Authentication of Whey Protein Powders

by Portable Mid-Infrared Spectrometers

Combined with Pattern Recognition Analysis
Ting Wang, Siow Ying Tan, William Mutilangi, Didem P. Aykas, and Luis E. Rodriguez-Saona

C: Food Chemistry
Abstract: The objective of this study was to develop a simple and rapid method to differentiate whey protein types
(WPC, WPI, and WPH) used for beverage manufacturing by combining the spectral signature collected from portable
mid-infrared spectrometers and pattern recognition analysis. Whey protein powders from different suppliers are produced
using a large number of processing and compositional variables, resulting in variation in composition, concentration,
protein structure, and thus functionality. Whey protein powders including whey protein isolates, whey protein concentrates
and whey protein hydrolysates were obtained from different suppliers and their spectra collected using portable mid-
infrared spectrometers (single and triple reflection) by pressing the powder onto an Attenuated Total Reflectance (ATR)
diamond crystal with a pressure clamp. Spectra were analyzed by soft independent modeling of class analogy (SIMCA)
generating a classification model showing the ability to differentiate whey protein types by forming tight clusters with
interclass distance values of >3, considered to be significantly different from each other. The major bands centered at 1640
and 1580 cm−1 were responsible for separation and were associated with differences in amide I and amide II vibrations
of proteins, respectively. Another important band in whey protein clustering was associated with carboxylate vibrations
of acidic amino acids (1570 cm−1 ). The use of a portable mid-IR spectrometer combined with pattern recognition
analysis showed potential for discriminating whey protein ingredients that can help to streamline the analytical procedure
so that it is more applicable for field-based screening of ingredients.

Keywords: ingredient authentication, pattern recognition, portable mid-infrared spectrometer, whey protein powder

Practical Application: A rapid, simple and accurate method was developed to authenticate commercial whey protein
products by using portable mid-infrared spectrometers combined with chemometrics, which could help ensure the
functionality of whey protein ingredients in food applications.

Introduction
Whey proteins are versatile food ingredients providing func-
tional properties including gelation, high solubility, water holding
capacity, foaming/emulsification and sensory characteristics
enabling their use in numerous food applications (Lee and others
1992; Bouaouina and others 2006; Dissanayake and others 2010).
Commercial products include whey protein concentrates (WPC),
isolates (WPI) and hydrolysates (WPH) differing on protein
concentration and composition related to processing methods
(Foegeding and Luck 2011; Jelen 2011). Separation processes
and thermal conditions have the most significant impact on
composition, determining the concentrations of protein and
nonprotein components, and also modifying protein structure
which impart functionality to food products (Foegeding and Luck
2011; Jelen 2011). Advances in processing technology, including
ultrafiltration, microfiltration, reverse osmosis, and ion-exchange,
have resulted in development of several different finished whey
products: WPC containing between 50% and 85% protein on a
MS 20141783 Submitted 10/26/2014, Accepted 7/23/2015. Authors Wang,
Aykas, and Rodriguez-Saona are with Dept. of Food Science and Technology, The
Ohio State Univ.,110 Parker Food Science and Technology Building, 2015 Fyffe
Road, Columbus, Ohio, 43210, U.S.A. Authors Tan and Mutilangi are with
Pepsi-Cola Company, 100 Stevens Ave, Valhalla, N.Y. 10595, U.S.A. Author
Aykas is with Dept. of Food Engineering, Faculty of Engineering, Adnan Menderes
Univ., Aydin, 09100, Turkey. Direct inquiries to author Rodriguez-Saona (E-mail:
rodriguez-saona.1@osu.edu).
dry basis, and WPI containing between 90% and 98% protein and
very small amounts of lactose and fat, reduced lactose whey, dem-
ineralized whey and hydrolyzed whey (Huffman 1996; Tunick
2008). The higher level of protein in WPI has found applications
in nutritional supplements, sports and health drinks, and protein-
fortified beverages (Foegeding and Luck 2011). WPH is produced
by enzymatic hydrolysis; hydrolysates may contain peptides with
bioactivity, enhanced digestibility, reduced allergenicity, improved
heat stability, and altered functional properties of gelation, foaming
and emulsification. WPH has become an important ingredient in
infant formula and in enhanced-performance products for muscle
recovery (Foegeding and others 2002; Foegeding and Luck
2011).
The desirable nutritional, functional and biological traits of
whey ingredients have positioned them as the high-quality and
preferred ingredient option for the foods and beverages (Smithers
2015). Ingredient authentication of commercial whey products
would limit fraud but also help ensure manufacturers in obtain-
ing the desired functionality in their product. Electrophoresis,
ion exchange chromatography, gel-permeation or size exclusion
chromatography, and high performance liquid chromatography
(HPLC) techniques have been proven to be effective approaches
towards characterization of individual protein type within com-
plex mixtures of whey products (Kilara 2008). In protein research,
infrared (NIR and mid-IR) spectroscopy has been applied for
the qualitative or quantitative determination of protein ingredi-
ents (Baer and others 1983; Marchi and others 2009) and to study

C 2015 Institute of Food Technologists

 R

doi: 10.1111/1750-3841.13006 Vol. 80, Nr. 10, 2015 r Journal of Food Science C2111
Further reproduction without permission is prohibited
 Authentication of whey protein powders . . .

secondary structure of proteins (Curley and others 1998). van der
Ven and others (2002) reported the combination of mid-infrared
spectroscopy and multivariate data analysis in protein hydrolysate
characterization. Their findings showed that mid-infrared spec-
tra correlated to various functional properties of whey and casein
hydrolysates to speed up product development (van der Ven and
others 2002). Miniaturization of vibrational spectrometers into
C: Food Chemistry
Fourier-transform mid-infrared (mid-IR) spectroscopy
Spectra measurements were performed using 2 portable mid-
infrared spectrometers, Cary 630 and 4500 series (Agilent Tech-
nologies, Santa Clara, Calif., U.S.A.). The portable Cary 630
spectrometer weighed 3.6 kg (8 lb) and was equipped with a
temperature-stabilized DTGS detector, permanently-aligned op-
tics, ZnSe beam splitter, 60° air bearing Michelson interferome-

commercially available portable and handheld infrared (IR) and ter and spectrum was collected using a single-reflection diamond
near-infrared (NIR) spectrometers has occurred within the last few ATR sampling interface with a 1 mm diameter sampling sur-
years, partly driven by developments in micro-electro-mechanical face, 200 μm active area providing 2 μm depth of penetration
systems production, and development of lasers, optical compo- for IR energy at 1700 cm−1 . The 4500 Series FTIR is equipped
nents and detectors that can be thermoelectrically air-cooled with a Michelson interferometer, Zinc selenide beamsplitter, low-
enabling miniaturization of spectrometers without sacrificing powered solid-state laser, wire-wound element infrared source and
performance (Sorak and others 2012). Portable vibrational spec- thermoelectrically cooled DTGS detector. All internal compo-
trometers are uniquely positioned for rapid on-site identification nents are mounted on shock-dampening platforms to protect from
and nondestructive analysis for authenticity of incoming ingredi- damage in the field. The 4500 series was interfaced with a triple
ents because of its speed, ruggedness, compactness, ease of use and reflection diamond crystal ATR accessory with a 2 mm diameter
transportation. The objective of this study was to develop a sim- sampling surface and 200 μm active area providing 6 μm effective
ple and rapid method to differentiate whey protein types (WPC, penetration depth for IR energy at 1700 cm−1 .
WPI, and WPH) used for beverage manufacturing by combin- The powder samples were pressed onto the diamond crystals
ing a portable mid-infrared spectrometer (Figure 1) and pattern using a pressure clamp with a slip clutch press providing consistent
recognition analysis. pressure. The IR spectra were collected using MicroLab software
(Agilent Technologies, Santa Clara, Calif., U.S.A.) operating in
the wavenumber ranges from 4000 to 700 cm−1 with resolution
Materials and Methods of 4 cm−1 , and 64 scans were co-added to increase signal to noise
Whey protein samples ratio. Two independent spectra were collected for each powder
Commercial whey protein ingredients used for beverage prod- sample. Both systems were equipped with compression clamps
ucts were coded with blind numbers and included 12 WPI rotated to a maximum pressure (10000 psi) to ensure the best
products (1 to 3 lots/product, n = 25), 9 WPC products contact and highest sampling sensitivity.
(1 to 3 lots/product, n = 14) and 9 WPH products (3 lots/product,
n = 18) were obtained from several manufacturers. Sample infor- Multivariate classification analysis
mation, including moisture, protein, ash, fat, and sugar was pro- The spectra collected were analyzed by the supervised pattern
vided for each powder from individual manufacturers. The pH recognition discriminant analysis, soft independent modeling of
of powders was measured by dissolving 5 g of each powder into class analogy (SIMCA), using the chemometrics modeling soft-
240 mL of deionized water and immersing a micro-pH electrode ware Pirouette (v4.0, Infometrix Inc., Woodville, Wash., U.S.A.).
probe (Micro-Inlab Pro, Mettler Toledo, Greifensee, Switzerland) SIMCA is a classification algorithm based on principal component
interfaced to a SevenCompact pH/ION S220 meter (Mettler (PC) analysis that is applied to each category of interest separately
Toledo, Switzerland). In addition, a soy protein powder, gelatin generating a “training set” to build the classification model (Bran-
powder and milk proteins were included as non-whey proteins to den and Hubert 2005). A SIMCA model consists of a collection
test the predictive ability of the classification analysis. of PCA models, one for each defined classes. PCA is separately
calculated on the objects of each class and cross-validation is used
to choose the number of retained components of each class model.
In this way, SIMCA defines subspaces (class models); then, a new
object is projected in each subspace and compared to it in order
to assess its distance from the class. Finally, the object assignation
is obtained by comparing the distances of the object from the class
models (Ballabio and Todeschini 2009). Spectral data were vector-
length normalized and a cross validation algorithm was used to
determine the number of principal components describing the
systematic variation from spectral data (Wold 1978; Wold and oth-
ers 1983), thus, avoiding over-parameterization or modeling of the
noise (Bro and others 2008). SIMCA models were evaluated in
terms of the matrix of scores and loadings, discriminating power
and interclass distances (Albano and others 1978; Branden and
Hubert 2005). Scores-plot projections of the original data onto
principal components axes were used to visualize sample cluster-
ing (patterns, groupings, or outliers) while the loadings plot shows
the impact of the variables (wavenumbers) on each vector. A crit-
Figure 1–Portable mid-infrared spectrometers, (A) Cary 630 and (B) 4500 ical distance obtained by means of Fisher statistics determined the
series (Agilent Technologies, Santa Clara, Calif., U.S.A.) equipped with
a single- and triple-reflection diamond ATR sampling interfaces. Whey
boundaries of surrounding each class in the SIMCA model. In-
powder samples were pressed onto the diamond crystals using a pressure terclass distance is the critical parameter determined by measuring
clamp with a slip clutch press. the distance between the geometric centers of 2 clusters (Albano

C2112 Journal of Food Science r Vol. 80, Nr. 10, 2015

Authentication of whey protein powders . . .
and others 1978), and this is conducted by the comparison of
the F-statistic with a given confidence interval, which is normally
95% (Meza-Márquez and others 2010). When F value is greater
than the critical F value and the probability is less than the critical
value (α), the interclass distance can be identified as significantly
different and hence the unknown spectrum does not belong to
the class being modeled (Meza-Márquez and others 2010). Sam-
ple residuals and Mahalanobis distances were used to determine
outliers (Hruschka 2001). An external validation sample set using
non-whey proteins was used to evaluate the predictive ability of
the calibration model.
Results and Discussion
The spectrum of commercial whey protein powders was col-
lected by ATR technique using a portable mid-infrared spec-
trometer (Figure 2A). Whey protein powders (WPI, WPC, and
WPH) showed distinct bands associated with functional group
vibrations in the information-rich region of 1800 to 900 cm−1 .
Both portable mid-IR units provided similar spectral patterns but
the triple reflection ATR accessory (4500 series) gave higher sig-
nal intensity compared to the single reflection (Cary 630) re-
solving minor spectral bands from low concentration components
(Figure 2B). The large broad bands in the raw spectrum centered
at approximately 1640 cm−1 were associated with amide I (α-helix
and β-sheet structures of proteins); while the signal at 1520 cm−1
corresponded to amide II (combined δ(N-H) and ν(C-N)) and is
attributed to the peptide bond group vibrations in proteins (Kong
A
acyl chain of TAG
C-H stretching of
O-H groups
Absorbance
WPH
WPC
WPI
3800 3600 3400 3200 3000 2800 2600 2400 2200 2000 1800 1600 1400 1200 1000 800
Wavenumber (cm-1)
B
Absorbance
3-reflections
3800 3400 3000 2600
Wavenumber
and Yu 2007). The band at approximately 1450 cm−1 resulted
from the deformation bending of C–H in the >CH2 groups and
the band at 1390 cm−1 corresponded to the stretching of C=O in
the COO− groups (Coates 2000). A distinct feature in the spectra
was the broad peak band at approximately 1250 cm−1 which cor-
responded to the combined signal from the amide III groups and
P=O stretching in phosphodiesters (>PO2− ) (Coates 2000; Kong
C: Food Chemistry
and Yu 2007). The bands between 1200 and 900 cm−1 were asso-
ciated with vibrations of C–O–C and C–O, attributed to sugars,
acids and/or polysaccharide components of the samples (Shiroma
and Rodriguez-Saon 2009).
Prior to data analysis, each spectrum was vector-length normal-
ized to minimize the effects of systematic error in data collection
(Zeaiter and others 2005; Rinnan and others 2009). The SIMCA
classification plots (Figure 3) showed well separated clustering
among the powder samples whose orientation in the 3D space cor-
related with whey protein type (that is, WPC, WPI, and WPH).
Figure 3A shows that the portable Cary 630 single reflection ATR
unit grouped the whey protein powders into 5 clusters defined by
pH (neutral or acidic) associated with WPI (2 classes) and WPH
(2 classes) while all WPC samples were classified in a single cluster.
A similar clustering pattern was obtained using the triple reflection
ATR 4500 unit (Figure 3B). High protein concentration (>65%)
WPC are produced by ultrafiltration and diafiltration to effec-
tively remove lactose, minerals, and other low-molecular-weight
components while WPI and WPH are manufactured using vari-
ous technologies such as ion exchange, microfiltration followed by
Figure 2–(A) Mid-Infrared spectra of
CH groups of lipids
Whey protein powders by mid-IR
Spectroscopy and (B) differences in
Amide II
Amide I
absorbance band intensity using portable
υC-O-C of single reflection (Cary 630) and
triple-reflection (4500 series) mid-IR
C=O groups of lipids
polysaccharides
spectrometers with Attenuated Total
Reflectance (ATR) accessories.
1-reflection
2200 1800 1400 1000
Vol. 80, Nr. 10, 2015 r Journal of Food Science C2113
 Authentication of whey protein powders . . .

ultrafiltration, electrodialysis, nanofiltration, controlled enzymatic Table 1–Composition of whey protein powders used for beverage
hydrolysis (Morr and Ha 1993; Jelen 2011); the filtration pro- formulations.
cess strongly influences the composition of the final powder. The Nutrient component Whey Whey Whey
pH (6 to 7) and protein content (78.5 ± 2.5%) of WPC powders protein protein protein
were comparable among samples (Table 1). concentrate isolate hydrolyzate
SIMCA results showed that WPI classification was based on Protein (% db)a Range 74–80 90–98 80–90
powder pH level with classes discriminated as having pH ranges Average 78.5 92.6 83.1
C: Food Chemistry

of 5.8 to 7.5 (WPI –- I) and 3.0 to 3.8 (WPI – II); the group-
ing was most likely associated with different processing conditions
and technologies used for manufacturing the WPI powders. The
pH of whey (produced from acid or sweet cheese whey) dur-
ing ultrafiltration, dialysis or diafiltration processes influences both
the mineral composition and the functional properties of WPC
(Hiddinck and others 1981; De Rham and Chanton 1984) and
WPI (Lupano and others 1992). Table 1 showed that protein levels
(92.6 ± 3.1%) of WPI samples were comparable among samples.
van der Ven and others (2002) reported on the capability to utilize
FTIR spectra to distinguish protein hydrolysates prepared from
different protein (casein or whey) sources and proteolytic enzymes
(acidic or neutral/alkaline). Finally, WPH grouping was based on
the presence of samples containing casein glycomacropeptide pro-
tein (cGMP). Glycomacropeptide (GMP) is a C-terminal part
of kappa- casein released in whey during cheese making by the
action of chymosin, providing unique biological and functional
(wide pH range of solubility, emulsifying and foaming) properties
(Neelima and others 2013).
Examination of the discrimination power plot provides in-
formation on spectral patterns (specificity) and band intensity
important for classification (discrimination or recognition) pur-
poses. Figure 3C showed that the SIMCA model generated using
the spectra collected by the Cary 630 equipped with a single
reflection ATR unit had a major discriminating band at 1630
(amide I) and 1578 (amide II) cm−1 . The absorbance in the amide
I band region is predominantly due to the stretching vibration of
the carbonyl peptide bond, whose frequency is highly sensitive
to protein secondary structure (Kong and Yu 2007). The amide I
Std Dev 2.5 3.1 3.8
pH Range 6–7 3.0–6.5 6.1–7.5
Average 6.6 5.1 7.0
Std Dev 0.4 1.4 0.5
a
Reported protein levels provided by the manufacturer’s product information.
band with a maximum centered near 1633 cm−1 is characteristic
of β-sheet (Farrell Jr. and others 2001). The amide II mode is
the out-of-phase combination of the NH in- plane bend and the
CN stretching vibrations (Barth and Zscherp 2002) and is sensi-
tivity to the protonation state of the peptide unit (DeFlores and
others 2009). The shoulder at 1570 cm−1 arises from the anti-
symmetric COO– stretching vibrations of the carboxylate moiety
of the amino acid side-chain groups of glutamate and aspartate
and is highly characteristic of the spectral changes associated with
calcium binding to the carboxylate ligands (Barth 2000; Fabian
and Vogel 2002). Finally, the band at approximately 1485 cm−1
could be attributed to aliphatic moieties of amino acid side chains
and the amide II’ vibrational modes (Barth 2000; Barth and
Zscherp 2002). The discriminating power plot for the model cre-
ated using spectra from the portable 4500 series unit (Figure 3D)
also showed the importance of the amide I and II vibrations, the
shoulder at 1570 cm−1 due to Ca++ binding carboxylate modes
and the band centered at 1485 cm−1 in the classification of the
powders.
SIMCA classification models showed large interclass distance
values (Table 2) for discrimination among WPI, WPC and WPH
samples, demonstrating the ability of mid-infrared spectroscopy
combined with chemometrics to achieve reliable resolution of

A B Figure 3–Soft independent modeling of class analogy

(SIMCA) classification plots for calibration models
PC2 WPH-II PC2 using (A) a portable Cary 630 and (B) 4500 series
WPH-II Attenuated Total Reflectance (ATR) mid-IR
spectrometers. SIMCA discriminating power based on
WPH-I WPH-I the mid-infrared spectra of whey protein powders
acquired using a (C) Cary 630, and (D) 4500 series
PC3 PC1 portable mid-IR spectrometers.
PC3 PC1
WPI-I WPI-I
WPC WPI-II
WPC
1-Bounce WPI-II 3-Bounce

C D 1586
1638 1578 1569
Discriminating Power

800
Discriminating Power

1640
800
1485

400 1569
400

1485

0 0

1578 1464 1580 1465

Wavenumber (cm -1) Wavenumber (cm-1)

C2114 Journal of Food Science r Vol. 80, Nr. 10, 2015

Authentication of whey protein powders . . .

Table 2–Interclass distances generated from whey protein pow- respectively. Overall, the interclass distances showed excellent pre-
der calibration model using bench-top (A) and portable (B) mid- dictive ability of the model for discriminating among the different
IR spectroscopy.
commercial whey protein powders.
Cary 630-1 reflection Finally, the predictive ability of the model was tested using soy
CS1@3a CS2@3 CS3@3 CS4@3 CS5@3 protein isolate, gelatin protein hydrolyzate and milk protein isolate
to determine the model performance in predicting proteins from
CS1b 0.0 different sources. SIMCA’s Coomans plot determined class mem-

C: Food Chemistry
CS2 4.6 0.0
CS3 7.6 9.1 0.0 bership in terms of distance from the boundaries (95% confidence
CS4 10.3 10.8 16.3 0.0 limits) of the categories generated by the classification model. In a
CS5 52.3 47.8 13.2 84.3 0.0 Coomans plot (Figure 4), the 2 axes represent the distance of each
4500 series- 3 Reflections sample from a specific class (for example, WPC, WPI, or WPH),
CS1@3 CS2@3 CS3@3 CS4@3 CS5@3 so that each class model is drawn as a rectangle corresponding to
CS1 0.0 the critical distance (P = 0.05) from the class. Any sample having
CS2 3.3 0.0 a distance to the corresponding class rectangle greater than the
CS3 5.9 6.1 0.0
CS4 7.3 9.1 11.1 0.0
critical distance is considered as being outside the class model and,
CS5 34.4 35.9 9.5 49.4 0.0 as a consequence, rejected as not belonging to that specific class
a
(graphically, it is plotted outside the rectangle defining the class
CS1@3: this indicated the number of factor used in SIMCA model. Classes were
assigned as: CS 1: WPI (I), pH of protein powder ranging from 5.8 to 7.5, CS 2: WPC, model). Moreover, the samples plotted onto the lower left square
CS 3: WPH (I), CS 4: WPI (II), pH ranging from 2.8 to 3.8, CS 5: WPH (II), CGMP. of the diagram are assigned to both classes. The results showed
pH was measured by dissolving 5g of whey powder in 8 oz of deionized water.
b
CS: CS indicates class assigned to individual category of whey powder products. that these “confounding” proteins were reliably predicted as not
belonging to any of the whey protein classes by using the models
generated from spectra by Cary 630 (Figure 4A and B) and 4500
series (Figures 4C and D) spectrometers. Nevertheless, the mod-
unique spectral markers. Interclass distance is the distance between els developed by the 4500 series with the triple reflection ATR
the geometric centers of two clusters; it represents the performance accessory gave higher critical distances for the non-whey protein
of separation with larger interclass distance resulting in better sep- than the Cary 630 with the single reflection ATR because of the
aration (Albano and others 1978). Interclass distances greater than increase signal intensity and ability to resolve minor spectral bands.
3 are considered significant for identification of data points as
members of a group (Kvalheim and Kratang 1992). All interclass
distances had values larger than 3, clearly demonstrating class dis- Conclusion
crimination and statistical significance (P < 0.05). Interestingly, Portable ATR-IR spectroscopy combined with pattern recog-
class 5 (WPH-III) showed large interclass distances compared to nition chemometrics had the ability to distinguish and authen-
the rest of the classes (9.5 to 84), which indicated that the CGMP ticate commercial whey protein ingredients. Although whey
powder had marked different compositional characteristics from protein powders were clustered according to type (WPC, WPI,
others. Most whey powder samples fell in classes WPC, WPI (I) and WPH), WPI powders were further discriminated by pH
and WPH (I), showing interclass distances of 4.6 and 3.3 (WPC while WPH powders containing casein glycomacropeptide pro-
vs. WPI (I)), 7.6 and 5.9 (WPI (I) vs. WPH (I)) and 9.1 and 6.1 tein (cGMP). Soy protein isolate, gelatin hydrolysate and milk
(WPC vs. WPH (I)) for models developed with the Cary 630 protein isolates were reliably predicted as not belonging to any
single reflection and 4500 series triple reflection ATR-IR units, of the whey protein classes. The portable mid-IR spectrometer

A SPI
B Figure 4–SIMCA’s Coomans plot for class
SPI membership of soy protein isolate (SP), gelatin
hydrolyzate (GP) and milk protein isolates (MPI)
GPH using the portable Cary 630 (A and B) and 4500
GPH series (C and D) mid-IR spectrometers.
WPH@3
WPI@3

MPI

MPI

WPC@3 WPC@3
C SPI D SPI

GPH
WPH@3
WPI@3

MPI
MPI
GPH

WPC@3
WPC@3

Vol. 80, Nr. 10, 2015 r Journal of Food Science C2115

 Authentication of whey protein powders . . .
equipped with a triple reflection accessory showed improved speci-
ficity for rejecting samples extraneous to the whey protein classes.
This research showed the feasibility of portable mid-IR spectrom-
eters as an effective tool in discriminating whey protein ingredients
for “field-based” authentication.
Acknowledgment
C: Food Chemistry
We would like to thank The Pepsi-Cola Company for providing
materials for the research and for their financial support. The
authors hereby declare that there is no conflict of interest.
References
Albano C, Dunn III W, Edlund U, Johansson E, Norden B, Sjöström M, Wold S. 1978. Four
levels of pattern recognition. Anal Chim Acta 103:429–43.
Baer RJ, Frank JF, Loewenstein M, Birth GS. 1983. Compositional analysis of whey powders
using near infrared diffuse reflectance spectroscopy. J Food Sci 48:959–61.
Ballabio D, Todeschini R. 2009. Multivariate classification for qualitative analysis. In: Da-Wen
S, editor. Infrared spectroscopy for food quality analysis and control. London, UK: Academic
Press/Elsevier. p 83–104.
Barth A. 2000. The infrared absorption of amino acid side chains. Prog Biophys Mol Biol
74:141–73.
Barth A, Zscherp C. 2002. What vibrations tell us about proteins. Q Rev Biophys 35:369–430.
Bouaouina H, Desraumax A, Loisel C, Legrand J. 2006. Functional properties of whey proteins
as affected by dynamic high-pressure treatment. Intl Dairy J 16:275–84.
Branden KV, Hubert M. 2005. Robust classification in high dimensions based on the SIMCA
method. Chemometr Intell Lab Sys 79:10–21.
Bro R, Kjeldahl K, Smilde AK, Kiers HAL. 2008. Cross-validation of component models: a
critical look at current methods. Anal Bioanal Chem 390:1241–51.
Coates J. 2000. Interpretation of infrared spectra, a practical approach. In: Meyers RA, editor.
Encyclopedia of analytical chemistry. Chichester, NY: John Wiley & Sons Ltd. p 10815–
37.
Curley D, Kumosinski T, Unrah J, Farrell H. 1998. Changes in the secondary structure of bovine
casein by Fourier transform infrared spectroscopy: effects of calcium and temperature. J Dairy
Sci 81:3154–62.
DeFlores LP, Ganim Z, Nicodemus RA, Tokmakoff A. 2009. Amide III 2D IR spectroscopy
provides enhanced protein secondary structural sensitivity. J Am Chem Soc 131:3385–91.
De Rham O, Chanton S 1984. Role of ionic environment in insolubilization of whey protein
during heat treatment of whey products. J Dairy Sci 67:939–49.
Dissanayake M, Kelly AL, Vasiljevic T. 2010. Gelling properties of microparticulated whey
proteins. J Agric Food Chem 58:6825–32.
Fabian H, Vogel HJ. 2002. Fourier transform infrared spectroscopy of calcium-binding proteins.
In: Vogel HJ, editor. Methods in molecular biology, Totowa: Human Press, p 57–74.
Farrell Jr. HM, Wickham ED, Unruh JJ, Qi PX, Hoagland PD. 2001. Secondary structural
studies of bovine caseins: temperature dependence of b-casein structure as analyzed by cir-
cular dichroism and FT IR spectroscopy and correlation with micellization. Food Hydrocoll
15:341–54.
Foegeding EA, Davis JP, Doucet D, McGuffey MK. 2002. Advances in modifying and under-
standing whey protein functionality. Trends Food Sci Technol. 13:151–9.
Foegeding EA, Luck PJ. 2011. Whey protein products. In: Roginski H, Fox PF, Fuquay JW,
editors. Encyclopedia of dairy sciences. London, UK: Academic Press. p 873–8.
C2116 Journal of Food Science r Vol. 80, Nr. 10, 2015
Hiddink J, De Boer R, Nooy PFC. 1981. Effect of various pre-treatments on the ultrafiltration
of sweet cheese whey at about 55 °C. Milchwissenschaft 36:657–63.
Hruschka WR. 2001. Data analysis: wavelength selection methods. In: Williams P, Norris K,
editors. Near-infrared technology in the agricultural and food industries. St. Paul, M inn.,
U.S.A.: American Association of Cereal Chemists, Inc. p 39–58.
Huffman LM. 1996. Processing whey protein for use as a food ingredient. Food Technol 50:49–
52.
Jelen P. 2011. Whey processing. In: Roginski H, Fox PF, Fuquay JW, editors. Encyclopedia of
dairy sciences. London, UK: Academic Press. p 731–7.
Kilara A. 2008. Whey and whey products. In: Chandan RC, Kilara A, Shah NP, editors. Dairy
processing and quality assurance. Ames, Iowa: Wiley-Blackwell. p 337–55.
Kong J, Yu S. 2007. Fourier transform infrared spectroscopic analysis of protein secondary
structures. Acta Biochim Biophys Sin 39:549–59.
Kvalheim OM, Kratang TV. 1992. SIMCA - Classification by means of disjoint cross validates
principal components models. In: Brereton RG, editor. Multivariate pattern recognition in
chemometrics. Ams terdam, N.Y.: Elsevier. p 209–49.
Lee S, Morr CV, Ha EYW. 1992. Structural and functional properties of caseinate and whey
protein isolate as affected by temperature and pH. J Food Sci 57:1210–4.
Lupano CE, Dumay E, Cheftel JC. 1992. Gelling properties of whey protein isolate: influence
of calcium removal by dialysis or diafiltration at acid or neutral pH. Intl J Food Sci Tcchnol
21:615–28
Marchi MD, Bonfatti V, Cecchinato A, Martino GD, Carnier P. 2009. Prediction of protein
composition of individual cow milk using mid-infrared spectroscopy. Ital J Anim Sci 8:399–
401.
Meza-Márquez OG, Gallardo-Velázquez T, Osorio-Revilla G. 2010. Application of mid-
infrared spectroscopy with multivariate analysis and soft independent modeling of class
analogies (SIMCA) for the detection of adulterants in minced beef. Meat Sci 86:
511–9.
Morr CV, Ha EYW. 1993. Whey protein concentrates and isolates: Processing and functional
properties. Crit Rev Food Sci Nutr 33:431–76.
Neelima, Sharma R, Singh Rajput Y, Mann B. 2013. Chemical and functional properties of
glycomacropeptide (GMP) and its role in the detection of cheese whey adulteration in milk:
a review. Dairy Sci Technol 93:21–43.
Rinnan A, vanden Berg F, Engelsen SB. 2009. Review of the most common pre-processing
techniques for near-infrared spectra. Trends Anal Chem. 28:1201-22.
Shiroma C, Rodriguez-Saona L. 2009. Application of NIR and MIR spectroscopy in quality
control of potato chips. J Food Comp Anal. 22:596–605.
Smithers GW. 2015. Whey-ing up the options— yesterday, today and tomorrow. Intl Dairy J
48:2–14. doi:10.1016/j.idairyj.2015.01.011.
Sorak D, Herberholz L, Iwascek S, Altinpinar S, Pfeifer F, Siesler HW. 2012. New developments
and applications of handheld Raman, mid-infrared, and near-infrared spectrometers. Appl
Spectrosc Rev 47 (2):83–115.
Tunick MH. 2008. Whey protein production and utilization: a brief history. In: Onwulata
C, Huth P, editor. Whey processing, functionality and health benefits. Ames, Iowa: Wiley-
Blackwell. p 1–13.
vander Ven C, Muresan S, Gruppen H, Bont DB, Merck KB, Voragen AGJ. 2002. FTIR spectra
of whey and casein hydrolysates in relation to their functional properties. J Agric Food Chem
50:6943−50.
Wold S. 1978. Cross-validatory estimation of the number of components in factor and principal
components models. Technometrics 20:397–405.
Wold S, Martens H, Wold H. 1983. The multivariate calibration problem in chemistry solved
by the PLS method. Matrix Pencils. 973:286–93.
Zeaiter M, Roger J-M, Bellon-Maurel V. 2005. Robustness of models developed by multi-
variate calibration. Part II: the influence of pre-processing methods. Trends Anal Chem 24:
437–45.