Subscriber access provided by UNIV OF WISCONSIN - MADISON
Article
Detection of starch adulteration in onion powder by FT-NIR and FT-IR spectroscopy Santosh Lohumi, Sangdae Lee, Wang-Hee Lee, Moon S. Kim, Changyeun Mo, Hanhong Bae, and Byoung-Kwan Cho J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/jf500574m • Publication Date (Web): 04 Sep 2014 Downloaded from http://pubs.acs.org on September 9, 2014
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Agricultural and Food Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 28
Journal of Agricultural and Food Chemistry
1
Detection of starch adulteration in onion powder by FT-NIR and
2
FT-IR spectroscopy
3
Santosh Lohumi1, Sangdae Lee1, Wang-Hee Lee1, Moon S. Kim2, Changyeun Mo3, Hanhong
4
Bae4, Byoung-Kwan Cho1,*
5
1
6
Chungnam National University, 99 Daehak-ro, Yuseoung-gu, Daejeon 305-764, Republic of
7
Korea
8
[email protected], [email protected], [email protected]
9
2
Department of Biosystems Machinery Engineering, College of Agricultural and Life Science,
Environmental Microbial and Food Safety Laboratory, USDA-ARS, 10300 Baltimore Ave.,
10
Beltsville, MD 20705, USA
11
[email protected]
12
3
13
Gwonseon-gu, Suwon, Gyeonggi-do 441-100, Republic of Korea
14
[email protected]
15
4
16
[email protected]
17
*Corresponding author: Byoung-Kwan Cho
18
Tel: +82-42-821-6715, Fax: +82-42-823-6246, E-mail: [email protected]
National Academy of Agricultural Science, Rural Development Administration, 150 Suinro,
School of Biotechnology, Yeungnam University, Gyeongsan 712-749, Republic of Korea
1 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 2 of 28
19
Abstract: Adulteration of onion powder with cornstarch was identified by Fourier transform
20
near-infrared (FT-NIR) and Fourier transform infrared (FT-IR) spectroscopy. The reflectance
21
spectra of 180 pure and adulterated samples (1–35 wt% starch) were collected and preprocessed
22
to generate calibration and prediction sets. A multivariate calibration model of partial least-
23
squares regression (PLSR) was executed on the pretreated spectra to predict the presence of
24
starch. The PLSR model predicted adulteration with an ܴଶ of 0.98 and a standard error of
25
prediction (SEP) of 1.18% for the FT-NIR data, and an ܴଶ of 0.90 and SEP of 3.12% for the FT-
26
IR data. Thus the FT-NIR data were of greater predictive value than the FT-IR data. Principal
27
component analysis on the preprocessed data identified the onion powder in terms of added
28
starch. The first three principal component loadings and beta coefficients of the PLSR model
29
revealed starch-related absorption. These methods can be applied to rapidly detect adulteration in
30
other spices.
31
Keywords: Starch, onion powder, adulteration, Fourier transform NIR and IR spectroscopy,
32
partial least-squares regression, principal component analysis
2 ACS Paragon Plus Environment
Page 3 of 28
33
Journal of Agricultural and Food Chemistry
Introduction
34
Spices are used pervasively in our daily lives, such as in the flavoring of food, nutrition,
35
perfumery, medicinal uses, and as preservatives.1 However, spices are common agricultural
36
commodities that are prone to adulteration. For legal or other reasons, adulteration is defined as
37
the addition of extraneous substances that are not normally contained within the original
38
material.2 Food adulteration can be categorized in two ways. Incidental adulteration occurs when
39
foreign substances are incorporated into a foodstuff due to negligence or ignorance. This can
40
occur in the field during harvesting, for example. The second is intentional adulteration, which
41
occurs by either the intent to cause physical or chemical harm, or the motivation for economic
42
gain.3 Adulteration has been commonplace throughout human history, and continues today with
43
several notable instances involving the spice industry. In 1994, ground paprika in Hungary was
44
found to be adulterated with lead oxide, causing the deaths of several people and sickening
45
dozens of others. More recently, ground capsicum was found to contain dyes not approved for
46
use in food.4
47
Recent methods used to authenticate spices have been based on morphological, microscopic,
48
chemical, or DNA-based characterizations. However, these analytical methods may not be
49
convenient for routine sample analysis or require a certain degree of expertise. In some cases,
50
chemical standards may be rare or expensive, or lack identifiable markers.5 Thus, there is still a
51
need to develop a fast, chemical-free, and cost-effective method for the detection of adulteration,
52
which can be applied over a wide range of food products, including spices.
53
Both Fourier transform near-infrared (FT-NIR) and Fourier transform infrared (FT-IR)
54
spectroscopies have been successfully used for qualitative and quantitative analyses of
3 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 4 of 28
55
agricultural products, as well as for pharmaceutical and petrochemical commodities. These
56
spectroscopic methods are based on the chemical composition and moisture content of the
57
biological materials.6 Chemical bonds between light atoms (such as C–H, O–H, and N–H)
58
generally have molecular vibrations, which result in overtones and combination bands that are
59
detectable in the NIR region, 780–2,500 nm.7 The specific patterns exhibited in the NIR region
60
can reveal the chemical and physical compositions of the material being studied. The mid-IR
61
(MIR) region (400–4000 cm−1) monitors the fundamental vibrational and rotational motions of
62
molecules, which provides a chemical profile of the sample. The MIR region is usually
63
considered as a reproducible region of the electromagnetic spectrum in which very small
64
differences in sample composition can be reliably measured.8 In terms of penetration power, NIR
65
radiation can generally penetrate much further into a sample than that of the MIR. The effective
66
path-length through the sample is impacted by the penetration depth of the IR waves. Therefore,
67
NIR spectroscopy is very useful in exploring bulk materials with little or no sample preparation.
68
The difference in the penetration depths between the NIR and MIR regions influences the
69
precision of developed multivariate analytical models as NIR region can penetrate several
70
millimeter into the sample while penetration power of MIR region in microns.9, 10, 11
71
Various studies have been investigated the potential of NIR and MIR spectroscopies to
72
determine melamine adulteration in infant formula powder,12 in dairy milk13 and in soya bean
73
meal.14 The effectiveness of FT-IR spectroscopy and hyperspectral imaging has already been
74
evaluated for the detection of adulteration in black pepper powder.15 In addition, FT-IR
75
spectroscopy in combination with a partial least-squares (PLS) model has also proved competent
76
to detect melamine adulteration in dairy milk.16 The use of FT-Raman and FT-IR spectroscopies
77
was investigated in the characterization of irradiated starch meaning a starch that irradiated at 3, 4 ACS Paragon Plus Environment
Page 5 of 28
Journal of Agricultural and Food Chemistry
78
5, or 10 kGy using a Gammacell 220 Co60 gamma irradiator.17 This research group developed a
79
regression model with the spectroscopic data and observed that NIR performed better than MIR
80
in terms of predicting chemical parameters (e.g., reducing sugars, ethanol, total phenolics, and
81
flavonoids) involved in the fermentation process.
82
Our study sought to quantitatively detect starch in onion powder using FT-NIR and FT-IR
83
spectroscopic techniques. Spectral analysis of eighteen concentrations of starch in onion powder
84
samples, ranging from 0 to 35%, was conducted. Based on the results, different adulterant
85
concentrations were identified and categorized by developing a multivariate analytical method.
86
The ultimate objective of this study was to predict and classify the added starch concentration in
87
onion powder using FT-NIR and FT-IR spectroscopic techniques.
88
89
Materials and Methods
90
Sample and adulterant
91
Onion powder and cornstarch (purities of 100% and 95%, respectively) were obtained from a
92
local market in Korea. Cornstarch was used as the adulterant in the onion powder because of its
93
cheap price, no significant effects on human health, and similar physical property. The samples
94
were stored at room temperature until use.
95
96
Sample preparation
5 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 6 of 28
97
The moisture content of the samples was equilibrated by drying in an oven at 74°C for 2 h prior
98
to mixing because spectroscopic measurement is highly sensitive to moisture content. The given
99
temperature was chosen to dry the sample effectively without interfering with the chemical
100
integrity of the samples.15 Both the dried sample and adulterant material weights were measured
101
with an electronic balance (Model AR2130, Ohaus Corp., USA). Then, the cornstarch was mixed
102
with the onion powder to final concentrations (w/w) of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, 18, 20,
103
25, 30, and 35%. Each combination was manually mixed and then transferred to a snap-cap vial.
104
Further mixing was accomplished by placing the filled vials onto a high speed shaker (Vortex-
105
Genie® 2, Scientific Industries, Inc., Model G560, USA) to reduce the variation in particle size.
106
Ten samples of each of the pure and adulterated powders were used, and sample quantities were
107
approximately 50 mg for the FT-NIR and 30 mg for the FT-IR studies.
108
109
FT-NIR spectra collection
110
NIR reflectance spectra of the pure and adulterated onion powder samples were obtained using
111
an FT-NIR spectrometer (Antaris II FT-NIR analyzer, Thermo Scientific Co., USA) equipped
112
with an InGaAs detector. The sample holder was furnished with an accessory that contained a
113
central hole (20 mm thick × 20 mm diameter) to maintain a uniform shape and thickness for the
114
sample over the irradiated surface during spectra collection. A reference scan was conducted
115
with a golden slit before scanning each sample. The instrument measured the diffuse reflectance,
116
and a total of 32 successive scans from each sample was collected over the wavelength range
117
between 4,000 and 10,000 cm−1 (1,000–2,500 nm) at 4 cm−1 intervals. Averaged spectra were
118
used for analysis.
6 ACS Paragon Plus Environment
Page 7 of 28
Journal of Agricultural and Food Chemistry
119
120
FT-IR spectra collection
121
FT-IR measurements were performed in the mid-infrared region (4000–650 cm−1) using a
122
Nicolet 6700 (Thermo Scientific Co., USA) FT-IR spectrophotometer, configured with an
123
attenuated total reflectance (ATR) sampling technique, deuterated triglycine sulfate (DTGS)
124
detector, and KBr beam splitter controlled by OMNIC software. Samples to be analyzed were
125
placed on a diamond crystal sampling plate and clamped with a pointed tip. A background scan
126
was obtained before every sample scan with an empty sample plate. In addition, the ATR
127
crystals and pointed tip were cleaned to remove any interference from the preceding sample. A
128
total of 32 successive scans from each sample were collected at 4 cm−1 intervals, and averaged
129
spectra of each sample were used for analysis.
130
131
Spectral data analysis
132
The spectroscopic (NIR and mid-IR) detectors receive light coming from the sample in the form
133
of diffuse reflectance after absorption, specular reflectance, and scattered light. Only the diffuse
134
reflectance contains chemical information, whereas the latter two carry no useful information as
135
there is no large variation in particle size between starch and onion powder (based on the result
136
from Cilas 1090 particle size analyzer). The collected spectra might be affected by systemic
137
noise due to light scattering, variation in particle size, and instrumental drift. Thus, in order to
138
determine an accurate chemical composition from spectroscopic measurements, the raw spectra
139
must be corrected at different levels by applying preprocessing methods.18 In this study, both FT-
140
NIR and FT-IR spectral data were treated with 5 pre-processing methods: three data 7 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
methods
(minimum,
maximum,
and
range),
standard
Page 8 of 28
141
normalization
normal
variate
142
transformation (SNV), and multiple scatter correlation (MSC). The original FT-NIR spectra of
143
all the samples, and the original spectra pre-processed by the SNV method are depicted in Figs.
144
1(a) and 1(b), respectively. The raw FT-IR spectra for each sample and SNV-preprocessed
145
spectra are shown in Figs. 2(a) and 2(b), respectively.
146
147
For the data analysis, principal component analysis (PCA) was carried out on the SNV- and
148
MSC-processed spectra from both FT-NIR and FT-IR. PCA can readily be applied to
149
spectroscopic data to perceive the nature and scattering level of the data. The first principal
150
component (PC1) describes the maximum of variation or spread in the samples. A second
151
principal component (PC2), orthogonal to the first, (i.e., completely uncorrelated), describes the
152
maximum of the variation not described by the first eigenvector, and so on. With this procedure,
153
the most important features of the data set can be seen in a low dimensionality plot.19 A
154
multivariate calibration model of a partial least-squares regression (PLSR) was then developed
155
using all the preprocessed spectral data to predict the extent of adulteration in the pure onion and
156
adulterated onion powder samples. PLSR is particularly suited when the matrix of predictors has
157
more variables than observations, such as from spectral data. PLSR has been used in food
158
authentication studies based in spectroscopic data.11 Multivariate analysis was performed using
159
MATLAB software version 7.0.4 (The Mathworks, Nitick, MA, USA).
160
The whole dataset (180 samples) was split into two sets: a calibration set consisting of 126
161
samples (7 samples from each group), and a prediction set consisting of 54 samples (3 samples
162
from each group). The PLSR models were built with the calibration set using a full cross-
163
validation method (leave-one out), removing one observation at one time from the sample sets 8 ACS Paragon Plus Environment
Page 9 of 28
Journal of Agricultural and Food Chemistry
164
until all samples have been removed once. This is the way generally used in developing PLSR
165
model for detecting a specific material with spectral dataset.20
166
167
Result and Discussion
168
Interpretation of spectral analyses
169
The FT-NIR original spectra of the pure and adulterated onion powder, and the SNV-
170
preprocessed spectra (Figs. 1(a) and 1(b)) reveal differences in the absorption intensities between
171
1920–1980 nm (marked) and can easily be identified. The absorption bands in this region
172
correspond to the O–H stretch and O–H band combination and the H–O–H deformation
173
combination, which represent the starch content.21 Starch is a semicrystalline polymer composed
174
of two polysaccharides, amylose and amylopectin. Amylose is a mostly linear chain consisting of
175
up to 3000 glucose molecules interconnected primarily by α-1,4-glycoside linkages.13 The
176
spectral fluctuations from 1400 to 1600 nm correspond to the first overtone of the hydroxyl
177
group. The precise position of these bands is very sensitive to hydrogen bonding in the starch
178
molecule.22
179
The regression coefficient plot (Fig. 3) of the PLS model for the FT-NIR data indicate some
180
meaningful bands that are attributable to the starch. The absorption band around 1420 nm
181
corresponded to a combination of O–H first overtone. Another absorption band at 1920 nm
182
related to CONH, which referred to as amide linkage. The absorption bands at 2312 nm and 2455
183
nm corresponded to C–H stretching and deformation vibrations, which represent starch content.
184
Figures 4(a) and 4(b) show the results of the PCA for the SNV-pretreated FT-NIR spectral data.
9 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 10 of 28
185
A data grouping was observed from the scatter plot of the PC scores (Fig. 4(a)), but the data for
186
samples adulterated at low concentration appear to overlap more. This confirmed our assumption
187
that the different spectral attributes among the samples were associated with the concentration of
188
the starch in the onion powder. Further, we assessed the potential of PCA to distinguish among
189
the low starch concentration samples (1–10%), but the data were overly scattered and
190
overlapping (data not shown). The first three PC loadings account for nearly 99% of the total
191
variance in the samples. PC loadings are correlation coefficients between the PC scores and the
192
original variables, which measure the importance of each variable in accounting for the
193
variability in the PC. It is possible to interperate the first few PCs in term of overall effect of a
194
contrast between groups of variables based on the structures of PC loadings. The PC loading plot
195
(Fig. 4(b)) shows the highest loading for PC1 and PC3 at ~1960 nm, related to the starch. PC1
196
and PC2 show other important bands between 1435 and 1450 nm, corresponding to the first O–H
197
stretching overtone, also assigned to the starch.
198
Typically, MIR region can be divided into two distinct groups: the region of the spectrum below
199
1500 cm−1 is called fingerprint region.23 Peaks in this region are very difficult to assign to
200
specific molecular vibration because each different compound produce a different and unique
201
absorption pattern in this part of spectrum.16 The MIR region from 4000 – 1500 cm-1 is known as
202
the functional group region. Most of the relevant information that is used to interpret the MIR
203
spectrum usually obtains from the functional group region.24 The most notable differences in the
204
FT-IR spectra in this study can be categorized into three main regions, which interpret the
205
significance of key bands (Figs. 2(a) and 2(b)). The first region was the fingerprint region (650–
206
1500 cm−1), which provided more complex spectra and fairly indistinguishable individual bands
207
than the region at higher wavenumbers. The peaks in the region 1045–1087 cm−1 are due to C–C 10 ACS Paragon Plus Environment
Page 11 of 28
Journal of Agricultural and Food Chemistry
208
bond vibrations related to the carbohydrate; those at 1170–1250 cm−1 correspond to absorption
209
bands for C–H, C=O, and C–C vibrations;25 and the triplet bands for the C–O–C stretching
210
absorption are found at 1147 and 1010 cm−1.26 The vibrations originate from the C–O–C bonds
211
related to the α-1,4-glycoside linkages, as these bonds are common in carbohydrates. The C–H
212
stretching region, found between 2850 and 3000 cm−1, includes a distinct band at 2920 cm–1.26
213
Finally, the higher wavenumber region (3000–3500 cm−1) indicates the O–H stretching vibration
214
bands. These peaks could be a result of moisture uptake in the samples during spectral analysis.
215
The beta coefficient (Fig. 5), which indicates the spectral differences among the samples, shows
216
some important bands representative of the carbohydrates (starch) that contribute to the
217
quantitative classification of the onion powder and adulterated samples.
218
PCA was executed on both the MSC- and SNV-preprocessed FT-IR spectral data to extract
219
relevent information. However, the data might be classified according to moisture content or
220
particle size, rather than their chemical composition characteristics, because the IR data patterns
221
are not only determined by absorption characteristics but also scattering characteristics. These
222
spectral variations were removed by oven drying all the samples and pretreating the spectra with
223
MSC- and SNV-preprocessing methods. Neither the MSC- nor the SNV-preprocessed FT-IR
224
data were able to quantify the level of adulteration by means of data clustering using PCA. The
225
data were scattered over a vast range and no distinct separations among the data groups were
226
found from the PCA score plot (data not shown). PCA is a popular primary technique for data
227
clustering. It is not, however optimized for class separability. The loading plots of the first three
228
PCs are depicted in Fig. 6. The PC1 loading accounts for 96.1% of the total sample variance,
229
producing a pattern similar to the raw FT-IR spectra (Fig. 2(a)). The loadings of PC1 and PC2
230
show the highest intensity at ~980 cm−1. Sankaran et al. (2010) also found a similar peak for 11 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 12 of 28
231
starch in the spectral range 950–1000 cm−1.27 In addition, Danwille (2011) observed a sharp peak
232
at 975 cm−1 arising from an –OCH3 bond vibration related to cellulose.15
233
234
Partial least squares regression analysis
235
A PLSR model was used to predict the added starch concentration in onion powder. The samples
236
were categorized based on the adulterant concentration, and a PLSR model was developed using
237
the preprocessed spectra of samples from both FT-NIR and FT-IR spectroscopy. The highest
238
correlation coefficient value, ܴଶ , of 0.98 with a standard error of prediction (SEP) of 1.18% was
239
observed by using the SNV-preprocessed FT-NIR spectra. The optimal number of factors used in
240
the PLS models were automatically selected based on the lowest value of predicted root mean
241
square error (RMSE) in the cross validation process. The PLSR results are listed in Tables 1 for
242
the FT-IR spectral data.
243
The PLSR result obtained for the FT-NIR data shows a strong relatioship (ܴଶ = 0.98) between
244
the actual and predicted concentration values. The result suggests that four principal components
245
were enough to extract 99% of the relevant information used to detect starch adulteration in
246
onion powder.
247
Based on the pattern of the FT-IR spectra, the prediction efficiency of the PLSR model was
248
evaluated using the SNV-preprocessed spectra of the full spectral region, as well as two different
249
selected regions (i.e., the fingerprint region (650–1500 cm−1) and the functional group region
250
(1500–4000 cm−1). The highest correlation coefficients of prediction (ܴଶ ), 0.90 and 0.89, were
251
obtained for the full spectral and fingerprint regions, respectively. In addition, these two spectral
12 ACS Paragon Plus Environment
Page 13 of 28
Journal of Agricultural and Food Chemistry
252
regions showed nearly the same ܴଶ and ܴ௩ଶ values, with a small difference in SEC and SEV
253
(Table 1). The prediction by PLSR of the functional group region was acceptable (ܴଶ = 0.84) but
254
lower than the first two groups.
255
Jawaid et al. (2013) and Sherman (1997) have suggested that the functional group region is better
256
for classification and prediction than the fingerprint region, because most of the information used
257
to interpret the MIR spectrum is usually obtained from the former, whereas the fingerprint region
258
is normally complex with many frequently overlapped bands.16, 28 In our study, the fingerprint
259
region gave better results than the functional group region with respect to the quantative
260
prediction of starch adulteration in the onion powder samples. Moreover, Filippov (1992) also
261
discribed the use of fingerprint region of MIR for the identification of carbohydrates.29 For more
262
information readers can be referred to reference no. 29 and references therein.
263
From the PLSR and PCA results, the FT-NIR spectral data were better able to quantitatively
264
predict the extent of starch adulteration in onion powder samples than the FT-IR spectral data. It
265
should be noted that limited amounts of sample are irradiated by MIR radiation during the FT-IR
266
analysis. Additionally, the penetration of MIR radiation is usually less than that of the NIR.
267
These reasons could be responsible for the low classification rate for the quantitative
268
discrimination of adulteration using FT-IR spectroscopy. It is known that onions contain
269
calcium, proteins, water-soluble carbohydrates (e.g., glucose, fructose, and sucrose), and other
270
chemical components.30 The concentration of added starch in the onion powder proportionally
271
influences the concentration of these chemical compounds on a percentage basis. These chemical
272
compounds have different absorption intensities at specific wavelengths in both the NIR and
273
MIR regions. In our study, in addition to the starch concentration, these intrinsic chemical
274
compounds might be responsible for distinct the classification. 13 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 14 of 28
275
In summary, two different vibrational spectroscopic techniques were investigated to
276
quantitatively detect cornstarch contamination in onion powder. The potential of these
277
techniques was evaluated by adding starch at various concentrations (1–35 wt%) into onion
278
powder and by properly mixing the samples. Both FT-NIR and FT-IR spectroscopic methods
279
demonstrated good potential; however, FT-NIR spectroscopy in combination with PLSR
280
afforded a higher prediction accuracy (ܴଶ = 0.98) than FT-IR spectroscopy (ܴଶ = 0.90). SNV
281
preprocessing, which has been useful in other quantitative applications, improved the accuracies
282
overall. These techniques would seem to be promising for the effective, chemical-free, and rapid
283
detection of product adulteration, as they require no sophisticated laboratory or well-trained
284
personnel to perform the analysis. In addition, for quality control measurements, both FT-NIR
285
and FT-IR spectroscopic techniques are fast and offer a solution for authenticity analysis of
286
agricultural and food products. Further studies could be performed to establish qualitative
287
methods for the detection of adulteration in spices.
288
289
ABBREVIATIONS USED
290
FT-NIR, Fourier transform near infrared; FT-IR, Fourier transform infrared; MIR, mid infrared;
291
PLSR, partial least-square regression; PCA, principal component analysis; SEP, standard error of
292
prediction; DNA, deoxyribonucleic acid; InGaAs, indium gallium arsenide; KBr, potassium
293
bromide; ATR, attenuated total reflectance; DTGS, deuterated triglycine sulfate; SNV, standard
294
normal variate transformation; MSC, multiple scatter correlation; PC, principal component;
295
RMSE, root mean square error; R2, correlation coefficient; ܴଶ , ܴ௩ଶ , ܴଶ are correlation coefficient
14 ACS Paragon Plus Environment
Page 15 of 28
Journal of Agricultural and Food Chemistry
296
of calibration, validation, and prediction, respectively; SEC, standard error of calibration; SEV,
297
standard error of validation ; SEP, standard error of prediction;
298
299
ACKNOWLEDGMENT
300
This research was supported by High Value-added Food Technology Development Program,
301
Ministry of Agriculture, Food and Rural Affairs (MAFRA), Republic of Korea.
15 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 16 of 28
302
REFERENCES
303
(1) Bhattacharjee, P.; Singhal, R.S.; Achyut, S.G. Supercritical carbon dioxide extraction for
304
identification of adulteration of black pepper with papaya seeds. Journal of the Science of Food
305
and Agriculture. 2003, 83, 783-786.
306
(2) Lakshmi, V. Food adulteration. International journal of science invention today. 2012, 1(2),
307
106-113.
308
(3) Everstine, K. Economically motivated adulteration, implication for food protection and
309
alternate approaches to detection. 2013, Ph.D. thesis. Faculty of the graduate school of the
310
University of Minnesota.
311
(4) ASTA (American Spice Trade Association). Spice adulteration – White paper.
312
www.astaspice.org/files/public/SpiceAdulteration.pdf.
313
(5) Shaw, P. C.; Ngan, F. N.; But, P. P. H.; Wang, J. Molecular marker in Chinese medicinal
314
materials. World scientific publishing co. Pte.ltd. 2002, Chapter 1 (1-5).
315
(6) Lohumi, S.; Mo, C.; Kang, J. S.; Hong, S. J.; Cho, B. K. Nondestructive evaluation for the
316
viability of watermelon (Citrullus lanatus) seeds using Fourier transform near infrared
317
spectroscopy. Journal of Biosystems Engineering. 2013, 38(4):312 – 317.
318
(7) Osborne, B. G.; Fearn, T.; Hindle P. T. Practical NIR spectroscopy with applications in food
319
and beverage analysis. Longman Scientific and Technical, Singapore, 1993, 2nd edition.
320
(8) Wen-Sun, D. Infrared spectroscopy for food quality and control 2009, 1st edition. ISBN.
16 ACS Paragon Plus Environment
Page 17 of 28
Journal of Agricultural and Food Chemistry
321
(9) Bras, L. P.; Bernardino, S. A.; Lopes, J. A. ; Menezes, J. C. Multiblock PLS as an approach
322
to compare and combine NIR and MIR spectra in calibrations of soybean flour. Chemomatrics
323
and Intelligent Laboratory System. 2004, 75(2005):91 – 99.
324
(10) Lammertyn, J.; Peirs, A.; Baerdemaeker, J. D.; Nicolai, B. Light penetration properties of
325
NIR radiation in fruit with respect to non-destructive quality assessment. Postharvest Biology
326
and Technology. 1999, 18(2000):121 – 132.
327
(11) Domagala, W. U. The use of the spectrometric technique FTIR-ATR to examine the
328
polymers surface. InTech, (http://creativecommons.org/licenses/by/3.0). 2012, Chepter 3:85 –
329
104.
330
(12) Mauer, L. J.; Chernyshova A. A.; Hiatt, A.; Deering, A.; Davis, R. Melamine detection in
331
infant formula powder using near- and mid-infrared spectroscopy. Journal of Agricultural and
332
Food Chemistry, 2009, 57, 3974 - 3980;
333
334
(13) Balabin, R. M.; Smirnov, S. V. Melamine detection by mid and near-infrared spectroscopy:
335
a quick and sensitive method for dairy products analysis including liquid milk, infant formula,
336
and milk powder. Talanta, 2011, 15;85(1): 562-8
337
(14) Haughey, S. A.; Graham, S. F.; Concouet, E.; Elliott, C. T. The application of near-infrared
338
spectroscopy (NIRS) to detect melamine adulteration of soya bean meal. Food Chemistry, 2012,
339
doi: 10.1016/j.foodchem.1012.01.068.
340
(15) Danwille, J. F. S. Detection and quantification of spice adulteration by near infrared
341
hyperspectral imaging. 2011, Master of Science in food science (Master thesis). 17 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 18 of 28
342
(16) Jawaid, S.; Talpur F.N.; Sherazi S.T.H.; Nizamani S. M.; Khaskheli A. A. Rapid detection
343
of melamine adulteration in dairy milk by SB-ATR–Fourier transform infrared spectroscopy.
344
Food chemistry, 2013, 141, 3066-3071.
345
(17) Ramazan K.; Irudyaraj, J.; Seetharaman, K. Characterization of irradiated starches by using
346
FT-Raman and FTIR spectroscopy. Journal of agricultural and food chemistry 2002, 50(14):
347
3912-8.
348
(18) Norris, K.; Williams, P. Optimization of mathematical treatment of raw near-infrared signal
349
in the measurement of protein in hard red spring wheat. I. Influence of particle size, Cereal
350
Chemistry 1983, vol. 61, pp. 158–165.
351
(19) Benar, P.; Adilson, R. Principal component analysis of the hydroxymethylation of sugarcane
352
lignin. Journal of wood chemistry and technology 1999. 19 (1&2), 151-165.
353
(20) Kandpal, L. M.; Lee, H.; Kim, M. S.; Mo, C.; Cho, B. K. Hyperspectral reflectance imaging
354
technique for visualization of moisture distribution in cooked chicken breast. Sensor 2013. 13,
355
13289 – 13300.
356
(21) Aenugu, H. P. R.; Kumar, D. S.; Parthiban, N.; Ghose S. S.; Banji, D. Near infrared
357
spectroscopy – an overview. International journal of ChemTeCh research 2011. vol. 3, no. 2,
358
825-836.
359
(22) Noah, L.; Robert, P.; Millar, S.; Champ, M. Near-infrared spectroscopy as applied to starch
360
analysis of digestive contents. Journal of agriculture and food chemistry 1997. 45, 2593-2597.
361
(23) – Griffiths, P. R.; de Haseth, J. A. Fourier transform infrared spectroscopy, 2nd edition.
362
2007, John Wiely and Sons, Inc., Hoboken, NJ. 18 ACS Paragon Plus Environment
Page 19 of 28
Journal of Agricultural and Food Chemistry
363
(24) Hsu, C. P. S. Handbook of instrumental techniques for analytical chemistry. Separation
364
sciences research and product development Mallinckrodt, Inc. 1997, Chapter 15:247 – 282.
365
(25) Fagan, C. C.; Donnell, O. Application of Mid-infrared Spectroscopy to food processing
366
system in nondestructive testing of food quality. J. Irudayaraj, C. Reh. Eds. 2008, pp. 119-142.
367
(26) Razali, M. A. A.; Sanusi, N.; Ismail, H.; Othman, N.; Ariffin, A. Application of response
368
surface methodology (RSM) for optimization of cassava starch grafted polyDADMAC synthesis
369
for cationic properties. Starch - journal. 2012, 64, 935-943.
370
(27) Sankaran, S.; Ehshani, R.; Etxeberria, E. Mid-infrared spectroscopy for the detection of
371
Huanglongbing (greeening) in citrus leaves. Talanta. 2010, 83, 574-581.
372
(28) Sherman, C. P. Infrared spectroscopy. Handbook of Instrumental Techniques for Analytical
373
Chemistry. Separation sciences research and product development mallinckrodt, Inc. Chapter 15,
374
1997, 247 – 282.
375
(29) Filippov, M. P. Practical infrared spectroscopy of pectic substances. Food Hydrocoloids,
376
1992, 6(1), 115 – 142.
377
(30) Wall, D.; Mishra, M.; Corgan, J. N. Dehydrate onion bulb weight and water-soluble
378
carbohydrates before and after maturity. Journal of American society for horticultural science.
379
1999, 124(6):581–586.
19 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 20 of 28
380
LIST OF FIGURE CAPTIONS
381
Figure 1. (a) Original FT-NIR spectra of pure onion and starch onion mixtures at different
382
concentrations and (b) SNV-preprocessed original spectra.
383 384
Figure 2. (a) Original FT-IR spectra of pure onion and starch onion mixtures at different
385
concentrations and (b) the same spectra after SNV-preprocessing.
386 387
Figure 3. Beta coefficient derived from the PLSR for the FT-NIR spectral data. Arrows indicate
388
the characteristics wavelengths of starch.
389 390
Figure 4. (a) Principal components score plot for the first three PCs for discrimination among
391
different adulteration concentration in onion powder and (b) first three principal component
392
loadings.
393 394
Figure 5. Beta coefficient derived from the PLSR for the FT-IR spectral data. Arrows indicate
395
the characteristics wavelengths of starch.
396
First three principal component loadings for starch-adulterated onion powder
397
Figure 6.
398
discrimination.
399 400 401 402 20 ACS Paragon Plus Environment
Page 21 of 28
Journal of Agricultural and Food Chemistry
TABLES Table 1. Ability of PLS to predict concentration of added starch in onion powder for SNVpreprocessed FT-IR spectra.
calibration
spectral region (cm−1)
ܴଶ (1
SEC (%)(a
validation ܴ௩ଶ (2
SEV(%)(b
prediction ܴଶ (3
SEP (%)(c
factors
650–4000
0.93
2.52
0.92
2.65
0.90
3.12
3
650–1500
0.93
2.58
0.92
2.70
0.89
3.24
3
1500–4000
0.92
2.65
0.90
3.04
0.84
4.04
4
403
(1
Calibration R2, (2Validation R2, (3Prediction R2. (aSEC, (bSEV, and (cSEP are the standard error
404
of calibration, validation, and prediction, respectively.
21 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 22 of 28
FIGURES
(a)
(b) Figure 1 22 ACS Paragon Plus Environment
Page 23 of 28
Journal of Agricultural and Food Chemistry
(a)
(b)
Figure 2 23 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 24 of 28
Figure 3
24 ACS Paragon Plus Environment
Page 25 of 28
Journal of Agricultural and Food Chemistry
(a)
(b) Figure 4
25 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 26 of 28
Figure 5
26 ACS Paragon Plus Environment
Page 27 of 28
Journal of Agricultural and Food Chemistry
Figure 6
27 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 28 of 28
TABLE OF CONTENTS GRAPHIC
28 ACS Paragon Plus Environment
Article
Detection of starch adulteration in onion powder by FT-NIR and FT-IR spectroscopy Santosh Lohumi, Sangdae Lee, Wang-Hee Lee, Moon S. Kim, Changyeun Mo, Hanhong Bae, and Byoung-Kwan Cho J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/jf500574m • Publication Date (Web): 04 Sep 2014 Downloaded from http://pubs.acs.org on September 9, 2014
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Agricultural and Food Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 28
Journal of Agricultural and Food Chemistry
1
Detection of starch adulteration in onion powder by FT-NIR and
2
FT-IR spectroscopy
3
Santosh Lohumi1, Sangdae Lee1, Wang-Hee Lee1, Moon S. Kim2, Changyeun Mo3, Hanhong
4
Bae4, Byoung-Kwan Cho1,*
5
1
6
Chungnam National University, 99 Daehak-ro, Yuseoung-gu, Daejeon 305-764, Republic of
7
Korea
8
[email protected], [email protected], [email protected]
9
2
Department of Biosystems Machinery Engineering, College of Agricultural and Life Science,
Environmental Microbial and Food Safety Laboratory, USDA-ARS, 10300 Baltimore Ave.,
10
Beltsville, MD 20705, USA
11
[email protected]
12
3
13
Gwonseon-gu, Suwon, Gyeonggi-do 441-100, Republic of Korea
14
[email protected]
15
4
16
[email protected]
17
*Corresponding author: Byoung-Kwan Cho
18
Tel: +82-42-821-6715, Fax: +82-42-823-6246, E-mail: [email protected]
National Academy of Agricultural Science, Rural Development Administration, 150 Suinro,
School of Biotechnology, Yeungnam University, Gyeongsan 712-749, Republic of Korea
1 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 2 of 28
19
Abstract: Adulteration of onion powder with cornstarch was identified by Fourier transform
20
near-infrared (FT-NIR) and Fourier transform infrared (FT-IR) spectroscopy. The reflectance
21
spectra of 180 pure and adulterated samples (1–35 wt% starch) were collected and preprocessed
22
to generate calibration and prediction sets. A multivariate calibration model of partial least-
23
squares regression (PLSR) was executed on the pretreated spectra to predict the presence of
24
starch. The PLSR model predicted adulteration with an ܴଶ of 0.98 and a standard error of
25
prediction (SEP) of 1.18% for the FT-NIR data, and an ܴଶ of 0.90 and SEP of 3.12% for the FT-
26
IR data. Thus the FT-NIR data were of greater predictive value than the FT-IR data. Principal
27
component analysis on the preprocessed data identified the onion powder in terms of added
28
starch. The first three principal component loadings and beta coefficients of the PLSR model
29
revealed starch-related absorption. These methods can be applied to rapidly detect adulteration in
30
other spices.
31
Keywords: Starch, onion powder, adulteration, Fourier transform NIR and IR spectroscopy,
32
partial least-squares regression, principal component analysis
2 ACS Paragon Plus Environment
Page 3 of 28
33
Journal of Agricultural and Food Chemistry
Introduction
34
Spices are used pervasively in our daily lives, such as in the flavoring of food, nutrition,
35
perfumery, medicinal uses, and as preservatives.1 However, spices are common agricultural
36
commodities that are prone to adulteration. For legal or other reasons, adulteration is defined as
37
the addition of extraneous substances that are not normally contained within the original
38
material.2 Food adulteration can be categorized in two ways. Incidental adulteration occurs when
39
foreign substances are incorporated into a foodstuff due to negligence or ignorance. This can
40
occur in the field during harvesting, for example. The second is intentional adulteration, which
41
occurs by either the intent to cause physical or chemical harm, or the motivation for economic
42
gain.3 Adulteration has been commonplace throughout human history, and continues today with
43
several notable instances involving the spice industry. In 1994, ground paprika in Hungary was
44
found to be adulterated with lead oxide, causing the deaths of several people and sickening
45
dozens of others. More recently, ground capsicum was found to contain dyes not approved for
46
use in food.4
47
Recent methods used to authenticate spices have been based on morphological, microscopic,
48
chemical, or DNA-based characterizations. However, these analytical methods may not be
49
convenient for routine sample analysis or require a certain degree of expertise. In some cases,
50
chemical standards may be rare or expensive, or lack identifiable markers.5 Thus, there is still a
51
need to develop a fast, chemical-free, and cost-effective method for the detection of adulteration,
52
which can be applied over a wide range of food products, including spices.
53
Both Fourier transform near-infrared (FT-NIR) and Fourier transform infrared (FT-IR)
54
spectroscopies have been successfully used for qualitative and quantitative analyses of
3 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 4 of 28
55
agricultural products, as well as for pharmaceutical and petrochemical commodities. These
56
spectroscopic methods are based on the chemical composition and moisture content of the
57
biological materials.6 Chemical bonds between light atoms (such as C–H, O–H, and N–H)
58
generally have molecular vibrations, which result in overtones and combination bands that are
59
detectable in the NIR region, 780–2,500 nm.7 The specific patterns exhibited in the NIR region
60
can reveal the chemical and physical compositions of the material being studied. The mid-IR
61
(MIR) region (400–4000 cm−1) monitors the fundamental vibrational and rotational motions of
62
molecules, which provides a chemical profile of the sample. The MIR region is usually
63
considered as a reproducible region of the electromagnetic spectrum in which very small
64
differences in sample composition can be reliably measured.8 In terms of penetration power, NIR
65
radiation can generally penetrate much further into a sample than that of the MIR. The effective
66
path-length through the sample is impacted by the penetration depth of the IR waves. Therefore,
67
NIR spectroscopy is very useful in exploring bulk materials with little or no sample preparation.
68
The difference in the penetration depths between the NIR and MIR regions influences the
69
precision of developed multivariate analytical models as NIR region can penetrate several
70
millimeter into the sample while penetration power of MIR region in microns.9, 10, 11
71
Various studies have been investigated the potential of NIR and MIR spectroscopies to
72
determine melamine adulteration in infant formula powder,12 in dairy milk13 and in soya bean
73
meal.14 The effectiveness of FT-IR spectroscopy and hyperspectral imaging has already been
74
evaluated for the detection of adulteration in black pepper powder.15 In addition, FT-IR
75
spectroscopy in combination with a partial least-squares (PLS) model has also proved competent
76
to detect melamine adulteration in dairy milk.16 The use of FT-Raman and FT-IR spectroscopies
77
was investigated in the characterization of irradiated starch meaning a starch that irradiated at 3, 4 ACS Paragon Plus Environment
Page 5 of 28
Journal of Agricultural and Food Chemistry
78
5, or 10 kGy using a Gammacell 220 Co60 gamma irradiator.17 This research group developed a
79
regression model with the spectroscopic data and observed that NIR performed better than MIR
80
in terms of predicting chemical parameters (e.g., reducing sugars, ethanol, total phenolics, and
81
flavonoids) involved in the fermentation process.
82
Our study sought to quantitatively detect starch in onion powder using FT-NIR and FT-IR
83
spectroscopic techniques. Spectral analysis of eighteen concentrations of starch in onion powder
84
samples, ranging from 0 to 35%, was conducted. Based on the results, different adulterant
85
concentrations were identified and categorized by developing a multivariate analytical method.
86
The ultimate objective of this study was to predict and classify the added starch concentration in
87
onion powder using FT-NIR and FT-IR spectroscopic techniques.
88
89
Materials and Methods
90
Sample and adulterant
91
Onion powder and cornstarch (purities of 100% and 95%, respectively) were obtained from a
92
local market in Korea. Cornstarch was used as the adulterant in the onion powder because of its
93
cheap price, no significant effects on human health, and similar physical property. The samples
94
were stored at room temperature until use.
95
96
Sample preparation
5 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 6 of 28
97
The moisture content of the samples was equilibrated by drying in an oven at 74°C for 2 h prior
98
to mixing because spectroscopic measurement is highly sensitive to moisture content. The given
99
temperature was chosen to dry the sample effectively without interfering with the chemical
100
integrity of the samples.15 Both the dried sample and adulterant material weights were measured
101
with an electronic balance (Model AR2130, Ohaus Corp., USA). Then, the cornstarch was mixed
102
with the onion powder to final concentrations (w/w) of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, 18, 20,
103
25, 30, and 35%. Each combination was manually mixed and then transferred to a snap-cap vial.
104
Further mixing was accomplished by placing the filled vials onto a high speed shaker (Vortex-
105
Genie® 2, Scientific Industries, Inc., Model G560, USA) to reduce the variation in particle size.
106
Ten samples of each of the pure and adulterated powders were used, and sample quantities were
107
approximately 50 mg for the FT-NIR and 30 mg for the FT-IR studies.
108
109
FT-NIR spectra collection
110
NIR reflectance spectra of the pure and adulterated onion powder samples were obtained using
111
an FT-NIR spectrometer (Antaris II FT-NIR analyzer, Thermo Scientific Co., USA) equipped
112
with an InGaAs detector. The sample holder was furnished with an accessory that contained a
113
central hole (20 mm thick × 20 mm diameter) to maintain a uniform shape and thickness for the
114
sample over the irradiated surface during spectra collection. A reference scan was conducted
115
with a golden slit before scanning each sample. The instrument measured the diffuse reflectance,
116
and a total of 32 successive scans from each sample was collected over the wavelength range
117
between 4,000 and 10,000 cm−1 (1,000–2,500 nm) at 4 cm−1 intervals. Averaged spectra were
118
used for analysis.
6 ACS Paragon Plus Environment
Page 7 of 28
Journal of Agricultural and Food Chemistry
119
120
FT-IR spectra collection
121
FT-IR measurements were performed in the mid-infrared region (4000–650 cm−1) using a
122
Nicolet 6700 (Thermo Scientific Co., USA) FT-IR spectrophotometer, configured with an
123
attenuated total reflectance (ATR) sampling technique, deuterated triglycine sulfate (DTGS)
124
detector, and KBr beam splitter controlled by OMNIC software. Samples to be analyzed were
125
placed on a diamond crystal sampling plate and clamped with a pointed tip. A background scan
126
was obtained before every sample scan with an empty sample plate. In addition, the ATR
127
crystals and pointed tip were cleaned to remove any interference from the preceding sample. A
128
total of 32 successive scans from each sample were collected at 4 cm−1 intervals, and averaged
129
spectra of each sample were used for analysis.
130
131
Spectral data analysis
132
The spectroscopic (NIR and mid-IR) detectors receive light coming from the sample in the form
133
of diffuse reflectance after absorption, specular reflectance, and scattered light. Only the diffuse
134
reflectance contains chemical information, whereas the latter two carry no useful information as
135
there is no large variation in particle size between starch and onion powder (based on the result
136
from Cilas 1090 particle size analyzer). The collected spectra might be affected by systemic
137
noise due to light scattering, variation in particle size, and instrumental drift. Thus, in order to
138
determine an accurate chemical composition from spectroscopic measurements, the raw spectra
139
must be corrected at different levels by applying preprocessing methods.18 In this study, both FT-
140
NIR and FT-IR spectral data were treated with 5 pre-processing methods: three data 7 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
methods
(minimum,
maximum,
and
range),
standard
Page 8 of 28
141
normalization
normal
variate
142
transformation (SNV), and multiple scatter correlation (MSC). The original FT-NIR spectra of
143
all the samples, and the original spectra pre-processed by the SNV method are depicted in Figs.
144
1(a) and 1(b), respectively. The raw FT-IR spectra for each sample and SNV-preprocessed
145
spectra are shown in Figs. 2(a) and 2(b), respectively.
146
147
For the data analysis, principal component analysis (PCA) was carried out on the SNV- and
148
MSC-processed spectra from both FT-NIR and FT-IR. PCA can readily be applied to
149
spectroscopic data to perceive the nature and scattering level of the data. The first principal
150
component (PC1) describes the maximum of variation or spread in the samples. A second
151
principal component (PC2), orthogonal to the first, (i.e., completely uncorrelated), describes the
152
maximum of the variation not described by the first eigenvector, and so on. With this procedure,
153
the most important features of the data set can be seen in a low dimensionality plot.19 A
154
multivariate calibration model of a partial least-squares regression (PLSR) was then developed
155
using all the preprocessed spectral data to predict the extent of adulteration in the pure onion and
156
adulterated onion powder samples. PLSR is particularly suited when the matrix of predictors has
157
more variables than observations, such as from spectral data. PLSR has been used in food
158
authentication studies based in spectroscopic data.11 Multivariate analysis was performed using
159
MATLAB software version 7.0.4 (The Mathworks, Nitick, MA, USA).
160
The whole dataset (180 samples) was split into two sets: a calibration set consisting of 126
161
samples (7 samples from each group), and a prediction set consisting of 54 samples (3 samples
162
from each group). The PLSR models were built with the calibration set using a full cross-
163
validation method (leave-one out), removing one observation at one time from the sample sets 8 ACS Paragon Plus Environment
Page 9 of 28
Journal of Agricultural and Food Chemistry
164
until all samples have been removed once. This is the way generally used in developing PLSR
165
model for detecting a specific material with spectral dataset.20
166
167
Result and Discussion
168
Interpretation of spectral analyses
169
The FT-NIR original spectra of the pure and adulterated onion powder, and the SNV-
170
preprocessed spectra (Figs. 1(a) and 1(b)) reveal differences in the absorption intensities between
171
1920–1980 nm (marked) and can easily be identified. The absorption bands in this region
172
correspond to the O–H stretch and O–H band combination and the H–O–H deformation
173
combination, which represent the starch content.21 Starch is a semicrystalline polymer composed
174
of two polysaccharides, amylose and amylopectin. Amylose is a mostly linear chain consisting of
175
up to 3000 glucose molecules interconnected primarily by α-1,4-glycoside linkages.13 The
176
spectral fluctuations from 1400 to 1600 nm correspond to the first overtone of the hydroxyl
177
group. The precise position of these bands is very sensitive to hydrogen bonding in the starch
178
molecule.22
179
The regression coefficient plot (Fig. 3) of the PLS model for the FT-NIR data indicate some
180
meaningful bands that are attributable to the starch. The absorption band around 1420 nm
181
corresponded to a combination of O–H first overtone. Another absorption band at 1920 nm
182
related to CONH, which referred to as amide linkage. The absorption bands at 2312 nm and 2455
183
nm corresponded to C–H stretching and deformation vibrations, which represent starch content.
184
Figures 4(a) and 4(b) show the results of the PCA for the SNV-pretreated FT-NIR spectral data.
9 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 10 of 28
185
A data grouping was observed from the scatter plot of the PC scores (Fig. 4(a)), but the data for
186
samples adulterated at low concentration appear to overlap more. This confirmed our assumption
187
that the different spectral attributes among the samples were associated with the concentration of
188
the starch in the onion powder. Further, we assessed the potential of PCA to distinguish among
189
the low starch concentration samples (1–10%), but the data were overly scattered and
190
overlapping (data not shown). The first three PC loadings account for nearly 99% of the total
191
variance in the samples. PC loadings are correlation coefficients between the PC scores and the
192
original variables, which measure the importance of each variable in accounting for the
193
variability in the PC. It is possible to interperate the first few PCs in term of overall effect of a
194
contrast between groups of variables based on the structures of PC loadings. The PC loading plot
195
(Fig. 4(b)) shows the highest loading for PC1 and PC3 at ~1960 nm, related to the starch. PC1
196
and PC2 show other important bands between 1435 and 1450 nm, corresponding to the first O–H
197
stretching overtone, also assigned to the starch.
198
Typically, MIR region can be divided into two distinct groups: the region of the spectrum below
199
1500 cm−1 is called fingerprint region.23 Peaks in this region are very difficult to assign to
200
specific molecular vibration because each different compound produce a different and unique
201
absorption pattern in this part of spectrum.16 The MIR region from 4000 – 1500 cm-1 is known as
202
the functional group region. Most of the relevant information that is used to interpret the MIR
203
spectrum usually obtains from the functional group region.24 The most notable differences in the
204
FT-IR spectra in this study can be categorized into three main regions, which interpret the
205
significance of key bands (Figs. 2(a) and 2(b)). The first region was the fingerprint region (650–
206
1500 cm−1), which provided more complex spectra and fairly indistinguishable individual bands
207
than the region at higher wavenumbers. The peaks in the region 1045–1087 cm−1 are due to C–C 10 ACS Paragon Plus Environment
Page 11 of 28
Journal of Agricultural and Food Chemistry
208
bond vibrations related to the carbohydrate; those at 1170–1250 cm−1 correspond to absorption
209
bands for C–H, C=O, and C–C vibrations;25 and the triplet bands for the C–O–C stretching
210
absorption are found at 1147 and 1010 cm−1.26 The vibrations originate from the C–O–C bonds
211
related to the α-1,4-glycoside linkages, as these bonds are common in carbohydrates. The C–H
212
stretching region, found between 2850 and 3000 cm−1, includes a distinct band at 2920 cm–1.26
213
Finally, the higher wavenumber region (3000–3500 cm−1) indicates the O–H stretching vibration
214
bands. These peaks could be a result of moisture uptake in the samples during spectral analysis.
215
The beta coefficient (Fig. 5), which indicates the spectral differences among the samples, shows
216
some important bands representative of the carbohydrates (starch) that contribute to the
217
quantitative classification of the onion powder and adulterated samples.
218
PCA was executed on both the MSC- and SNV-preprocessed FT-IR spectral data to extract
219
relevent information. However, the data might be classified according to moisture content or
220
particle size, rather than their chemical composition characteristics, because the IR data patterns
221
are not only determined by absorption characteristics but also scattering characteristics. These
222
spectral variations were removed by oven drying all the samples and pretreating the spectra with
223
MSC- and SNV-preprocessing methods. Neither the MSC- nor the SNV-preprocessed FT-IR
224
data were able to quantify the level of adulteration by means of data clustering using PCA. The
225
data were scattered over a vast range and no distinct separations among the data groups were
226
found from the PCA score plot (data not shown). PCA is a popular primary technique for data
227
clustering. It is not, however optimized for class separability. The loading plots of the first three
228
PCs are depicted in Fig. 6. The PC1 loading accounts for 96.1% of the total sample variance,
229
producing a pattern similar to the raw FT-IR spectra (Fig. 2(a)). The loadings of PC1 and PC2
230
show the highest intensity at ~980 cm−1. Sankaran et al. (2010) also found a similar peak for 11 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 12 of 28
231
starch in the spectral range 950–1000 cm−1.27 In addition, Danwille (2011) observed a sharp peak
232
at 975 cm−1 arising from an –OCH3 bond vibration related to cellulose.15
233
234
Partial least squares regression analysis
235
A PLSR model was used to predict the added starch concentration in onion powder. The samples
236
were categorized based on the adulterant concentration, and a PLSR model was developed using
237
the preprocessed spectra of samples from both FT-NIR and FT-IR spectroscopy. The highest
238
correlation coefficient value, ܴଶ , of 0.98 with a standard error of prediction (SEP) of 1.18% was
239
observed by using the SNV-preprocessed FT-NIR spectra. The optimal number of factors used in
240
the PLS models were automatically selected based on the lowest value of predicted root mean
241
square error (RMSE) in the cross validation process. The PLSR results are listed in Tables 1 for
242
the FT-IR spectral data.
243
The PLSR result obtained for the FT-NIR data shows a strong relatioship (ܴଶ = 0.98) between
244
the actual and predicted concentration values. The result suggests that four principal components
245
were enough to extract 99% of the relevant information used to detect starch adulteration in
246
onion powder.
247
Based on the pattern of the FT-IR spectra, the prediction efficiency of the PLSR model was
248
evaluated using the SNV-preprocessed spectra of the full spectral region, as well as two different
249
selected regions (i.e., the fingerprint region (650–1500 cm−1) and the functional group region
250
(1500–4000 cm−1). The highest correlation coefficients of prediction (ܴଶ ), 0.90 and 0.89, were
251
obtained for the full spectral and fingerprint regions, respectively. In addition, these two spectral
12 ACS Paragon Plus Environment
Page 13 of 28
Journal of Agricultural and Food Chemistry
252
regions showed nearly the same ܴଶ and ܴ௩ଶ values, with a small difference in SEC and SEV
253
(Table 1). The prediction by PLSR of the functional group region was acceptable (ܴଶ = 0.84) but
254
lower than the first two groups.
255
Jawaid et al. (2013) and Sherman (1997) have suggested that the functional group region is better
256
for classification and prediction than the fingerprint region, because most of the information used
257
to interpret the MIR spectrum is usually obtained from the former, whereas the fingerprint region
258
is normally complex with many frequently overlapped bands.16, 28 In our study, the fingerprint
259
region gave better results than the functional group region with respect to the quantative
260
prediction of starch adulteration in the onion powder samples. Moreover, Filippov (1992) also
261
discribed the use of fingerprint region of MIR for the identification of carbohydrates.29 For more
262
information readers can be referred to reference no. 29 and references therein.
263
From the PLSR and PCA results, the FT-NIR spectral data were better able to quantitatively
264
predict the extent of starch adulteration in onion powder samples than the FT-IR spectral data. It
265
should be noted that limited amounts of sample are irradiated by MIR radiation during the FT-IR
266
analysis. Additionally, the penetration of MIR radiation is usually less than that of the NIR.
267
These reasons could be responsible for the low classification rate for the quantitative
268
discrimination of adulteration using FT-IR spectroscopy. It is known that onions contain
269
calcium, proteins, water-soluble carbohydrates (e.g., glucose, fructose, and sucrose), and other
270
chemical components.30 The concentration of added starch in the onion powder proportionally
271
influences the concentration of these chemical compounds on a percentage basis. These chemical
272
compounds have different absorption intensities at specific wavelengths in both the NIR and
273
MIR regions. In our study, in addition to the starch concentration, these intrinsic chemical
274
compounds might be responsible for distinct the classification. 13 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 14 of 28
275
In summary, two different vibrational spectroscopic techniques were investigated to
276
quantitatively detect cornstarch contamination in onion powder. The potential of these
277
techniques was evaluated by adding starch at various concentrations (1–35 wt%) into onion
278
powder and by properly mixing the samples. Both FT-NIR and FT-IR spectroscopic methods
279
demonstrated good potential; however, FT-NIR spectroscopy in combination with PLSR
280
afforded a higher prediction accuracy (ܴଶ = 0.98) than FT-IR spectroscopy (ܴଶ = 0.90). SNV
281
preprocessing, which has been useful in other quantitative applications, improved the accuracies
282
overall. These techniques would seem to be promising for the effective, chemical-free, and rapid
283
detection of product adulteration, as they require no sophisticated laboratory or well-trained
284
personnel to perform the analysis. In addition, for quality control measurements, both FT-NIR
285
and FT-IR spectroscopic techniques are fast and offer a solution for authenticity analysis of
286
agricultural and food products. Further studies could be performed to establish qualitative
287
methods for the detection of adulteration in spices.
288
289
ABBREVIATIONS USED
290
FT-NIR, Fourier transform near infrared; FT-IR, Fourier transform infrared; MIR, mid infrared;
291
PLSR, partial least-square regression; PCA, principal component analysis; SEP, standard error of
292
prediction; DNA, deoxyribonucleic acid; InGaAs, indium gallium arsenide; KBr, potassium
293
bromide; ATR, attenuated total reflectance; DTGS, deuterated triglycine sulfate; SNV, standard
294
normal variate transformation; MSC, multiple scatter correlation; PC, principal component;
295
RMSE, root mean square error; R2, correlation coefficient; ܴଶ , ܴ௩ଶ , ܴଶ are correlation coefficient
14 ACS Paragon Plus Environment
Page 15 of 28
Journal of Agricultural and Food Chemistry
296
of calibration, validation, and prediction, respectively; SEC, standard error of calibration; SEV,
297
standard error of validation ; SEP, standard error of prediction;
298
299
ACKNOWLEDGMENT
300
This research was supported by High Value-added Food Technology Development Program,
301
Ministry of Agriculture, Food and Rural Affairs (MAFRA), Republic of Korea.
15 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 16 of 28
302
REFERENCES
303
(1) Bhattacharjee, P.; Singhal, R.S.; Achyut, S.G. Supercritical carbon dioxide extraction for
304
identification of adulteration of black pepper with papaya seeds. Journal of the Science of Food
305
and Agriculture. 2003, 83, 783-786.
306
(2) Lakshmi, V. Food adulteration. International journal of science invention today. 2012, 1(2),
307
106-113.
308
(3) Everstine, K. Economically motivated adulteration, implication for food protection and
309
alternate approaches to detection. 2013, Ph.D. thesis. Faculty of the graduate school of the
310
University of Minnesota.
311
(4) ASTA (American Spice Trade Association). Spice adulteration – White paper.
312
www.astaspice.org/files/public/SpiceAdulteration.pdf.
313
(5) Shaw, P. C.; Ngan, F. N.; But, P. P. H.; Wang, J. Molecular marker in Chinese medicinal
314
materials. World scientific publishing co. Pte.ltd. 2002, Chapter 1 (1-5).
315
(6) Lohumi, S.; Mo, C.; Kang, J. S.; Hong, S. J.; Cho, B. K. Nondestructive evaluation for the
316
viability of watermelon (Citrullus lanatus) seeds using Fourier transform near infrared
317
spectroscopy. Journal of Biosystems Engineering. 2013, 38(4):312 – 317.
318
(7) Osborne, B. G.; Fearn, T.; Hindle P. T. Practical NIR spectroscopy with applications in food
319
and beverage analysis. Longman Scientific and Technical, Singapore, 1993, 2nd edition.
320
(8) Wen-Sun, D. Infrared spectroscopy for food quality and control 2009, 1st edition. ISBN.
16 ACS Paragon Plus Environment
Page 17 of 28
Journal of Agricultural and Food Chemistry
321
(9) Bras, L. P.; Bernardino, S. A.; Lopes, J. A. ; Menezes, J. C. Multiblock PLS as an approach
322
to compare and combine NIR and MIR spectra in calibrations of soybean flour. Chemomatrics
323
and Intelligent Laboratory System. 2004, 75(2005):91 – 99.
324
(10) Lammertyn, J.; Peirs, A.; Baerdemaeker, J. D.; Nicolai, B. Light penetration properties of
325
NIR radiation in fruit with respect to non-destructive quality assessment. Postharvest Biology
326
and Technology. 1999, 18(2000):121 – 132.
327
(11) Domagala, W. U. The use of the spectrometric technique FTIR-ATR to examine the
328
polymers surface. InTech, (http://creativecommons.org/licenses/by/3.0). 2012, Chepter 3:85 –
329
104.
330
(12) Mauer, L. J.; Chernyshova A. A.; Hiatt, A.; Deering, A.; Davis, R. Melamine detection in
331
infant formula powder using near- and mid-infrared spectroscopy. Journal of Agricultural and
332
Food Chemistry, 2009, 57, 3974 - 3980;
333
334
(13) Balabin, R. M.; Smirnov, S. V. Melamine detection by mid and near-infrared spectroscopy:
335
a quick and sensitive method for dairy products analysis including liquid milk, infant formula,
336
and milk powder. Talanta, 2011, 15;85(1): 562-8
337
(14) Haughey, S. A.; Graham, S. F.; Concouet, E.; Elliott, C. T. The application of near-infrared
338
spectroscopy (NIRS) to detect melamine adulteration of soya bean meal. Food Chemistry, 2012,
339
doi: 10.1016/j.foodchem.1012.01.068.
340
(15) Danwille, J. F. S. Detection and quantification of spice adulteration by near infrared
341
hyperspectral imaging. 2011, Master of Science in food science (Master thesis). 17 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 18 of 28
342
(16) Jawaid, S.; Talpur F.N.; Sherazi S.T.H.; Nizamani S. M.; Khaskheli A. A. Rapid detection
343
of melamine adulteration in dairy milk by SB-ATR–Fourier transform infrared spectroscopy.
344
Food chemistry, 2013, 141, 3066-3071.
345
(17) Ramazan K.; Irudyaraj, J.; Seetharaman, K. Characterization of irradiated starches by using
346
FT-Raman and FTIR spectroscopy. Journal of agricultural and food chemistry 2002, 50(14):
347
3912-8.
348
(18) Norris, K.; Williams, P. Optimization of mathematical treatment of raw near-infrared signal
349
in the measurement of protein in hard red spring wheat. I. Influence of particle size, Cereal
350
Chemistry 1983, vol. 61, pp. 158–165.
351
(19) Benar, P.; Adilson, R. Principal component analysis of the hydroxymethylation of sugarcane
352
lignin. Journal of wood chemistry and technology 1999. 19 (1&2), 151-165.
353
(20) Kandpal, L. M.; Lee, H.; Kim, M. S.; Mo, C.; Cho, B. K. Hyperspectral reflectance imaging
354
technique for visualization of moisture distribution in cooked chicken breast. Sensor 2013. 13,
355
13289 – 13300.
356
(21) Aenugu, H. P. R.; Kumar, D. S.; Parthiban, N.; Ghose S. S.; Banji, D. Near infrared
357
spectroscopy – an overview. International journal of ChemTeCh research 2011. vol. 3, no. 2,
358
825-836.
359
(22) Noah, L.; Robert, P.; Millar, S.; Champ, M. Near-infrared spectroscopy as applied to starch
360
analysis of digestive contents. Journal of agriculture and food chemistry 1997. 45, 2593-2597.
361
(23) – Griffiths, P. R.; de Haseth, J. A. Fourier transform infrared spectroscopy, 2nd edition.
362
2007, John Wiely and Sons, Inc., Hoboken, NJ. 18 ACS Paragon Plus Environment
Page 19 of 28
Journal of Agricultural and Food Chemistry
363
(24) Hsu, C. P. S. Handbook of instrumental techniques for analytical chemistry. Separation
364
sciences research and product development Mallinckrodt, Inc. 1997, Chapter 15:247 – 282.
365
(25) Fagan, C. C.; Donnell, O. Application of Mid-infrared Spectroscopy to food processing
366
system in nondestructive testing of food quality. J. Irudayaraj, C. Reh. Eds. 2008, pp. 119-142.
367
(26) Razali, M. A. A.; Sanusi, N.; Ismail, H.; Othman, N.; Ariffin, A. Application of response
368
surface methodology (RSM) for optimization of cassava starch grafted polyDADMAC synthesis
369
for cationic properties. Starch - journal. 2012, 64, 935-943.
370
(27) Sankaran, S.; Ehshani, R.; Etxeberria, E. Mid-infrared spectroscopy for the detection of
371
Huanglongbing (greeening) in citrus leaves. Talanta. 2010, 83, 574-581.
372
(28) Sherman, C. P. Infrared spectroscopy. Handbook of Instrumental Techniques for Analytical
373
Chemistry. Separation sciences research and product development mallinckrodt, Inc. Chapter 15,
374
1997, 247 – 282.
375
(29) Filippov, M. P. Practical infrared spectroscopy of pectic substances. Food Hydrocoloids,
376
1992, 6(1), 115 – 142.
377
(30) Wall, D.; Mishra, M.; Corgan, J. N. Dehydrate onion bulb weight and water-soluble
378
carbohydrates before and after maturity. Journal of American society for horticultural science.
379
1999, 124(6):581–586.
19 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 20 of 28
380
LIST OF FIGURE CAPTIONS
381
Figure 1. (a) Original FT-NIR spectra of pure onion and starch onion mixtures at different
382
concentrations and (b) SNV-preprocessed original spectra.
383 384
Figure 2. (a) Original FT-IR spectra of pure onion and starch onion mixtures at different
385
concentrations and (b) the same spectra after SNV-preprocessing.
386 387
Figure 3. Beta coefficient derived from the PLSR for the FT-NIR spectral data. Arrows indicate
388
the characteristics wavelengths of starch.
389 390
Figure 4. (a) Principal components score plot for the first three PCs for discrimination among
391
different adulteration concentration in onion powder and (b) first three principal component
392
loadings.
393 394
Figure 5. Beta coefficient derived from the PLSR for the FT-IR spectral data. Arrows indicate
395
the characteristics wavelengths of starch.
396
First three principal component loadings for starch-adulterated onion powder
397
Figure 6.
398
discrimination.
399 400 401 402 20 ACS Paragon Plus Environment
Page 21 of 28
Journal of Agricultural and Food Chemistry
TABLES Table 1. Ability of PLS to predict concentration of added starch in onion powder for SNVpreprocessed FT-IR spectra.
calibration
spectral region (cm−1)
ܴଶ (1
SEC (%)(a
validation ܴ௩ଶ (2
SEV(%)(b
prediction ܴଶ (3
SEP (%)(c
factors
650–4000
0.93
2.52
0.92
2.65
0.90
3.12
3
650–1500
0.93
2.58
0.92
2.70
0.89
3.24
3
1500–4000
0.92
2.65
0.90
3.04
0.84
4.04
4
403
(1
Calibration R2, (2Validation R2, (3Prediction R2. (aSEC, (bSEV, and (cSEP are the standard error
404
of calibration, validation, and prediction, respectively.
21 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 22 of 28
FIGURES
(a)
(b) Figure 1 22 ACS Paragon Plus Environment
Page 23 of 28
Journal of Agricultural and Food Chemistry
(a)
(b)
Figure 2 23 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 24 of 28
Figure 3
24 ACS Paragon Plus Environment
Page 25 of 28
Journal of Agricultural and Food Chemistry
(a)
(b) Figure 4
25 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 26 of 28
Figure 5
26 ACS Paragon Plus Environment
Page 27 of 28
Journal of Agricultural and Food Chemistry
Figure 6
27 ACS Paragon Plus Environment
Journal of Agricultural and Food Chemistry
Page 28 of 28
TABLE OF CONTENTS GRAPHIC
28 ACS Paragon Plus Environment
