Goodacre. Aucott and I. 123. 105 K. Kays. Van Horn. B. 2007. Chem. Cao. 119 I. Herderich. S. Wu. J. K. Sone. Balabin and S. Wiley. P. L. Soc. Kester and M. R. B. vol. Sherazi and M. Chem. Bellingham. 2004. 110 B. Herrero. Al-Qadiri. 1–29. J. Smirnov. Lohninger and R. 1986. 64. McDonald. G. S. R. M. D. A. A. Argyri and G. Chernyshova. Rannou and G. S. P. Dahl and K. Barr. E. Rev. 507–514. J. Karoui and J. M. 110. 98–102. Appl. Salinas. M. Argyri. 95 H. 124 T. 90 L. Mensink. 136.. K. Al-Jowder.. 2004. X. Himmelsbach and S. L. 72–83. Food Chem. X. 8187–8193. Kramer. 203–230. E.. S. 145 W. S. A. 1211–1217. Liu and M. Agric. Bradley. B. 2011. Giangiacomo. M. H. 81. Anal. Schmidt. 2009. Analyst. High Throughput Screening. 93 R. C. Xu. 23. 106 K. Al-Jowder. 535–541.

Remaud and G. Appl. M. in Xxii International Conference on Raman Spectroscopy. Le Dreau. 1989. Huang and Y. Bailey. Kochhar. Anal. 195 S. Soc. P. Q.. Jones. 157 B. Kyrikou. in Biophotonics: Photonic Solutions for Better Health Care. Eur. 2010. Zervou.. Sowoidnich and H. 2003. Int. R. G. L. Spraul. Radt. Giansante. M. M. 2004. 5447–5454. 2007. G. C. Magri. W. ASABE. 62. Narayanana. 2004. 170–176. I. 2008. 37. P. Estrella. 183 A. M. A. Q. pp. 5706–5727 5725 . N. D. Herrero. 1194–1198. L. O’Hagan. 212 J. 61. Stamler. 2010. 297–307. 215 J. Rev. 2008. Golman. Rev. J. Kosir. 44. Agric. A. Gil. Elliott and J. 180 C. Food Chem. J. 4. 424–430. Intell. G. M. 57. Atherton. Deckert. Hernandez. 41. 81. Appl. 359. G. W. 177 A. V. T. Lett. ed. S. Y. V. 3. H. 196 A. 971–973. Chem. Surricchio... F. H. 51. J. 159–171. G. Korifi.. Y. K. Int. Spraul. 11270–11275. Shende. Bucci and A. Goodacre. C. A. 40. 2009. Ther. S. 194 J.1039/C2CS35138B This journal is c The Royal Society of Chemistry 2012 Chem.. 103. Sharman. 6145–6150. R. Sci. M. Rossmann. Anal. Popp and K. S. 191 F. F. Malekfar. Itzkan. E. R. Detavernier and A. 2002. Holmes. P. Rosch and J. 78. Keller. Nicholson and M. Magri. E. Cancer. Popp.org | doi:10. Spectrosc. El-Abassy. J. M. Mozaffari. Touris. J. Ma. I. Brown. G. 65. Small. J. 206 P. Schwarz. Food Chem. Mavromoustakos. Halsall.. 198 W. Am. Nutr. Artaud and N. A. I. Chem. 888–894. 37. 92. 5293–5303. 151 M. Elsevier. Nat. 2005. Farquharson. M. Xu. Cross. Sadeghi and M. 147 E. M. 2003. Sbirrazzuoli and L. I. 216–220. M. B. B. F. Pap. Tu. Moing. Rev. Dunn. J. J. 188 U. Donfack and A. Asher. 711–717. Rev. Agric. 96. Goodacre. I.. F. Abbasian. C. E. 1992. Heinz and P. Kell.. 256–267. D. Beauvoit.. S. Davies and P. Food Rev. Diano. Lucotti and G. ed. J. Pietsch. J. F. 1996. K. Soc. K. A.View Online 146 J. Thaning. Phys. Song. K. control and physiology. 1996. 40–41. D. 2008. Cambero. Biol. J.. Goodacre. Gomezcordoves and I.. 163–166. Feld. 62. Le Duarte. G. 81. 2010. D. C. Analyst. 2183–2190. 2009. 495–502. 150 R. 152 M. D. Schmid. Syst. J. 2007. 16. 2010. B. D.. 2010. B. C. Salter. Siperko. 114. Qu and Y. Food Res. LiChan. L. 488.. 53. 361–370. Bianchi. Spectrosc. P. W. Farquharson. Lampen. Spectrosc. vol. Anal. Chim. 26.. Chim. Kemsley.. Herrero. 127. Chim. Nutr. 2004. M. M. 2005. C. H. S. Acta. Biais. 921–930. 59. G. 53. B. Photosynthesis: Metabolism. Fleischm... Desimon. M. Y. 40. Soc. Maucourt. Delfino. Schafer. ASABE. 551–561. W. Agric. 182 H. B. Food Agric. Phys. A. Hendra and A. Delgadillo. M. 512–523. I. R. 2009.. 193 M. Y. P. P. S. 214 J. Rosch. Zhao. Liz-Marzan. Anal. 37. 232–240. Guillou. A. 7. Agric. G. Blanch. 2001. J. B. R. Beyer. 207–213. Anal. Food Chem. B. 2008. Caballero. Chem.. S. E. Y. B. de la Hoz and M. J. R.. K. 203 C. B. Humpfer. R. Shi. Kneipp. B. 23–30. Molinet. Pesapane.. Abst. van den Berg. 2884–2894. G. Brown. Kidric.. H. Wang. Chem. 38. A. J. 594–594. Damm. M. Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Griffin. Shadi. Dresser. A.. J. J. Crit. McQuilla. Lee. Qian and S. E. Rolin and A.. 148 S. Chan. I. Food Chem. J. 1751–1764. Carmona. J. Kokkinofta. Caja. 1267. D. Korotic. G. Bianchi. M. Food Sci. L. 1667–1670. Matthews. Y. J.. C. Food Chem. Holmes. 2007. Tuchin and D. 197–201. 57. Glaeser... Food Chem. 79. Chu. H. 175 R. Acad. Defernez and I. F. Tirona and R. 199 M. 164 R. Goodacre. 215–221.. Allwood. I. J. J. 117. Jolicoeur. C. B. D. Y. Anal. Food Sci.. Sensors. Food Chem. Gil. Perelman. Broadhurst. M. Raman Spectrosc. I. J. Petrakis. Nie. Reson. Braumann. U. A.. Technol. F. 35–51. 599.. Agric. Longman Scientific and Technical. Ooghe. Bailo and V. Soc. Jimenez-Colmenero and P. Annu. 2000. Camerlingo. Lepore. F. Shende. 347–381. pp. 2010. 2009. J. Shapira. G. Dumas. Rev. 63. 37. 5587. 2049–2061.. Nicholson. 45–69. 3153–3157. 189 H. Sci. S. J. Gallaher. E. Epidemiol. Technol. 53. Godejohann. I. Spasic. 62–72. J. 955–964. Nature. Ther. 65. 1001–1011. Hildebrandt. Chem. I. 209 D. 6. B. Schmidt. Food Chem. 50. Soc. 187 M. Maksymiuk and S. 200 B. Rev. S. Acta Phys. Pharmacol. I. Chem. Food Chem. C. 2003. 70. 170 R. Dupuy. 185 A. Chem. J. Cheung. H. 205 D. in Encyclopedia of Food Science and Nutrition. Fauhl. 859–866. Chemom. Food Chem. 2011. 6201–6206. 1091–1092. in’t Zandt and M. H. J. U. Phytochemistry. I.. Dunn. R. N. Lett. Clin. V. ed. 207 P. Jacob. Phys. 165 G. A201–A210. F. 645–655. Broadhurst. Inscore. L. vol. 598. Gift. D. Wang and R. Belton. Anal. Nikbakht. V. Y. Tang and C. Chem. vol. Z. Krause. R. J. 1994. Smith. 1985. L. Bailey. Martin. J. J. P. 2009. 197 A. Bothwell and J. 1992. P. Leake. Mortter and M. Chem. 115. J. 2008. N. S. Aparicio-Ruiz. Trends Food Sci. 220. 41. Drexler. 201 W. 7. C. 56. Spangenberg and J. 2007. S. Zerbi. Pharmacol. Herraiz.. Raman Spectrosc. T. A. Roma. 2008. 1457–1462. 2008. Moussa. 178 C. 160 D. 153 W.. Acta. J. J. Kim.. P. 115. Trans. M. 46. W. 1997. Bictash. 156 E. A. 783–826. 211 G. Herrero... S. Phys. M. Ahn. 192 A. Rev. L. 2008. Agric. Goodacre and J. 176 W.. Baumann. R. Bianchi and R. Ogrinc. 2007. Mita and M. Soc. Lizzani-Cuvelier. Wackerbarth. 1992. Kim. J. ed. Leake and R. 174 R. T. Pyrolysis. Meurens. Y. H. Acta.rsc. F. Harz. Technol. B. Agric. 912–920.. 50. I. Carmona. 181 A. Lachenmeier. Lopez-Diez. 173 R. Appl. Agric. Marini. C. 1286–1294. P. Tseng and D. 168 E. 1993. Shockcor. 186 H. Frank. 40. L. 48. D. J. D. M. J. M. 2010.. 167 V. Appl. Colquhoun. K. Cambridge Philos. Anklam. He. Chem. Ziegler. B. C. Ordonez. Maraschin and M. Hanote. Kronfeldt. I. D. in Nondestructive Sensing for Food Safety. Radovic and E. 202 K. M.. 159–170. 922–931. Perezilzarbe. 171 F. Dasari and M. Popp. T. 155 X. 110–118.. 387–426. 1993. Clin. 37. Goodacre. 184 C. Reniero and C. 163 G. Anal. I. G. Int. E. M. Lipert. N. L. Lab. Harlow. 86. M. M. Kim. Krause. Dardenne. 1009–1017. Spraul. 2012. Guillou. P. Lindon and J. K. Gregor. Clin. Baeten.. Kuhlmann. P. 154 K. 215. 1531–1535. K. Kneipp. J. 166 C. Schreiner. Dresser. Garcia-Gonzalez and R. pp. J. J.. 64. Rev. Inscore. 216 B.. 44. W. F. Rev. 1997. 208 A. Kell and G. Cheng. Anal. Huyghebaert. L. 2009. Martin. 1998. 2011. Proc. Leary and S. Schmilovitch. Y. 149 S. Am. M. Griffin. Huang and S. Kell. 436–443. Lazzari. Chem. Zenone. Xu and R. Lees and Y. Sengupta. Chen and S. 2005. Luo and M. 521–527. 1974. UK. A59–A66. M. Martin. 2010. W. 2002. Trends Food Sci. J. AOAC Int. G. S. 161 D. 2003. A. 362–370. 132-CHED. 17. Agric. McIntyre. A. Pierna. M. E9910–E9910. Ooghe. Lee and Y. D. Qin.. J. Y. 204 N. Raman Spectrosc. Magn. B. J. D. M. U. Martel. 159 P. 2009. P. I. N. J. Materny. B. M. 1998. Martin. 2007. 1993. Chao and M.. 2009. M. Pol. Lin. 7285–7290. H. S. Rev. 42. Food Chem. J. K. 1873–1882. P. Y. M. Moskovits. 179 R. Kell and G. L. R. A. and Natural Resources. 238–255. 604–610. 2008. D. Hong. Asher. Irudayaraj and R. Soc. Griffin and J. 213 C.. Tintchev. pp. Champion and L. del Castillo and M. Spectrosc. W. S. Goodacre. Heinzmann. Y. H. Kell and G. Soc. Nicholson. Alvarez-Puebla and L. Rev. Liouni. E. L. Bailey. Martin.. Lawlor. Theocharis and T. Vaidyanathan. 2143–2149. Goodacre. Ellis. 436–443. 81. 42. Harz. Mizrach. J. R. Ko. F. D. 3206–3213. U. 162 H. F. 190 B. J. Bianchi. McGrath. Heise. R. 6991. M. C. 2003. McManus. 42. Cordella. Bioanal. S. L. Rev. Chem. Microbiol. 210 G. Porter. Food Chem. Mod. Natl. M. Int. Naulet and H. Trans. J. C. 158 W. Rapid Methods Autom.. B. U. Colquhoun. 2006. 6862–6870. 1540–1547. Goodacre. D. 172 R. 1998. 2010. 376. 1. Chem. 493–510. 2002. J. Appl. A. Deborde. 2011. 169 H. J. Bianchi and D. Technol. A. 2008. 1996. 1284–1289. Chem. British Journal of Clinical Pharmacology. Food Res. Quality. Dennis. L. Dunn.

M. 282 A. U313–U313. 1087–1096. Efron and R. 910–918. Med. M. Heberger. 29. Q. B. C. Bortolotti. 2011. John Wiley & Sons. S. A. Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Everitt. Hooijerink and A. Malabat. 206. Dixon. Chem. B. R. Muller and W. 89. 321–327. K. 243 L. S. B. Cody. H. D. 2010. Bhavanam and P. 2002. Food Microbiol. A. L. X. Caldarelli and B. Monin. A. A. Madigan. Technol. J. Alonso-Salces and K. Van Peteghem and Y. 56. Welch. 250 A. R. R. I. H. Woolfe and S. Schnurer. 2004.. B. Keppler. J. D. B. Bioanal. L. Food Compos. Paglia and R. Nielen. 139–147. J. Proceedings. Colquhoun. Axelson. Acta.. DiNubile. J. B. Hener. B. 242 P. Am. Di Anibal. Rezzi and D. J. Rev. 296–307. Westerhuis. McCracken. Hooijerink. A. M. Defernez. Buccolieri and C. Chem. 270 C. J. Smedsgaard and J. 39–51. J. M. 2002. 231. D. Rossmann. 2009. 268 M. 262. M. K. 218 D.. L. 234 M. 227 F. 22. T. B. A. Smit. 589–599. L. Newton. Oscar-Berman. Craig. 2006. 5. R. J. M.. Vlahov. Czerwenka. Agric. Samms. 75–82. Forensic Sci. Revilla-Ruiz. 2007. R. 99–105. Ellis. Lolli. G. 81. 2009. Maiuolo. Chem. 262 N. Laurent. Ebbels. Food Additives Contaminants Part A. E. C. Bremer. H. Bessant. 121. Rezzi. Trost. 80. Kempf. Denner. 256 M. J. Broadhurst. 226 A. Meat Sci. Velioglu and S. 236 B. Szekeres. P. Spasic and D. 2. Durst. B. 20. Technol. U. 2005. 766–773. T. A. Axelson. Goldmann. 228 A.. Monfreda. Assoc. A. Zamora. I. 51. A. 222 S. Doetkott and M. Toxin Rev. Standal. 357–374. D. Z. 267 M. D. Guillou. Berlin.. Q. Vis. Di Donna. Trygg and F. E. Frankhuizen. Bhattacharjee. 283 W. 230. J. pp. Food Addit. Mozayani.. Fernandez. Baldizzone. Technol. D.. Appl. Berge. B. Rowland and D. 135–158. J. Oliver. Kristal.. D. 2002.a View to the Future. 257–261. Anal. K. P. Manetti.. Jiang. A. K. T. J. R. 244 W. S. BioEssays. E. Fantoni and C.. Passalacqua.. Simoneau. Ratel. Spectrosc.. S. Food Chem. Y. 1997. G. McCoy and D. 81. Anal. 3404–3408. S.1039/C2CS35138B 5726 Chem.. J. 38–47. Schneider. Microbiol. Luykx. Food Chem. Toth and T. L. Kell. 1216–1223. Steiner. 7. Jeckelmann. R. 55. Baker. E. Greule. 2004. Food Chem. 255 H. M. 2008. G. Larondelle. Postgrad.. R. M. Savoy-Perroud.. 2005. Buccolieri and A. E. C. 011001. Osella.. D.. J. O’Hagan. J. M. 93. S. Shaw. Buehringer. J.. Schleucher. Musso and D. Appl. M. 231–241. Somorjai. A. 261 L. 2006. Schreier. 248 W. P. 2011. 2364–2367. Shintu. K. vol. Lutter. Cajka and L. M. pp. 1472–1481. A. 4640–4642.. J. 2009. S. P.rsc.. Sherwood. Abst. 21. Bioanal. J. H. A. Wilson and I. 238 B. G. Delatour. Yeung. M. 229 M. Bertelli. A. 57. Liang. Agric. Mottier. Cordewener. 50. 2010. P. L. Ding. A. Nielsen. Mazzotti. Cooks. 230 S. Varga. T. Contam. Papotti. Knowles. Herrman and S. Wu. 2011. Guillou. Cramer. 247 N. Cody and J. 30. Soc.. Forensic Sci. World Mycotoxin Journal. Panigrahi. H. 232–238. D. 2012. Int. 220 G. F. Paternostro. An Introduction to the Bootstrap. Dunn. Wulfert. D. Auer and R. 325–331. Kell. 37. Chen. 281 O. 2004. P. J. 219 L. Nas. 2009. Zhou. Laramee. Zheng. Kronewald.. K. Soc. D. Qiao. K. Efron. Beuerle. Bot. R. Smilde. Szigeti. Royal Soc Chemistry. Al-Jowder. Hajslova. 1–13. Martinez. C. Musselman and J. A. 552. J. 2009. 193–202. Heberger. Handl.. Lundstedt and J. P. Jackson. LC GC North America. 14. 348. 68–75. Acta. S. 81–89. 2004. 2165–2171. R. 240 P. Stat. Casuscelli. Masoum. M. Desmarchelier. Mass Spectrom. J. von der Ohe. Giordano. J. 68. A. 58. G. 273–286. S. Anal. 21. 933–941. Chem. Haefliger and N. S. Standal and D. 3559. Anal. 1139–1145. Washington D. R. Talanta. 2009. F. Beger. 28. 1990.. Schonfeld. 2008. Gil and I. 2010. J. T.. 2000. Goodacre. 13–24. Stice. 124. Trends Anal. E. Finstad and M. Oliveira. Kell. 2009. C. J. G. Bertholet. Franca and R. Wittig. E. 3rd edn. Agric. Rutledge. Anal. M. C. B. Anal.. Cambridge. 249 P. J. Trends Food Sci.. 2008. 260 R. C. Carabias-Martinez. Agric. Callao. I. Dane and R. G. J. Scussel. M. Aursand. Kempf. Anal. Xu and R. F. F. Engel. Tibshirani. Hachimi and G. Food Compos. 77. F. 176–196. M.. Anal. S. Chem. 246 M. 26. Curr. 23. Dunn. 2003. Lafargue. 1144–1153. Campos-Gimenez. Mukherjee and D. 278 K.. Garozzo. J. diCamillo. Berdague. Tangni. Beuerle. J. 2009. J. F. J. 756–761. Q. 901–908. S. J. S. Chem. C. G. Food Sci. Rapid Commun. D. 62–70. U. 76. 2010. Monard. I. J. C. Varga. ed. A. van Velzen. Int. T. Anal. W. 235 R. Primrose. 221 I. G. C. J. Am. T. Ruisanchez and M. 52. 286 J. D. Houck and A. 259 J. 245 J. Exp. 22. Hu. Reniero. J. 542–548. 20.. 2297–2302. T. M. Opin. Chung. S. N. Nicolaou. Food Microbiol. Meyer. 72. 2011. L. M. 204–218. Metabolomics. 116. Tubman.. 2010. A. 2009. ed. Karaca. J. 993–998. J. 2010. M. Aceto. 394. Golland. M. 271 J. 387. J. Eur. Tang and S. Pfaff and R. 2011. J. Van Hove. 143. 1998. H. Zomer and J. B. 4. 2010. SpringerVerlag Berlin. Logue. F. A. 85. Suri. 3491–3502.. 181–184. London. Anselme. Food Chem. 1. Smilde. 2003. J. 2005. A. Kell. Waite. 985–991. G. Petrakis and P. 2007. 224 D. 46–46. 207–215... S. E. J. Mod. 8495–8501. V. Food Res. 1981.View Online 217 O. B.. Sci. Kell and S. P. E. Contam. 2002. Oliveira. 22. W. A. 285 M. 693–698. K. M. Panchenko. D.. Oddone. 254 M. T. 222–226. TrAC. 232 D. P. Nutr. Soc. P. 2011.. Methods. C. A. A. P. Peck and S. 56. C. Chem.C. 196–214. Rodriguez-Gonzalo and P. A. Laramee and H. L. Trends Biotechnol. TrAC. Anal. Webb. 284 D. J. C. 1994. A. M. Meat Sci. Pap. Augusti. 5706–5727 This journal is c The Royal Society of Chemistry 2012 . Sindona. Li. Crawford. Kocsube... G. Chem. 1158. Cody. Good. van Dorsten. Biometrika. C. Chim. Bengsch. S. van Duijnhoven and F. Pacheco and V. Pyrolysis. Kell. Stat. Berrueta. Y. C. Food Chem. 2011. Perrin. Y. 279 S. Downs. Bartok. Cardinal and E. Durst. 71. 263 R. 1499–1510. Belton. G. Biol. Lin. 48. Goodacre. L. 275 R. 272 R. Meir. D. M. Mol. Brown. M. 251 L. 2009. Goodacre. 2426–2436. Appl. 2005. J. Robert and T. 143–149. 1993. Mass Spectrom. 276 R. M. M. 30. Food Chem. M. Y. B. 237 J. F. P. 264 S. E. D. 2011. 700–707. 1361–1366. Goodacre. Ifa. H. Hidalgo and R. 266 A. 273 R. J... M. K. Borjesson. Brescia. 1997. 501–515. Plessi.org | doi:10. G. 165–180. W. 223 M. Bowirrat. A. W. J. 2008. 2007. 253 P. B. 1137. Food Chem. V.. S. M. 2001. S231–S243. J. 76. Bianchi. Kemsley. 243–247. 51–59. 3. Grova. T. Anal. S. T.. Rychen. 58. AOAC Int. I. 258 S. Am. Yang. Trends Anal. Mass Spectrom. Serra and M. J. Agric. J. R. McDonald and F. H. 239 B.. N. Jambrenghi. Salemi and D. S. Aursand. 265 W. Food Chem. J. 225 C. 5104–5108. 2007. Lindner. M. Meat Sci. vol. 2004. 22. in Learning Theory. C. Mariani and C. 231 D. H. Phys. Tice. E. Food Chem. Nilles. Chromatogr. De Luca and G. Blum. Sjostrom. J. Chim. Hammond and D. America. Metabolomics. D. Y. D. Chem. S. 2005. Feidt. G. L. B. 49. 2007. C. Kwong. 295–304. 2005. Marchello. P. 499–506. E. H. G. Int. 122. B. 269 J. Food Addit. J. Bioanal... W. Chen. 1997. D. L. Calmani. 41. Rapid Commun. Cozzolino. Food Addit.. B. C. M. Delgadillo. Yu. G. Cluster Analysis. 252 M. Jalali-Heravi. 203–214. B. S.. J. C. J. Tumino. Gold. 2010. Sep. 2007. Olsson. Agric. Dai. Dais. Morse and M. P. B. Metabolomics. Pussemier. Contam. 56.. 241 W. J. V. A. K. Li. in Magnetic Resonance in Food Science . Braverman. Chem. 1193–1200. Schreier and T. E. 399. 1. 2009. Mass Spectrom. Zhou and X. Franke. Kettrup and E. Brescia. M. 2010. Food Control.. 257 J. Motte. Jones. Axelson. Stocker. 40–41. Anal. 274 R. C. S. F. 32. 2005. Hoefsloot. Agiomyrgianaki. 233 M. 903–913. Vaclavik. 17. Chromatogr. Sacco. Crepineau. C. 2006. Carrino. Food Chem. Connor. J. Goodacre. 1995–2008.. 2010. P. Chapman & Hall/CRC. A.. I. J. 277 J. A. Kennedy. Mosandl and F.. Mol. 280 A. Food Res. H.

Hajslova. S121–S122.. R. Acta. Demirbas. C. K. 2415–2440. L. De Medici.. 1992. McManus. 290 J. Neofotis. Penza. L. 333 T. Second edn. A. Chem. S. A. 1933. 316 A. J. 1994. 309 C. 317 Y. J. Science. MA. J. J. Tryjanowski. Bino and R. Ivnitski. 24. 55. M. 326 S. H. Olshen and C. F. TrAC. 2008. 302 A. M. J. Goodacre. von der Kammer. Manly. Food Policy. Mass.. Wadsworth. Haverkamp and P. Lommen. Pretty. Rev. California. 307 R. S52–S61. Hotelling. J. 6464–6469. Tortorella and G. 587–604. Vaclavik. W. 385.. J. Sci. Beddington. Kistemaker. Schwack. Root. A. 2010. 425–436. 591–595. Soc. 288 T. V. 1983. A. K. 293 T. 305 N. Rudnitskaya. Woods. M. Garcia and R. Agric. 49. 329 H. Wu. R. Anderson and T. I. 335 R. 2009. I. P. Int. 33. 330 K. 696–699. Fenicia.. Breiman. Anal. 342 L. Legin and A. 337 D. Cambridge. K. 308 C. Q. Engelsen. 340 J.. E. Stone. 344 H. Mass Spectrom. Wielinga. Riddellova. 2008. Technol.. Cuellar and M. SpringerVerlag. 559–572. 301 A. M. Classification and regression trees. 1989. Nielen and J.1039/C2CS35138B This journal is c The Royal Society of Chemistry 2012 Chem. 145. Pimentel. 331 T. Hajslova. Kirshenbaum and M. B. 318 J. 2010. 4195–4203. Begley. P.. Kohonen. M. J. Learn. 5. D. Morlock and W. Sci. Knutsson. C. 21. 325 I. MIT Press. 322 J. M. 292 L. Ba ¨ ck. Hajslova. Fricker. Bioanal. 649–665. Gil. 20. 1986. 2010.. T. Chichester. C. Hastie. I. J. Godfray. 2009. Næs. 5706–5727 5727 . 30. G. J. 321 R. Riddellova and J. Metabolomics. D. G. 30. Rumelhart and J. Multivariate Statistical Methods: A Primer. Martens and T. The University of Michigan Press. 17. Bioelectron. Thomas and C.. 9. 80. Wiley. 41. Tomaniova and J. 1989. 29. Oxford. M. Agren. Nutr. Michigan. 377. Friedman. Sanchez-Brunete. 299 D. Rohlik. 1984. 2008. Dunn. Eur. Valero.. Energy Convers. 417–441 and 498–520. Biosens. 314 A. Trends Anal. B. Liu. L. Chemometrics for Pattern Recognition. I. New York. D. Cody. Food Sci. Moco. J. Karoly. Smilde. Anal. B. Hervas and M. 2003. Koza. J.View Online 287 L. 35–44. Ackerman. B. Skiby. B. Murphy. 21. 133–138. Casassa. Chapman & Hall. McClelland. Kim and Y. Energy Environ. 4. 2009. Vlasov. Dunn. Neural Computing: Theory and Practice. K. Cassano. 82. Brynda. 45. X. Breiman. 2012. John Wiley. E. 312 A. Cajka. F. G. Food Microbiol. 1848–1854. New Sci. Thompson. W. 327. J. M. Das. Rev. 2. Sensors. Int. Environ. Martin. Noonan and T. B. A. Wasserman. Anal. R. 298 H. J. R. 1997. 336 W. 294 T. 320 M. EhlingSchulz. J. Actuators. Pipek and J. H.. 2009. L. Knutsson. 2003. J. Technol. H. G. 373. C. D. B. Toulmin. Vaidyanathan. Harrigan and D. LC GC Europe. 21–29. H. 2002. Prog. M. 16–33. 372–377.. B. 323 F. A. 245–252. Ferreira and M. 1087–1112. Cajka. Legros. 78. 297 A. J. 2008. G. 2006. J. 353–U320. 296 G. J. 2011. Fogel and Z. Abdel-Hamid. 2001. Electrophoresis. Hall. 408–421. 35. Erban. Analyst. J. Kell. R. M. Chromatogr. Berlin. Sens. Mag. 2003. G. S. Friedman. 22. Chem. Bro and A. 1033–1053. Bioanal. Springer International. 36. Tadeo. 2011. 2004. 339 T. Chem. Parallel Distributed Processing. 227. 367. 28–29. Goodacre. K. P. G. Martens. London. 2011. Hajslova.. Genetic Programming: On the Programming of computers by Means of Natural Selection. 2007. 30. Handbook of Evolutionary Computation. Pearson. Wilkins. P. Poustka. Nielsen and S. Guo and D. 2003. Loscher. Zaccaria. 1999. Rosenzweig. Trends Food Sci. B. 1–18. L. Lofstrom. Brereton. Food Control. 778–791. Chim. 2011. Chem. Cambridge. D. 1217. A. Pacific Grove. 59. 4852–4861. Lommen. 2009. 6. Dwivedi. Crevillen. Bioelectron. 332 B. Bioanal.rsc. Luedemann. 311 S. 300 A. 303 S. Chem. O. Pimentel and M. Protoc. R. Hofmann. 15. C. Vaclavik. S123–S128.. Demirbas. Energy. Self-Organization and Associative Memory.. 328 R. Escarpa.. W. De Vos. Q. Anal. Downloaded on 20 August 2012 Published on 25 June 2012 on http://pubs. Crit.. 26. 2010. 2008. and Prediction. 1994. A. P. Nat. 1217. Am. 1989. Khanna and M. Cui. Rosenzweig and M. Anal. Jang. 1975. Experiments in the Microstructure of Cognition. F. Gonzalez. 2010. P. Bhatia. 1623–1632. 81–88. 2004. 287–290. 2009.. M. Nature. Biosens. Journal of Educational Psychology. B. 304 R. Caldow. de Koning. 2011. Chem. D. Segerman. R. 2002. Chemom. 2007. O. C. New York. 346 R. 2011. J. Multivariate Calibration. Talanta. 313 A. Kopka and R.. Technol. Zhang. 207. W. L. 76–87. Chichester. 250. Gonzalez. 2106–2116. The Elements of Statistical Learning: Data Mining. A. Rapid Commun. G. TrAC. Food Additives Contaminants Part A. Gallego-Urrea. Stone and M. N. 434–445. Seguin. Cajka. L. A. 1950–1957. D. 73. F. Sharman. 319 X. J. Kay. Hrbek and J. Adaptation in Natural and Artificial Systems. Webber. Garcia-Valcarcel. D. Baeumner. L. 812–818. Mueller. J. Food Microbiol. Clin. Keurentjes. Crute. Slavik. Oliveira. Garcia. 36.. 9567–9575. 289 H. Z. S. Inference.. 24. 1901. 17. Mafra. 44. Muir.. 295 L. M. Vaclavik. J. K. T. H. Dane and R. Sci. Inc. Wallington. Tibshirani and J. 645. P. 2011. A. D. 14. J. 324 V. Schwack. J. Holland. Chromatogr. Trends Biotechnol. Parry.. 136–146. J. A. Philos. MIT Press. Hrbek and J. V. 660S–663S. Tuoriniemi and M. J. Webber. 5–32. Hajslova. van Rotterdam. Ann Arbor. Vaclavik. E. 479–496. L. 5919–5926. Phytochem. Zachariasova. 394–404. Albero and A. I. J. 338 P.. G. Chem. Plant Sci.. Int. 473–483. S. 2011. T. M. Menzel. Fach. L. P. B. Nature. Q. Andersson. Anal. 341 J. 135. P. 306 R. IOP Publishing/Oxford University Press. 327 R. 343 L. Lawrence. A. S. K.org | doi:10. 2011. 599–624.. 75. Rastogi. Food Chem. 1611–1618.. Hooijerink. 145. Vaclavik. 310 D.. 334 H.. 39. Haddad. Schultz and M. Allwood. Imeson. H. Hrbek. Chem. 2010. Anal. 453.. Windig. B. Estrella. 291 L. Hajslova. J. Mach. 2009. Larsen. C. Atanasov and E. S. 2010. 2007.. Rothenbacher and W. 56–63. Trends Anal. Food Res. Fajardo. Van Nostrand Reinhold. A. Anal. A. 2004. S. Cajka. 2011. Energy Combust. Manage. Taufiqurrahmi and S. T. Pimentel. 586–595. J. B.. Michalewicz. Robinson. G. Technol. J.. Rawlins and A. 2010. Loeschner and T. Lacina. 315 D. E. Black and M.. 2001. C. 345 J. Homola and E.. S. Deisingh. 2010. Hassellov. Biomass Bioenergy. Vicarelli. Rojas. Schurek. V.