Download as pdf or txt
Download as pdf or txt
You are on page 1of 134

The Authentication of Whey Protein Powder Ingredients and Understanding

Factors Regulating Astringency in Acidic Whey Protein Beverages to Estimate


Astringency by Infrared Spectroscopy ― An Instrumental Approach

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University

By

Ting Wang

Graduate Program in Food Science and Technology

The Ohio State University

2014

Dissertation Committee:

Dr. Luis E. Rodriguez-Saona, Advisor

Dr. Monica Giusti

Dr. Michael E Mangino

Dr. Lynn Knipe


Copyrighted by

Ting Wang

2014
Abstract

Whey protein, recognized as a healthy ingredient nowadays, has been widely used

in beverages, bakery products, meat products, and infant formula with attractive

functionalities such as water binding, gelation, emulsification, and foaming. The

processing of whey protein from cheese manufacture has become highly automated and

the production amount has increased largely, which indicated that whey protein has

become to an essential ingredient in food industry. The advanced techniques involving

membrane filtration and ion exchange have made it possible to manufacture specific

product to fulfill the demand of customers. While whey protein has been successfully

transformed from waste to a value-added ingredient, it still confronts problems: The

adulteration of raw material and astringency induced when whey protein beverage was

formulated under acidic pH.

The overall objectives of this dissertation were to provide an accurate and

convenient instrumental method, Infrared Spectroscopy (IR), to authenticate the whey

protein powder and predict the astringency of acidic whey protein beverage, and

understand the factors regulating the whey protein astringency, hence benefit the

application of whey protein as a key ingredient in food products.

Whey protein powders including whey protein isolate (WPI), whey protein

concentrate (WPC), and whey protein hydrolysate (WPH) were obtained from different

ii
manufactures, with different sample characteristics and processing techniques. Each

powder was formulated to a corresponding beverage. The first batch was controlled under

uniform pH value (pH=3.5) after formulation from different powders, and beverage of the

second batch formulated by each whey protein powder was further adjusted to 5 different

pH values (ranging from 2.2 to 3.9). Quantitative sensory test of astringency on all whey

protein beverage samples was conducted at the Sensory Service Center in North Carolina

State University. Infrared spectra of all samples were collected in the wavenumber ranges

from 4000 to 700 cm-1. Classification model and prediction model were built through

multivariate analysis. Multiple linear regression and Pearson correlation were used to

reveal the factors modulating whey protein astringency.

SIMCA classification models with interclass distances larger than 3 were built to

authenticate raw ingredients, and the unknown samples were successfully predicted using

calibration models. The Partial Least Square Regression (PLSR) analysis showed strong

ability to predict astringency based on infrared spectra of different protein products (WPI,

WPC, and WPH beverages). IR bands associated with protein secondary conformation

and carbonyl side chain of amino acids were identified to be significant in both

authentication and prediction. Whey protein types and pH values were suggested to be

significant factors in the development of astringency in whey protein products.

In conclusion, with the comprehensive information obtained from food samples

spectra, IR combined with multivariable chemometrics was an accurate, easy and high

throughput method in understanding and predicting whey protein astringency, which will

be extremely attractive to both scientific research and food industry practice.

iii
Acknowledgments

I sincerely thank Dr. Luis Rodriguez-Saona for his advising, which is not only on

the knowledge and skills as a Ph. D student, but also on the personality in the future of

my life. He taught me how to become a confident people; he guided me how to choose

the career; he supported me to gain experience from food industry; He encouraged me to

conquer problems and fix mistakes. Whenever I became confused, he was always there to

help and gave me the right direction. He is the best advisor I have ever met and it is the

best decision I have ever made to choose him as my advisor.

I also would like to thank my dissertation committee members, Dr. Monica

Giusti, Dr. Michael E Mangino, Dr. Lynn Knipe and Dr. Celia E. Wills. The blessing,

help and guidance given by them always encouraged me; their feedback helped me with a

wonderful dissertation.

I would like to express my special thanks of gratitude to PepsiCo, Inc. for their

generous financial and samples support, as well as Siow Ying Tan, Mutilangi William

and Connie Cerdena who gave me this excellent opportunity to work with them on this

excited project.

I would like to thank the Sensory Service Center at North Carolina State

University, leading by Dr. MaryAnne Drake, for their great efforts on the sensory

evaluation of my samples.

iv
I would like to thank my husband, the most important one in my life.

v
Vita

September 14, 1982 ....................................... Yang Quan, China

2001 – 2005 ................................................... B.S. Bioengineering, South China

University of Technology

2008 - 2011 .................................................. M.S. Food Science and Technology, The

Ohio State University

2011 - present ............................................... Ph.D. Food Science and Technology, The

Ohio State University

Fields of Study

Major Field: Food Science and Technology

vi
Table of Contents

Abstract .............................................................................................................................. ii

Acknowledgments ............................................................................................................. iv

Vita .................................................................................................................................... vi

List of Tables..................................................................................................................... xi

List of Figures ................................................................................................................. xiii

Chapter 1: Literature Review ............................................................................................ 1

1.1 A Review of whey protein processing and functionality .......................................... 1

1.1.1 Whey and whey protein fractions ...................................................................... 1

1.1.2 Whey protein processing and production ........................................................... 5

1.1.3 Whey protein functionality .............................................................................. 14

1.1.4 Whey protein application ................................................................................. 21

1.1.5 Whey protein analytical methods ..................................................................... 22

1.2 Astringency and mechanisms in various food ........................................................ 26

1.2.1 Terminology of astringency ............................................................................. 26

1.2.2 Mechanisms of astringency in four main groups of astringent compounds ..... 27

1.2.3 Exploration of mechanisms of astringency caused by whey protein ............... 33


vii
1.3 The application of Infrared Spectroscopy (IR) on whey protein. ........................... 35

Chapter 2: Authentication of Whey Protein Powders by Combining Infrared

Spectroscopy and Pattern Recognition Analysis .............................................................. 40

2.1 Abstract................................................................................................................... 40

2.2 Introduction ............................................................................................................ 41

2.3 Materials and methods ............................................................................................ 43

2.3.1 Whey protein sample ....................................................................................... 43

2.3.2 Fourier-Transform Infrared Spectroscopy (FT-IR) .......................................... 43

2.3.3 Multivariate classification analysis .................................................................. 44

2.4 Results and discussion ............................................................................................ 46

2.5 Conclusion .............................................................................................................. 55

Chapter 3: A Novel Application of FT-IR Technique Combined with Chemometrics in

Analyzing and Predicting the Astringency of Acidic Whey Protein Beverages .............. 56

3.1 Abstract................................................................................................................... 56

3.2 Introduction ............................................................................................................ 57

3.3 Materials and methods ............................................................................................ 59

3.3.1 Whey protein sample ....................................................................................... 59

3.3.2 Astringency prediction ..................................................................................... 60

viii
3.3.3 FTIR Micro-spectroscopy ................................................................................ 60

3.3.4 Portable FTIR Spectroscopy ............................................................................ 61

3.3.5 Multivariate regression analysis ....................................................................... 61

3.4 Results and discussion ............................................................................................ 63

3.4.1 Perceived astringency scores for whey protein beverages ............................... 63

3.4.2 Comparing spectra information from whey protein powder and beverage ...... 65

3.4.3 SIMCA classification of whey protein beverages with fixed pH ..................... 67

3.4.4 PLSR calibration models of whey protein beverages with fixed pH ............... 69

3.4.5 PLSR calibration models of whey protein beverages with various pH ............ 71

3.4.6 PLSR prediction of a separate set of whey protein beverages ......................... 80

3.5 Conclusion .............................................................................................................. 82

Chapter 4 Understanding Factors Influencing Development of Astringency in Whey

Protein Products ............................................................................................................... 83

4.1 Abstract................................................................................................................... 83

4.2 Introduction ............................................................................................................ 84

4.3 Materials and methods ............................................................................................ 85

4.3.1 Whey protein sample ....................................................................................... 85

4.3.2 Astringency prediction ..................................................................................... 85

4.3.3 Fourier-Transform Infrared Spectroscopy (FT-IR) .......................................... 86

ix
4.3.4 Statistical analysis ............................................................................................ 86

4.3.5 Gas Chromatography (GC) analysis ................................................................ 87

4.4 Results and discussion ............................................................................................ 88

4.5 Conclusion .............................................................................................................. 96

4.6 Significance and future work .................................................................................. 96

List of References............................................................................................................. 98

x
List of Tables

Table 1.1. Characteristic information of protein in Whey. ................................................. 2

Table 1.2. Production (in tons) of whey protein products in the United States ................ 11

Table 1.3. Analytical methods for whey protein contents regulated by Code of Federal

Regulations Title 21. ........................................................................................................ 24

Table 1.4. Charges of Whey proteins and salivary proteins in different solutions and their

potential of interaction. .................................................................................................... 34

Table 1.5. A general band assignment of MIR applied to food matrix. ........................... 36

Table 1.6. Recent researches on induced conformational changes of individual whey

protein. ............................................................................................................................. 37

Table 2.1. Interclass Distances generated from whey protein powder calibration model

using bench-top (A) and portable (B) FT-IR Spectroscopy. ............................................ 53

Table 2.2. Number of protein samples used for spectral acquisition using the bench-top

and portable FT-IR Spectrometers ................................................................................... 54

Table 3.1. Sensory astringency scores perceived from whey protein beverage samples

under various pH value .................................................................................................... 64

Table 3.2. The multiple linear regression using pH levels and protein types as predictor

and astringency as response.. ........................................................................................... 65

xi
Table 3.3. Calibration and cross-validation results of multivariate models developed by

whey protein beverages with fixed pH (3.5) using FTIR-ATR Micro-spectroscopy

spectra. ............................................................................................................................. 71

Table 3.4. Calibration and cross-validation results of multivariate models developed by

whey protein beverages with various pH using FTIR-ATR Micro-spectroscopy, FTIR-

Reflectance Micro-spectroscopy, or Portable FTIR-ATR Spectrometer spectra. ............ 74

Table 3.5. Calibration and cross-validation results of multivariate models developed by

whey protein beverages with fixed and various pH using FTIR-ATR Micro-spectroscopy

spectra. ............................................................................................................................. 79

Table 3.6. Prediction values of astringency in formulated whey protein beverages using

IR spectra and calibration models by FTIR-reflectance Micro-spectroscopy.. ................ 81

Table 4.1. Multiple Linear Regression using pH and whey protein types as predictor and

astringency scores as response for WPI, WPC, and WPH. .............................................. 89

Table 4.2. Calibration and cross-validation results of PLSR models developed by pH of

whey protein beverages using FTIR-Reflectance Micro-spectroscopy spectra................ 90

Table 4.3. Calibration and cross-validation statistic performance of PLSR models

developed by using whey protein powder samples. ......................................................... 93

Table 4.4. Pearson correlation between concentrations of glutamic acid (GLU) and

aspartic acid (ASP) in whey protein powders and the corresponding astringency of

beverages at pH 3.5. ......................................................................................................... 95

xii
List of Figures

Figure 1.1. The amino acids sequence of Bovine β-lactoglobulin. .................................... 3

Figure 1.2. The amino acids sequence of Bovine α-lactalbumin........................................ 4

Figure 1.3. A typical industrial spray drying process. ...................................................... 10

Figure 1.4. Processing of whey protein concentrate......................................................... 12

Figure 1.5. Processing of whey protein isolate................................................................. 13

Figure 1.6. Generalized sequence of proteins-water interaction at increasing ERH. The

actual water content(s) at each Aw varies with the nature and types of protein and the

experimental conditions. .................................................................................................. 15

Figure 1.7. A mouth-feel wheel to classify the sensation elicited by red. ........................ 27

Figure 1.8. The example of hydrolyzable tannins (A) and condensed tannins (B) showing

their molecular structures. ................................................................................................ 28

Figure 1.9. Proposed mechanism for polyphenol-PRPs complex formation .................... 30

Figure 2.1. Mid-Infrared spectra of Whey protein powder from bench-top FT-IR

Spectroscopy (A) and portable FT-IR Spectroscopy (B) with Attenuated Total

Reflectance (ATR) ........................................................................................................... 47

Figure 2.2. Soft independent modeling of class analogy (SIMCA) classification plots for

calibration models using a benchtop (A) and portable (B) FTIR spectrometers. SIMCA

xiii
validation performance for calibration models using a benchtop (C) and portable (D)

FTIR spectrometers based on the infrared spectra of whey protein powders. .................. 49

Figure 2.3. Soft independent modeling of class analogy (SIMCA) discriminating power

based on the infrared spectra of whey protein powder from bench-top FT-IR

Spectroscopy (A) and Portable FT-IR Spectroscopy (B). ................................................ 51

Figure 3.1. Attenuated Total Reflectance (ATR) Mid-Infrared spectra of Whey protein

powder and beverage.. ...................................................................................................... 66

Figure 3.2. Second derivative spectral transformation collected using an Attenuated Total

Reflectance Infrared (ATR-IR) accessory equipped either with a ZnSe crystal plate

(bench-top) or Germanium crystal (Micro-spectroscopy)... ............................................. 67

Figure 3.3. Soft independent modeling of class analogy (SIMCA) classification plot and

discriminating power based on the infrared spectra of whey protein beverage samples.. 68

Figure 3.4. Partial least squares regression (PLSR) plots and loadings based on the

infrared spectra of whey protein WPI, WPC and WPH beverage samples with fixed pH

(3.5) using FTIR-ATR germanium Micro-spectroscopy. ................................................. 70

Figure 3.5. Partial least squares regression (PLSR) plots and loadings based on the

infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH

(ranging from 2.2 to 3.9) using FTIR-ATR germanium Micro-spectroscopy. ................. 73

Figure 3.6. Partial least squares regression (PLSR) plots and loadings based on the

infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH

(ranging from 2.2 to 3.9) using FTIR-reflectance Micro-spectroscopy. .......................... 76

xiv
Figure 3.7. Partial least squares regression (PLSR) plots and loadings based on the

infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH

(ranging from 2.2 to 3.9) using portable FTIR-ATR diamond equipment. ...................... 77

Figure 3.8. Partial least squares regression (PLSR) plots and loadings based on the

infrared spectra of whey protein WPI, WPC and WPH beverage samples with fixed pH

(3.5) and various pH (ranging from 2.2 to 3.9) using FTIR-ATR germanium Micro-

spectroscopy ..................................................................................................................... 78

Figure 4.1. The examination of validity using MLR to analyze WPI samples. Residuals vs

Fits for Astringency (A) and Normal plot of Residuals for Astringency (B). .................. 89

Figure 4.2. PLSR plots and loadings based on the infrared spectra of whey protein WPI,

WPC and WPH beverage samples and measured pH using FTIR-Reflectance Micro-

spectroscopy. .................................................................................................................... 91

Figure 4.3. PLSR plots and loadings based on the infrared spectra of whey protein

powder samples (WPI, WPC, and WPH). ........................................................................ 92

xv
Chapter 1: Literature Review

1.1 A Review of whey protein processing and functionality

1.1.1 Whey and whey protein fractions

Whey protein, discovered 3000 years ago, is generally recognized as a value

added key ingredient in modern food industry (Smithers, 2008). It is defined in the

United States Code of Federal Regulations Title 21:

“Whey is the liquid substance obtained by separating the coagulum from milk,

cream, or skim milk in cheese making. Whey obtained from a procedure, in which a

significant amount of lactose is converted to lactic acid, or from the curd formation by

direct acidification of milk, is known as acid whey. Whey obtained from a procedure in

which there is insignificant conversion of lactose to lactic acid is known as sweet whey.

Sweet whey has a maximum titratable acidity of not more than 0.16 percent, calculated

as lactic acid, and an alkalinity of ash of not more than 225 milliliters of 0.1N

hydrochloric acid per 100 grams. The acidity of whey, sweet or acid, may be adjusted by

the addition of safe and suitable pH-adjusting ingredients.”

Whey is a complex solution containing water, proteins, lactose, minerals, milk fat,

and lactic acid (Morr, 1989). Whey proteins, the components that play the most important

role in functional and nutritional properties when used in food products, are composed of

different types of proteins. The proteins involved in whey include α-lactalbumin (α-LA),

1
β-lactoglobulin(β-LG), bovine serum albumin (BSA), immunoglobulins (Igs), lactoferrin

(LF), lactoperoxidase, glycomacropeptides (GMPs) and others (Laetitia M. Bonnaillie

and Peggy M. Tomasula, 2008). The characteristic information of individual protein is

listed in Table 1.1. Whey proteins possess globular structures with high levels of

secondary, tertiary and quaternary structures (Sne ana ovanovi , irol ub ara ,

gn en a e, ). Particular whey fractions, β-lactoglobulins, α-lactalbumins,

enriched fractions containing lactoferrins or glycomacropeptides were mostly reported

whey protein with health benefits and widely used in food products.

Molecular Weight
Protein Isoelectric pH
(kg/mol)

β-lactoglobulin 18 5.4
Major whey protein α-lactalbumin 14 4.4
Bovine serum albumin 66 5.1
Immunoglobulins 150 5-8
Lactoferrin 77 7.9
Minor whey protein
Lactoperoxidase 78 9.6
Glycomacropeptides 8.6 <3.8
Table 1.1. Characteristic information of protein in Whey (Etzel 2004; Kilara and Vaghela
2004).

β-LG is the major protein in whey, with 58% of total whey protein content. The

functional characteristics dominate the functionality of whey protein concentrations and

whey protein isolates. Native β-LG is a small globular protein with nown primary,

secondary and tertiary structure (Sne ana ovanovi , irol ub ara , gn en a e,


2
). The primary structure of bovine β-LG molecule is composed of 178 amino acid

residues and the sequence information is listed in Figure 1.1. The secondary structure is

composed of α-helical (1 %), β-sheet (51%), reverse turn (17%), and aperiodic (17%)

respectively (Creamer et al., 1983). The tertiary structure contains two disulfide bonds

and one sulfhydryl group burying inside native protein that will be exposed and activated

by denaturation (heat, pH, or agents) (Swaisgood, 1986; Galani and Owusu Apenten,

1999; Sawyer et al., 1999; Fox, 2003).

Figure 1.1. The amino acids sequence of ovine β-lactoglobulin (Uniprot, 2014).

α-LA accounts 31% of the total whey protein. It is a compact and acidic globular

protein, with binding ability to metal ions such as Ca2+ and Zn2+ (Permyakov and

Berliner, 2000). The peptide linkages of 142 amino acids residuals, showing in Figure

1. , illustrate the primary structure of bovine α-LA. There are two domains in native α-

LA including a large u-helical domain connected to a small v-sheet domain by a calcium-

3
binding loop. The tertiary structure involves four disulfide bridges, which can stabilize

the protein molecule (Permyakov and Berliner, 2000).

Figure 1.2. The amino acids sequence of ovine α-lactalbumin (Uniprot, 2014).

α-LA is an important protein due to the functionality it can provide. It is one

component of lactose synthase in mammary secretory cells and participates the

interaction with galactosyltransferase to increase the enzyme’s affinity and specificity for

glucose (Hill and Brew, 197 ). With similar amino acid composition to human mil , α-

LA is also considered as the most suitable protein as the main ingredient in infant

formular (Heine et al., 1991). One research conducted on 23 recovered depressed patients

and 20 controls with the dietary supplement rich in whey protein α-LA showed the

enhancing effect on participants’ memory, irrespective of history of depression ( ooi et

al., 6). There have been more studies done to reveal the functionality of α-LA in

cognitive performance improvement, antimicrobial property, and others (Permyakov and

4
Berliner, 2000; Markus et al., ), which indicated that the significance of α-LA was

not limited in nutritional value.

After the development of isolation techniques, minor proteins in whey could be

separated to exhibit their biological potential. LF showed both bacteriostatic and

bactericidal activity against pathogenic microorganisms, including those responsible for

gastroenteric infections, food poisoning, listeriosis, and mastitis, hence attracted high

attention in the application as natural antibiotics and ingredient in infant formulas

(Dionysius et al., 1993). GMP was probably the least known minor protein in whey,

because of no aromatic amino acids, a net negative charge, and the low molecular weight

with only 64 amino acid residuals (Brody, 2000). However, more and more research

focused on GMP has found its promising bioactivities, which made it a prospective

ingredient in functional food and even medicinal application (Martín-Diana et al. 2006;

Nakano et al.2006; Thoma-Worringeretal et al., 2006). The biological activities related to

GMP included binding of cholera and Escherichia coli enterotoxins, inhibition of

bacterial and viral adhesion, suppression of gastric secretions, promotion of

bifidobacterial growth, and modulation of immune system responses (Brody, 2000).

1.1.2 Whey protein processing and production

Liquid whey obtained from the cheese-manufacturing vat was treated by further

processing steps to fulfill the requirement of customers. Liquid whey was traditionally

processed into whey powder and drying was always the final step in manufacturing whey

proteins. In the 1920s four different methods were involved in whey protein processing:

conventional hot roller milk driers; heating until a concentrated liquid was obtained,

5
cooling to solidification, and then extruding in a tunnel; two-stage steam heating; and a

combination of spray drying and rotary drum drying (Gillies 1974). Nowadays there had

been various advanced techniques developed by research beyond the traditional

processing methods, which generated more and more whey protein products with

different functional and biological properties. Each processing step (technique) and the

control of processing conditions (temperature, pH, and others) would alter the protein

properties and enhance the variability of available whey protein products.

Before further processing steps such as concentration, crystallization,

demineralization, and drying, the pretreatments were normally required to prepare the

liquid whey into good condition with less impurities left from cheese manufacturing.

After collected from the cheese vat, whey liquid normally contained small amount of

cheese fines and lipids, which required centrifugation (clarification) as one pretreatment

to remove the impurities (USDEC, 2004); Whey liquid also had starter culture

microorganisms left from cheese manufacturing, which was not expected to be present in

whey product and required pasteurization to inactivate them (USDEC, 2004). High

temperature and short time (72°C, 15sec) pasteurization was used to heat the whey liquid

to inactivate the starter culture microorganisms and kill possible pathogens (USDEC,

2004). Other pretreatments such as shifting pH, addition of phosphates, and addition of

CaCl2 or other ions were also helpful in removing phospholipoprotein complexes (PLPC)

and colloidal calcium phosphate to improve the performance of following processing

steps (Morr, 1993). After these pretreatments, the whey liquid was ready for further

processing.

6
Although Whey proteins had a high nutritional value, mostly because of the high

content of essential amino acids, especially sulfur-containing ones, whey powders dried

directly from fresh whey liquid was not a value-added product, due to the high lactose

and ash contents (Wit, 1998; USDEC, 2004). Many techniques have been developed to

selectively concentrate and isolate whey proteins and membrane technology and ion

exchange had been recognized as value-added process.

There have been four classes of membranes used to concentrate whey: reverse

osmosis (RO), nanofiltration (NF), ultrafiltration (UF) and microfiltration (MF) (USDEC,

2004). Basically these techniques involved passing whey liquid across a semi-permeable

membrane, while a pressure gradient was created by a combination of pumps and valves

(USDEC, 2004). When whey liquid flowed by the membrane with designed pore size, the

smaller molecules would permeate the membrane, and so called permeate; the larger

molecules and particles would not be able to cross the membrane and hence be

concentrated on the other side of the membrane, which were referred to as retentate

(USDEC, 2004).

As one of the most popular used membrane technique, UF can produce various

whey protein products with different nutrition ratio of the retentate and permeate, based

on the retention characteristics of the membranes used (Atra et al., 2005). In the retentate

part, particles with bigger molecular weight (size) such as protein and fat fraction were

kept and concentrated, while particles with smaller molecular weight (size), such as

lactose, minerals and vitamins, were penetrated and separated as permeate (Hinrichs,

2001; Kessler, 2002). UF was able to separate and concentrate substances with a

7
molecular weight between103 and 106 Da, which were much bigger than NF’s molecular

weight capacity, ranging from 100 and 500 Da (Atra et al., 2005). With this small pores

sizes in membrane, NF membranes could only allow the penetration of some monovalent

ions and water, hence it could be used to concentrate of the permeate produced during UF

of whey, which mostly consisted of lactose in the same concentration as in the water

phase of the original fluid (Atra et al., 2005). MF had the largest membrane pores sizes

among all four classes mentioned above, and could permeate smaller soluble proteins,

peptides, lactose, minerals, non-protein nitrogen components, and water easily (USDEC,

2004). The retained part was mainly fat globula, and it was considered extremely

effective for removing the PLPC at all pH and temperature conditions. Hence it was

either considered as an alternative process to pretreatment to remove PLPC, or as an

amendment method to remove a small amount of fat that was not recovered through

centrifugation during pretreatment (Morr, 1993; USDEC, 2004). RO had the smallest

membrane pores sizes and only water can permeate through the membrane, hence it could

filter the whey liquid in the way that the concentrated solids contained the same ratios as

before (USDEC, 2004).

There had been two major ion exchange processes available, including the Vistec

process and the Spherosil process (Morr, 1989). They were different in their basic

operation: the former one used a stirred-bed ion exchange reactor column, whereas the

Latter one used a fixed-bed ion exchanger (Morr, 1989). Since the stirred-bed process is

much superior, it had been used as a more popular method to produce commercial Whey

protein products (Morr, 1989). In processing, liquid whey was firstly adjusted to acidic

8
pH to obtain the positive charged protein molecules, and then it was pumped into a tank

(ion exchange tower) that contained negatively charged resin beads. The positively

charged protein was retained in the negatively charged resin, and other components such

as fat, lactose, and mineral were eluted out of the tower and separated from the proteins

(Huffman, 1996). Once the resin was fully loaded by proteins, a new mobile phase was

added to adjust the pH of the tank to alkaline and detach the protein from the resin for

following processes (Huffman, 1996).

Diafiltration (DF) was used to further process the retentate after UF of whey

liquid, and produce whey protein products with a higher protein content. With the

addition of DF volumes (normally wash-water), the retentate macromolecules were

further purified by eliminating more lactose and minerals (Román et al., 2012).

Crystallization was used to process permeate produced from membrane filtration

step, hence it was considered as a further processing or economical disposal of the

remaining lactose stream (Jelen, 1979). The lactose in permeate was supersaturated to

crystallize and high quality lactose could be utilized in the sweetener market (Jelen,

1979).

Spray drying was normally the last step in processing whey protein powders.

Typical drying operations consisted of evaporation in multistage vacuum evaporators,

followed by spray drying (Jelen, 2011). As one of the mostly used drying techniques,

spray drying was applied to not only whey powder, but also instant coffee, milk powder,

tea and soups, as well as healthcare and pharmaceutical products, due to the advantages

such as continuous operation, suitable to heat sensitive materials, short processing time,

9
and automatic control (Chegini and Taheri, 2013). The quality of whey powder processed

through spray drying was affected by operating conditions and process variables: the

method and conditions of atomization, the type of spray/air contact, drying air

temperature and feeding concentration, feeding temperature and the degree of feed

aeration were all important parameters (Chegini and Taheri, 2013). A typical industrial

spray drying process is showed in figure 1.3.

Figure 1.3. A typical industrial spray drying process (Chegini and Taheri, 2013).

10
With various processing techniques, specific whey protein products could be

provided based on the requirement of customers. There have been large amount of whey

products available in the U.S. market as shown in Table 1.2.

2004 2005 2006


Reduced lactose and mineral whey 38507 44620 41547
Whey protein concentrate 25-49.9 123266 125363 134928
Whey protein concentrate 50-89.9 38147 48783 59083
Whey protein isolate 12544 12517 13913
Lactose 301921 323854 335049
Table 1.2. Production (in tons) of whey protein products in the United States (USDA
National Agriculture Statistics Services)

The United States Code of Federal Regulations Title 21 defined Whey protein

concentrate (WPC) as:

“Whey protein concentrate is the substance obtained by the removal of sufficient

nonprotein constituents from whey so that the finished dry product contains not less than

25 percent protein. Whey protein concentrate is produced by physical separation

techniques such as precipitation, filtration, or dialysis. As with whey, whey protein

concentrate can be used as a fluid, concentrate, or dry product form. The acidity of whey

protein concentrate may be adjusted by the addition of safe and suitable pH-adjusting

ingredients.”

WPC in the market was mostly available as 35% or 85% protein contents, and

85% WPC was used more as human food ingredients (Foegeding and Luck, 2011). The

11
processing steps were listed in figure 1.4.

Figure 1.4. Processing of whey protein concentrate (USDEC, 2004).

Whey protein isolate (WPI) contained larger than 90% of protein and 4–6%

water, the remaining 4–6% of the ingredient was a combination of fat, lactose, and ash

(Foegeding and Luck, 2011). The processing steps were listed in Figure 1.5.

12
Figure 1.5. Processing of whey protein isolate (USDEC, 2004).

Whey protein hydrolysate (WPH) was obtained by hydrolyzation of WPC or WPI

using enzymes, acids or alkali reagent, cleaving the peptide bonds in whey protein

molecules producing peptides with different sizes and free amino acids, which could

reduce allergenicity, achieve specific dietary requirements, or to improve functional

properties (Adler-Nissen, 1986; Farrell et al., 1987; Asselin et al., 1988; de Freitas et al.,

1993; Cordle, 1994; Frokjaer, 1994; Nielsen, 1997; Clemente, 2000; Sinha et al., 2007;).

Compared to acid and alkaline hydrolysis, which were hard to control and tend to reduce

13
the nutritional qualities of WPH, enzymatic hydrolysis (either acid or neutral) has

become a better choice to gain biologically active components for food application (Sinha

et al., 2007). The degree of hydrolysis and the molecular weight distribution of the

constituent peptides were normally used to characterize WPH, which could affect the

functional and sensory properties (van der Ven et al., 2002).

1.1.3 Whey protein functionality

The functionality has been defined as “any property of a food or food ingredient,

except its nutritional ones, that affects its utilization (Pout-El, 1981).” When tal ing

about the functionality of whey protein, water binding, gelation, emulsification, and

foaming were those frequently required in food production and attracted more attention

from the researchers.

Protein-water interactions in food system were always the first considered

parameter, because of its wide influences in solubility, gelation, emulsifying, foaming

capacity and other functionalities. Kuntz and Kauzman (1974) defined protein-water

interaction at increasing water activities based on changes of water molecules associated

with proteins: “The initial molecules of water that associate with, sorb, or bind to a dry

protein display altered physical properties, e.g., reduced vapor pressure, enthalpy,

entropy, volume, chemical potential; decreased translational/rotational frequencies; and

increased specific gravity, heat of vaporization, and heat capacity; The bound water was

not freezable at normal temperatures or even as low as -40°C. As the equilibrium relative

humidity (ERH) is increased, additional water molecules progressively associate with

protein in less structured layers and as capillary-held water. These water molecules show

14
a gradual continuum in physical properties from the highly structured initial layers to

those typical of normal bulk liquid water” (Figure 1.6).

Dry protein

Sorption at Aw 0.05-0.4

Protein covered by monolayer of


adsorbed water molecules
20-30 g H 20/100 g protein

H2O at Aw 0.5-0.75

Protein with multilayers of adsorbed water


30-40g H2O/100g protein Swelling of protein “glassy” state

H2O at Aw 0.98

Hydrated protein with multilayer and capillary condensed water Swollen, solvated,
40-50g H2O/100g protein Partly solubilized protein

H2O at Aw <0.99

Swelling
Solvation
Dispersion
Dissolution

Figure 1.6. Generalized sequence of proteins-water interaction at increasing ERH. The


actual water content(s) at each Aw varies with the nature and types of protein and the
experimental conditions (Kinsella et al., 1986).

The factors affecting water binding of protein involved both interior nature of

proteins themselves and exterior conditions of the environments. As stated by Kinsella

15
and others (1986), amino acid composition, protein conformation, surface

polarity/hydrophobicity, ionic concentration, ion species, pH, and temperature were

important factors in the water binding of proteins. Amino acids with electrical charged

side groups or polar side groups tended to associate with more water molecules by either

ion-dipole or dipole-dipole interactions, on the other hand, nonpolar amino acids tended

to be involved in the minimal water interaction. Similarly, when more polar groups were

exposed to the aqueous phase, which was possibly based on the protein conformational

structure and surface area, the water binding would increase accordingly (Mangino,

1984). Salts exhibited different abilities when introduced to protein-water interaction,

which was recognized as salting in and salting out: less amount of salts could interact

with both charged groups on the protein surface and water molecules, and hence increase

the water-binding ability, whereas high amount of salts would compete with protein for

water molecules and decrease the water-binding ability (Mangino, 1984). pH was able to

affect the net charge and conformation of protein: when pH was away from the isoelectric

point, the solubility tended to increase due to enhanced water-binding ability of ionized

amino acids side groups, and when pH approached the isoelectric point, the protein was

prone to precipitate (Kuntz, 1971). Normally water sorption decreased while the

temperature increased and hydration around polar groups decreased too (Noguchi, 1981).

In general, whey protein was classified as heat induced gelation, and formed a

continuous network of macroscopic dimensions immersed in a liquid medium exhibiting

no steady-state flow. The gelation process was suggested to follow a two-stage process:

The initial reactions in the gelation process involved the denaturation of a protein

16
molecule by weakening and breaking of hydrogen and disulfide bonds in secondary and

tertiary structure to disrupt native protein conformational structures. The second reaction

involved polymerization or aggregation of the denatured protein molecules to produce a

three-dimensional protein structure and immobilized a large amount of water molecules

to form a gel (Schmidt, 1981; Brandenburg et al., 1992; Mangino, 1992). A proper

balance between repulsive and attractive protein-protein and protein-water interaction

was required in forming and maintaining the protein gel structure (Schmidt, 1981).

Properties of protein gels were also affected by intrinsic and extrinsic factors, which

included but not limited to the composition and concentration of the proteins, heating

temperature, pH, ionic strength , calcium concentration, and free sulfhydryl

concentration (Schmidt, 1981; Damodaran, 1989; Foegeding, 1989). A lot of researchers

have illustrated the significant influences of minerals on the characteristics of whey

protein gels, and they also have proved that salt levels and salt types had various impacts.

Schmidt and others (et al., 1978,1979) found that dialyzed WPC obtained the maximum

gel hardness values when protein suspensions contained 200mM NaCl or 11.1mM CaCl2,

and hardness decrease with increasing salt concentrations. Mulvihilland and Kinsella

(1988) confirmed these observations with similar results and further suggested that CaCl2

appeared to be far more effective than NaCl because much lower concentration of CaCl2

was required to produce similar increases in hardness. Kuhn and Foegeding (1991)

conducted the experiment to investigate the influence of monovalent and divalent salts of

on the rheological properties of WPI gels. They found that the charges of salts had

obvious impact on the shear strain values, and WPI gels formed with addition of

17
monovalent salt showed significant difference from WPI gels formed with divalent salts

(Kuhn and Foegeding, 1991). Protein concentration was also an important factor in heat

induced gel formation and the properties of the gel structure. A critical level of protein

concentration was required to establish the gel formation that should be high enough to

facilitate the protein-protein interaction, and this critical level might be different based on

the protein used. While the concentration was increased higher than the critical level, a

firmer gel might be formed (Schmidt, 1981). The temperature and pH value had great

effects on protein denaturation, hence were able to affect gel formation in various ways.

Emulsions are heterogeneous systems consisting of two immiscible liquids that

were noticed as dispersed phase and continuous phase, and the emulsifiers functioned to

reduce the interfacial energy and facilitate the dispersion of the dispersed phase (Kinsella

and Whitehead, 1989). When a protein was used as an emulsifier, both nutritional reasons

and functional nature of the proteins made them good candidates in food applications.

Proteins are macromolecules containing both hydrophilic and hydrophobic groups. To act

as an emulsifier, protein needed to firstly approach the interface of two phases and

partially unfold to expose the hydrophobic groups, which could happen as random

fluctuation at either solvent phase or water-oil interface (Mangino, 1984). To lower the

free energy in the two phases system, the protein had the chance to insert its exposed

hydrophobic groups into the oil phase and leave the hydrophilic groups inside the water

phase. The rigidity of the protein structure and the number and location of hydrophobic

groups were important factors that could affect the unfolding and hence the emulsifying

18
ability (Mangino, 1984). In addition, pH, salts, protein concentration, and temperature

could also influence the insertion of whey protein into oil phase (Yamauchi et al., 1980).

Singh and Dalgleish (1998) conducted experiments to investigate the emulsifying

properties of WPH with degrees of hydrolysis (DH) varying from 8 to 45%, and to

determine the minimal length of peptides that could be used to produce a stable oil-in-

water emulsion system. They observed that a minimum length of peptide, which was the

product of hydrolyzed total whey protein, was required for effective stabilization of

emulsions, and the mean length of the effective peptides was about 5 amino acids (Singh

and Dalgleish, 1998). They also suggested that the enzymes used to hydrolyze total whey

protein played important roles in the functional behavior of WPH (Singh and Dalgleish,

1998).

After establishment of the emulsion, the two phases system was

thermodynamically unstable, and would finally collapse. Interfacial tension between the

two phases, characteristics of the adsorbed film in the interface, magnitude of the

electrical charge on the dispersed globules, size and surface-volume ratio of the dispersed

globules, and viscosity of the dispersing phase were all factors influencing the stabilizing

of emulsion (Morr and Ha, 1993).

Food foams were most commonly formed by mechanically dispersing air into a

solution, with foaming agents such as protein, emulsifier, and polysaccharide food

stabilizer gum involved in the system to lower the high surface tension and high surface

energy (Mangino, 1984; Morr and Ha, 1993). Foam formed and stabilized by protein

shared quite similar mechanism as emulsion system involving protein molecules. The

19
protein molecules needed to firstly diffuse to the interface of air and solution, unfold

themselves and surround the nonpolar air phase to lower down the interfacial tension

(Mangino, 1984).

The stability of the foam was dependent upon factors such as total solids, protein

and carbohydrate concentration, pH, the concentration of calcium and other ions, and

whipping method (USDEC, 2004). Phillips and others (1991) reported that addition of

salts (Na2SO4, NaCl, and NaSCN) with low concentrations (0.1M) reduced maximum

foam expansion values of WPI solutions, however addition of Na2SO4 with high

concentration (1M) increased the maximum foam expansion compared to the control

WPC solution. More research revealed the effect of thermal treatment on the physical

properties of whey protein foams. Temperature of heat treatment and pH were claimed to

have major influence on the protein denaturation and protein solubility, hence were able

to impact foam stabilized by protein molecules. Nicorescu and others (2009)

demonstrated that foam overrun of a commercial WPI solution slightly decreased while

the temperature of thermal treatment upon protein was increased, they also found that the

temperature of thermal treatment at 80°C was best for stability against drainage, and

increasing the temperature above 80°C did not improve the functional properties of the

proteins. Richert and others (1974) also found that severe heat treatments generally

impaired the foaming characteristics of WPC dispersions. The optimum pH value to

stabilize foam was detected to be different based on the whey protein used: maximum

stability of foam involving WPC solution was reported at approximately pH 7 (Richert et

20
al., 1974), while the maximum foam stability for unheated WPI was obtained at pH 5.0

(Phillips, 1990).

1.1.4 Whey protein application

As mentioned in section 1.2, nowadays there is abundance of commercially

available whey protein products: WPC, WPI, and WPH were the major supplies, while

reduced lactose and mineral whey, whey powder, whey permeate, demineralized whey

and whey protein fraction were also significant components in food industry. These whey

derivatives differ not only in the amount of protein content, but also in the molecule

interaction that is related directly to the control of processing methods and parameters

(Jelen, 2011; Foegeding and Luck, 2011).

Hence whey protein products, especial commercial available products, generate

diverse and specific functions when introduced to the food application. The protein

content in WPC can range widely from 25% to 89%, the lower level (35%) and higher

level (80%) were normally used in different food products due to the nutritional value,

price, and functionality. WPC with 35% protein content could be used in bakery products

to form a heat and shear-induced produc; another usage was to put in stews and sauces

because of their thickening properties; possible application also included luncheon meats

and meat patties (Morr and Ha, 1993; Foegeding and Luck, 2011). All of these

applications are the outcomes of combining protein, lactose, and minerals (Morr and Ha,

1993; Foegeding and Luck, 2011). WPC with 80% protein content was normally

designed for gelation, emulsification, and foam formation functions in food products

since its higher protein composition, for example the high gel strength and water-binding

21
properties made WPC with 80% protein content a nice candidate for meat products

(Foegeding and Luck, 2011). WPI has even higher level of protein composition with a

minimum of 90%, which is obtained from the additional processing steps (either ion

exchange chromatography or microfiltration) (Tunick, 2008). By removing the impurities

from WPI by ionic charge, this ingredient gains excellent water binding, gelling,

emulsifying, and foaming abilities and is broadly used in nutritional supplements, sports

and health drinks, and protein-fortified beverages (Foegeding and Luck, 2011). In

contrast to WPC and WPI as the intact protein, WPH is produced by enzymatic

hydrolysis and hence contains peptides that are going to enhance the absorbance and

lower the allergenicity (Foegeding and Luck, 2011). In addition to these properties, WPH

also performs nicely in improving heat stability, producing bioactive peptides, tailoring

amounts and size of peptides for special diets, and altering the functional properties of

gelation, foaming and emulsification (Nielsen, 1997; Foegeding et al., 2002).

Consequently it is easy to find WPH as an important ingredient in infant formula and

enhanced-performance products (Sinha, 2007; Foegeding and Luck, 2011). Actually, the

differences of functionalities are not only seen among WPC, WPI and WPH, different

commercial whey protein products generated distinct properties and functionalities,

because variation present during whey protein manufacturing and processing from

original products (milk) will result in diverse composition and functional properties.

1.1.5 Whey protein analytical methods

The analysis of whey protein concentrates was described in Code of Federal

Regulations Title 21, which included the analytical methods for all major contents

22
present in whey protein powders (Table 1.3). In addition, various analytical methods to

determine the whey protein composition were developed, because the composition was

more important in the prediction of whey protein nutritional value, functionalities and

biological activities. Researchers had explored electrophoresis and liquid

chromatography as the analytical methods to fractionate whey proteins, and High

performance liquid chromatography (HPLC) has became the most popular method due to

its versatility, short analysis time, high resolution, and superior column performance

(Elgar et al., 2000). Electrophoresis had been traditionally adopted in the research of

proteins: native polyacrylamide gel electrophoresis (PAGE) separated the proteins by

their different charges and sizes, whereas sodium dodecyl sulphate polyacrylamide gel

electrophoresis (SDS-PAGE) worked by molecular mass differences only (Kinghorn et

al., 1995). Kinghorn and coworkers (1995) measured the major proteins in both acid

whey and WPC samples by native PAGE and SDS-PAGE, and compared the separation

achieved by both methods. They demonstrated that native PAGE could be used to

determine levels of α-Lac, β-Lg A, β-Lg B and BSA, wheress IgG proteins can not be

quantitated. SDS-PAGE could be used to quantitate BSA and IgG (using the heavy chain)

while the presence of glyco-α-Lac might interfere with the determination of α-Lac and β-

Lg levels. The accuracy of both PAGE methods was dependent on staining/de-staining

protocols and on reproducible densitometry (Kinghorn et al., 1995).

23
Criteria Analytical method Section Heading
Protein content >25% Total Nitrogen-- 16.036 Total Solids
Officials (Liquid sample)
Final Action

Kjeldahl Method 16.193 Protein--


(Dry sample), Official Final
Action
Fat content 1-10% Reese-Gottlieb Method 16.059 Fat
[Reference Method] (Liquid sample)
(11)--Official Final
Action
16.199 Fat in Dried
(Dry sample) Milk(45)--
Official Final
Action
Ash content 2-15% Ash (5)--Official Final 16.035 Total Solids
Action (Liquid sample)
Ash--Official Final 16.196 Dried Milk,
Action (Dry sample) Nonfat Dry
Milk, and
Malted Milk
Lactose content ≤6 % Gravimetric Method-- 16.057 Lactose
Official Final Action (Liquid sample)
Lane-Eynon General 31.061 Lactose--
Volumetric Method (Dry sample) Chemical
Methods--
Official Final
Action
Moisture 1-6% Moisture (41)--Official 16.192 Dried Milk,
content Final Action Nonfat Dry
Milk, and
Malted Milk
Solids content Method I--Official 16.032 Total Solids
Final Action
Titratable Acidity (2)--Official 16.023 Milk
Acidity Final Action
Equivalent
potentiometric method
Table 1.3. Analytical methods for whey protein contents regulated by Code of Federal
Regulations Title 21.

24
Chromatography technique was utilized as the main method in analytical and

preparative separation of whey proteins. Ion exchange chromatography, reversed phase

chromatography, size-exclusion chromatography, hydrophobic chromatography, and

affinity chromatography have been explored in the separation of major or minor whey

proteins and each method had its own merits and sensitivity. For instance, Elgar and

coworkers (2000) developed a reversed phase-HPLC method to simultaneously separate

and quantify the α-lactalbumin, β-lactoglobulin, bovine serum albumin, proteose peptone,

caseinomacropeptide, and immunoglobulin G in bovine whey samples including WPC

and WPI. The α-lactalbumin detected by reversed phase-HPLC showed similar results as

literature value and their reference value using sodium dodecyl polyacrylamide gel

electrophoresis (SDS–PAGE), whereas β-lactoglobulin showed lower level compared to

SDS-PAGE due to the solubility interference; Bovine serum albumin and

immunoglobulin G detected using reversed phase-HPLC also correlated well with their

reference value measured by SDS-PAGE, and proteose peptone and caseinomacropeptide

were within reasonable range of literature values or reference values separately (Elgar et

al., 2000). Doultani and coworkers (2003) conducted the ion exchange chromatography

to recover the major proteins in mozzarella cheese whey. Both cation exchange

chromatography packed with SP Sepharose Big Beads and anion exchange

chromatography packed with Q Sepharose Big Beads were used in the recover process,

because glycomacropeptide could not bind to the S-beads and pass through into the

effluent solution of cation exchange chromatography. The results turned out to be that

92% of the major whey proteins (α-lactalbumin, β-lactoglobulin, bovine serum albumin

25
and immunoglobulin G) and 96% of the glycomacropeptide were successfully recovered

with high throughput (Doultani et al., 2003). Hence it needed to be noticed that the

increasing robustness of HPLC with the development of efficient columns, buffers,

solvents, sensitive detector and higher instrument automation made it a more

comprehensive and reliable analytical method in the analysis of whey protein

compositions.

1.2 Astringency and mechanisms in various food

1.2.1 Terminology of astringency

Astringency is a complex phenomenon and it incorporates multiple sensations

involving the drying and roughness of oral surfaces with the additional perception of

tightening, or puckering of the oral mucosa and muscles around the mouth (Lee and

Lawless, 1991). There have been a lot of attempts to describe this complex terminology

and argument about sub-qualities such as pucker, sourness and bitterness were under

discussion for a long while (Lee and Lawless, 1991; Green, 1993). The American

Society for Testing and aterials (AST ) defined astringency as “the complex of

sensations due to shrinking, drawing or puckering of the epithelium as a result of

exposure to substances such as alums or tannins” (ASTM, 2004). Gawel et al. (2000)

identified a range of oral sensations from red wines by a descriptive panel and defined

astringency. The related terms were shown in figure 1.7.

26
Figure 1.7. A mouth-feel wheel to classify the sensation elicited by red wine (Gawel et
al., 2000).

1.2.2 Mechanisms of astringency in four main groups of astringent compounds

Astringency was identified in an abundance of food and beverages such as wine,

tea, cranberry juice, unripen fruits, organic acids, and others, which could be classified

into four “true” groups of astringent compounds: salts of multivalent metallic cations

(particularly aluminum salts), dehydrating agents (ethanol and acetone), mineral and

organic acids, and vegetable tannins (Joslyn and Goldstein, 1964).

The mechanisms of astringency caused by vegetable tannins have been well

studied and was thought to result from the interaction of tannins with specific salivary

proteins to form protein-tannin complexes, which would diminish saliva’s ability to

27
lubricate the mouth by subsequent aggregation and precipitation (Luck et al, 1994; Baxter

et al., 1997; Dinnella et al., 2009).

Tannins, commonly referred to as polyphenols, were categorized as condensed

tannins or hydrolyzable tannins as shown in figure 1.8 (Haslam et al., 1988; Bennick,

2002). Tannins in plants were recognized to have biological effects or defense effects,

which could protect plant leaves and unripe fruits from the browsing by their astringency

properties (Bennick, 2002). Traditionally, polyphenols triggering astringency have been

defined with molecular weights between 500 and 3000 Da (Bakker, 1998; Lesschaeve

and Noble, 2005), and it was generally accepted that the degree of polymerization and

molecular weight of polyohenol was positively related to its precipitating ability and

perceived intensity (Bate-Smith, 1973; Arnold et al., 1980; Peleg et al., 1999).

Figure 1.8. The example of hydrolyzable tannins (A) and condensed tannins (B) showing
their molecular structures.
28
It was generally in agreement that salivary proteins were comprised of proline-

rich proteins (PRPs), histidine-rich proteins (histatins or HRPs), α-amylase, and

mucinglycoproteins (de Freitas and Mateus, 2001; Charlton et al., 2002; Condelli et al,

2006; Bajec and Pickering, 2008), and PRPs and HRPs counted for the major proteins

responsible for tannins binding. The extended and unfolded structure of PRPs and HRPs

provided plenty of binding sites that were easily accessed by tannins (Charlton et al.,

1996; de Freitas and Mateus, 1 bstl et al., Dangles and Dufour, 6 Croft

and Foley, 2008; Canon et al, 2009).

Charlton et al. (2002) proposed a 3-stage model describing the binding and

precipitation of PRPs by polyphenols, which was confirmed and further modified by

bstl and cowor ers ( ) (Figure. 1.9). In the first stage, polyphenols interacted with

salivary proteins and bound to multiple sites, causing the previously randomly coiled

protein to coil around the polyphenol and compact. In this stage, hydrophobic interactions

and hydrogen bonding were both involved in the interaction between polyphenols and

PRPs (McRae and Kennedy, 2011). In the second stage, polyphenol bridges were formed

via cross-linking of protein-phenol complexes and polyphenol-PRPs dimers were yielded.

In the last stage, the dimers aggregate to form large complexes and precipitate (Charlton

et al., bstl et al., ). In the last two stages, hydrogen bonding was shown to be

involved (Hagerman et al., 1998; Simon et al., 2003; Cala et al., 2010). Potential

interactions between polyphenols and PRPs could include covalent bonds, ionic bonds,

hydrophobic interaction, or hydrogen bonding. However there was little evidence that

29
prove the presence of covalent or ionic bonds towards these interactions (Haslam et al.,

1991; Hagerman and Butler, 1978).

Figure 1.9. Proposed mechanism for polyphenol-PRPs complex formation ( bstl et al.,
2004).

Factors modulated astringency caused by vegetable tannins have been reported

separately for individual food or beverage. Beverage pH, viscosity, temperature,

sweetener and ethanol concentration were mostly studied and showed diverse relationship

toward astringency. It is generally understood that pH has great impact on the tannins

astringency. Kallithraka and coworkers (1997) found that lowering the pH of model

solutions and red wines resulted in the increase of the astringency intensity. Fontoin and

others (2008) showed similar relationship in model wine solutions. Experiments were

also conducted to reveal the relationship between pH and astringency in cranberry juice,

and results illustrated that juices with low pH were significantly higher in astringency,

regardless of sample temperature or viscosity. High viscosity might contribute to reduced

astringency, and possible mechanisms were suggested to be: (1) the lubricating ability

30
caused by thickening agent such as carboxymethylcellulose (Peleg and Noble, 1999); (2)

the complex formed by hydrocolloid gums and tannins which could reduce tannins

binding efficiency with proteins (Taira, Ono & Matsumoto, 1997). In addition, the

increased viscosity of the model wine may explain the association between high sucrose

concentrations and low astringency (Smith et al., 1996). Surprisingly the addition of

sweetener (aspartame) in grape seed tannins had no impact on the astringency parameter

(Smith et al., 1996). Another important modulating factor normally discussed in red wine

was proposed to be the concentration of ethanol, and higher concentrations were related

to lower intensity of perceived astringency (Vidal et al., 2004; Fontoin et al., 2008). The

recommended mechanisms contained: (1) the conformational changes of tannins in

higher ethanol wines, which could reduce the binding abilities of tannins, the self-

association of bound tannins, and the formation of protein aggregates (Fontoin et al.,

2008). (2) greater ethanol concentrations may increase the lubricity of the oral cavity and

reduce the perception of roughness (Demiglio and Pickering, 2008; Fontoin et al., 2008).

Although the mechanisms caused by salts of multivalent metallic cations were not

fully understood, aluminum sulfate, aluminum ammonium sulfate and aluminum

potassium sulfate have been defined to be astringent agents (Jellinek, 1985; Lee and

Lawless,1991; Breslin et al, 1993). Besides, researchers also stated that zinc was able to

induce astringency as a primary sensation (Keast, 2003; Lim and Lawless, 2005), while

iron, copper, and the minerals magnesium and calcium could induce astringency as the

secondary sensation (Lim and Lawless, 2005; Lawless et al., 2004). Current

understanding about the mechanisms of astringency triggered by metal cations focused

31
on the binding ability of aluminum toward some proteins and the strong binding ability

toward water (Trapp, 1985; Haslam et al., 1988; Khalil-Manesh et al, 1989).

Contrary to the effect of pH on phenols, addition of acid suppressed the

astringency of alum cations. Lawless et al. (1994) reported alum-acid mixtures (alum-

gallic acid or alum-citric acid) showed less intensity in astringency than the unmixed

components, and aluminum cations may act as Lewis bases, binding with atoms with

unshared pairs of electrons, such as carboxylate ions. Other researchers speculated that

chelation of the aluminum cation upon acid reduced the binding ability of aluminum ions

with salivary proteins (Peleg et al., 1998). The viscosity was also indicated as one of the

parameters regulating astringency caused by alum cation. The astringency was detected

to be at lowest level while the viscosity of aluminmum sulfate was increased from

unthickened station to 5-7cP by methylcellulose (Smith and Noble, 1998). The authors

considered that the rising viscosity was able to restore lubrication lost by salivary

proteins and possibly lower diffusivity of the astringent stimuli (Smith and Noble, 1998).

Organic acids such as acetic, fumeric, quinic, adipic, lactic, malic, tartaric, and

citric acid (Martin and Pangborn, 1971; Hyde and Pangborn, 1978; Rubico and

McDaniel, 1992; Hartwig and McDaniel, 1995; Sowalsky and Noble, 1998), can induce

sensations of astringency. The astringency elicited by organic acid is solely a function of

pH, in a pattern of inverse relationship (Hartwig and McDaniel, 1995; Sowalsky and

Noble, 1998; Lawless et al., 1996). No influence from either specific anion or acid

concentration was observed in modulating astringency caused by organic acid (Sowalsky

and Noble, 1998).

32
Dehydrating agents such as ethanol were considered to be astringent, and were

generally used in cosmetic and medication products, however there is no direct evidence

proving the astringency as an oral sensation.

1.2.3 Exploration of mechanisms of astringency caused by whey protein

While whey protein ingredient was attracting more attention from the food

industry with its versatile functionalities and applications, the main problem, astringency,

generated as whey protein beverage formulated under acidic pH has emerged into the

view of researchers. The mechanisms and modulating factors related to astringency

caused by whey protein beverages were still unclear and under great debate.

Recently it was generally in consensus that whey protein itself induced the

astringency. Sano and coworkers (2005) conducted an experiment to compare the sodium

phosphate buffer solution at pH 3.5 as control with whey protein dissolved in the same

buffer solution, they detected no astringency from buffer but found astringency from the

latter solution. This experiment indicated that astringency in acidic whey protein

beverages was resulted from whey protein instead of acid, which was further confirmed

by the dependent relationship between astringency and whey protein concentrations

(Sano et al., 2005; Kelly et al., 2010). However opposite conclusion was brought by other

scientists stating that the astringency of acidic whey protein solutions appeared to be

caused by their high acidity and not directly by the whey proteins (Lee and Vickers,

2008). In this experiment, the total acidity of the solutions was controlled in both control

solutions and whey protein solution, and comparison between both samples did not

support the results gained by Sano and coworkers (Lee and Vickers, 2008).

33
Sano and coworkers also proposed the mechanisms of whey protein beverages.

When acidic WPI solution (pH 3.5) was mixed with salivary protein (pH 7) in the oral

cavity, the pH solution could reach a pH of about 5, which was approximately the

isoelectric point of whey protein. Whey protein would precipitate at their isoelectric

points in the mouth (Sano et al., 2005). Latterly Beecher and coworkers (2008) suggested

another mechanism from their evaluation of four groups of whey protein solutions at pH

6.8, 3.4, 3.0 and 2.6. Based on their observations (Table 1.4), they concluded that the

astringency of acidic whey protein solution could be attributed to the interactions

between the positively charged whey proteins (pH < pI) and negatively charged saliva

proteins.

The pH of whey Charge of whey Charge of salivary Potential for


protein solution proteins proteins aggregation
6.8 - - Less likely
3.4 + - Strongest
3.0/2.6 + + Decreased
Table 1.4. Charges of Whey proteins and salivary proteins in different solutions and their
potential of interaction.

Vardhanabhuti and coworkers (2010) further supported this mechanism by

examining the astringency caused by a commercial lactoferrin at pH 3.5, 4.5, and 7.0.

Lactoferrin had isoelectric point equal to 8.8; hence it remained positively charged at all

pH mentioned above. Their results showed that lactoferrin was astringent at all pH values

34
(3.5, 4.5, and 7.0), indicating that interactions between positively charged lactoferrin and

salivary proteins played a role in astringency (Vardhanabhuti et al., 2010).

1.3 The application of Infrared Spectroscopy (IR) on whey protein.

Infrared spectroscopy is based on the absorption of electromagnetic energy of

vibrations between atoms in a molecule and generation of spectrum information about the

chemical composition and conformational structure of the matrix (Willard et al., 1981).

Those vibrations between atoms are determined by the mass of the constituent atoms, the

shape of the molecule, the stiffness of the bonds, and the periods of the associated

vibrational coupling (Karoui et al., 2010). The mid-infrared spectroscopy (MIR)

(wavenumber range 400 to 4000 cm−1) became especially attractive since measurements

in this range provide direct information concerning the specific constituents in the

sample, as well as their characteristic molecular structure (Etzion et al., 2004). The MIR

characteristic absorption bands generated are generally related to the major components

in food matrix (Table 1.5). IR has been widely used in food analysis with advantages as

excellent signal-to-noise ratio, simultaneous detection of all frequencies, reduced scan

time without loss of resolution, high energy throughput, superior wavelength resolution,

and superior wavelength accuracy obtained through the use of an internal laser to

maintain wavelength calibration (Griffiths and Haseth, 1986). The development of

attenuated total reflectance (ATR) technology has made the application more successful

as it provided a simple and reproducible method to handle liquids and paste samples

(Kuehl and Crocombe, 1984). Research on whey protein products, whether solid or

35
liquid, have been facilitated with the aid of IR and covered the areas including

conformational analysis, quality authentication, and compositional quantitation.

Vibration description and Approximate


Food Component
Chemical bonds Frequency (cm-1)
OH-stretching 3300
Water
OH-bending 1640
Amide I (C=O stretching) 1653
Amide II (N-H deformation and
1567
Protein C-N stretching)
Amide III (CN stretching, NH
1 9−13 1
bending)
Acyl chain C-H 3000~2800
C=O group 1750
Fat
Triglyceride ester linkage C-O 1175
Trans double bonds 966
Aqueous sugar C-O and C-H stretching 1100~1000
Cellulose 1170-1150, 1050, 1030
Pectin 1680-1600, 1260, 955
Table 1.5. A general band assignment of MIR applied to food matrix (Finch and
Lippincott, 1956; Lin and Brown, 1992; Van de Voort et al., 1992; Libnau et al., 1994;
Safar et al., 1994; Dupuy et al., 1996; Michell and Schimleck , 1996; Mascarenhas et al.,
2000; Yang and Irudayaraj, 2000; Ripoche and Guillard , 2001; McClure and Standfield,
2002; Sørense et a., 2003; Kong and Yu, 2007; Hashimoto and Kameoka, 2008; Huang et
al., 2008).

There have been a lot of approaches toward the whey protein conformational

change induced by factors such as pH, heat, interactions with other molecules and more.

Researchers adopted IR as a major technique to assist the identification of the molecular

changes. Table 1.6 showed some of the research done by controlling various factors.

36
Protein Factors induced Conformational change Reference
conformational change
β-lactoglobulin Heating temperature, Intermolecular Allain et
pH, and high pressure hydrogen-bonding β- al., 1999
homogenization of sheet structure
gelling conditions (1614 cm−1)
β-lactoglobulin Heated at 90 °C for 30 Heat increased β-sheet Ngarize et
min and high-pressure- structures at the expense al., 2004
treated at 600 MPa for of α-helix and high-
20 min pressure increased
random coil;
Intermolecular β-sheet
(1625 and 1680 cm-1)
BSA Heat ( −9 °C) Formation of an Murayama
irreversible and
intermolecular β-sheet Tomida,
on heating above 70 °C; 2004
Short-segment chains
connecting α-helical
segments (1630 cm-1)
BSA NaCl, lactose, sucrose, Loss of α-helical Boye et
-1
glucose, cysteine, urea, structure (1654 cm ) and al., 1996
NEM, and SDS, and pH, the rise of the ordered
in gelling non-native β-sheet
structure
(1616 and 1684 cm-1)
α -lactalbumin and β Interaction with phenolic Shift of Amide I and II Zhang et
-lactoglobulin acids peaks; Changes of al., 2014
percentage content of
component bands;
Reduction in the amount
of protein α -helix
structure
Table 1.6. Recent researches on induced conformational changes of individual whey
protein.

One of the most important applications of IR related to whey protein is the

detection of milk adulteration. Milk can be adulterated by addition of water, neutralizers,

salt, sugar, high solid contents, whey, or hydrogen peroxide (Kartheek et al., 2011),
37
which could bring economic problems and health concerns. Whey protein could be added

to milk as an inexpensive protein source, which would be hard to detect due to structure

similarity. Mendenhall and Brown (1991) examined 135 non-fat dry milk samples

containing various concentrations of WPC (1.26 to 33.00%). A strong correlation (Y>

0.99) between the actual concentrations and the values predicted by the Fourier transform

infrared spectroscopy (FTIR) method was found through the utilization of partial lease

square regression (PLSR), and detection limit was identified as 5.6% whey protein

concentrate (Mendenhall and Brown, 1991). Santos and coworkers (2013) spiked the

control milk samples with known levels of whey ranging from 1.87 to 30 g/L. A

classification model built using MIR spectra and multivariable chemometrics to

discriminate control milk and adulterated milk samples was successfully built with large

interclass distances (7.7 between control and whey), and the detection limit was identified

as 7.5 g/L. The authors further quantified the adulterant levels of whey in milk using the

same MIR spectra and PLSR and obtained satisfactory results for both calibration and

prediction models (r = .96). They concluded “MIR could provide the dairy industry with

a simple, rapid and non-destructive technique for detection and quantification of milk

adulteration” (Santos et al., 13).

Since the value and functionality of whey protein products were largely affected

by their composition, the market required a convenient and accurate analytical method.

IR combined with statistical analysis has been found to be a potential method in the

prediction of the major compositions. A study aimed to determine the potential of near

infrared spectroscopy in estimating whey protein products’ compositions was carried out

38
based on 47 whey powders (Baer et al., 1983). The calibration and prediction models

were developed individually for moisture, protein, fat, and lactose to evaluate the

feasibility of IR in correlating the spectra information and reference values. Good

correlation was obtained with high R-value and low standard error of prediction.

Researchers also made their effort to predict whey protein fractions such as α-lactalbumin

and β-lactoglobulin using FTIR spectra (De Marchi et al., 2010; Rutten et al., 2011).

They proved it was possible to use FT-MIR as a routinely assessment in practice,

however these application need to be cautious because the results might not be accurate

enough when variation of genetic levels was involved (De Marchi et al., 2010; Rutten et

al., 2011).

39
Chapter 2: Authentication of Whey Protein Powders by Combining Infrared
Spectroscopy and Pattern Recognition Analysis

2.1 Abstract

Whey proteins are attractive ingredients to the food industry because of their high

nutritional value and wide functionality. Whey protein powders are available from

different suppliers using various processing methods and conditions, resulting in

variability in their component make-up, levels and protein structure, having an impact in

their functionality. There are three major types of whey protein including whey protein

isolates (WPI), whey protein concentrates (WPC) and whey protein hydrolysates (WPH),

which provided diverse functionality to food applications. Our objective was to develop a

simple and rapid method to differentiate whey protein types by combining a portable

infrared spectrometer with pattern recognition analysis. Whey protein powders including

WPI (n = 56), WPC (n = 32) and WPH (n = 23) were obtained from different suppliers

and their spectra collected using a portable infrared spectrometer by pressing the powder

onto an ATR-IR diamond crystal with a pressure clamp. Spectra were analyzed by soft

independent modeling of class analogy (SIMCA) generating a classification model

showing excellent 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 band centered at 1393 cm-1 was responsible for separation and can be

associated with carboxylic acid side chains of amino acids present in whey proteins.
40
Portable IR units was able to quickly assess the quality of the incoming raw material

allowing for timely corrective measures during manufacturing. Portable systems are

simple to use and require minimal or no sample preparation, thus reducing assay time and

helping to streamline the analytical procedure so that it is more applicable for field-based

screening and higher sample throughput.

2.2 Introduction

Whey proteins are valuable food ingredients providing functional properties

including gelation, high solubility, water-holding capacity, foaming/emulsification and

sensory characteristics enabling their use in numerous food applications (Lee et al., 1992;

Bouaouina et al., 2006; Dissanayake et al., 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 techniques and thermal conditions are the most significant

factors in the variability of whey protein products, not only affecting the amount of

protein and non-protein components but also modifying protein structure ultimately

imparting 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: whey protein concentrates (WPC) containing between 50 - 85% protein

on a dry basis, whey protein isolate (WPI) containing between 90-98% protein and very

small amounts of lactose and fat, reduced lactose whey, demineralized whey and

41
hydrolyzed whey (Huffman, 1996). 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

containing peptides that enhances digestibility, reduces allergenicity, improves heat

stability, produces bioactive peptides, and alters the functional properties of gelation,

foaming and emulsification becoming an important ingredient in infant formula and

enhanced-performance products (Foegeding et al., 2002; Foegeding and Luck, 2011).

While the production of whey powder has steadily increased (Foegeding and

Luck, 2011), the market demand far exceeds supply resulting in premium prices for whey

protein powders. Ingredient authentication of commercial whey products would limit

fraud but also help ensure manufacturers achieve the desired functionality in their

product. Ion exchange chromatography, electrophoresis, gel-permeation or size exclusion

chromatography, and HPLC techniques have been proven to be effective approaches

towards characterization of individual protein type within complex mixtures of whey

products (Kilara, 2008). In protein research, infrared (IR) spectroscopy has been applied

for the qualitative or quantitative determination of whey protein ingredients (Baer et al.,

1983; Marchi et al., 2009) and to study secondary structure of proteins (Curley et al.,

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 infrared spectra correlated to various functional properties of whey

and casein hydrolysates to speed up product development (van der Ven et al., 2002).

However there is still a need to establish a fast and efficient method to easily characterize

42
whey protein products that are not limited to a single type to benefit the dairy industry to

support quality control. The objective of this research was to explore the potential of

infrared spectroscopy in authentication of whey protein powders by characterizing

commercial whey protein products (WPI, WPC, and WPH) through pattern recognition

analysis.

2.3 Materials and methods

2.3.1 Whey protein sample

Different commercial whey protein powder products coded with 3 digit number,

including WPI (n=56), WPC (n=32) and WPH (n=23) were obtained from different

manufacturers. Sample information, including moisture, protein, ash, fat, sugar and pH,

was also provided for each powder from individual manufacturer. The pH of powder was

measured by dissolving 5g of each powder into 8oz of deionized water. In addition, a soy

protein powder and gelatin powder were included for validation of the model. Protein

levels and pH info were available.

2.3.2 Fourier-Transform Infrared Spectroscopy (FT-IR)

Spectra measurements were performed using bench-top (Excalibur FTS 3100GX;

Varian, Palo Alto, CA) and portable (Cary 630, Agilent Technologies, Santa Clara, CA)

spectrometers. The bench-top included a dynamically aligned Michelson interferometer,

potassium bromide beam splitter, and a deuterated triglycine sulfate (DTGS) detector.

Spectra were observed using commercial software (Win-IR Pro, Version 3.4.2; Varian

Inc.). IR spectra were collected from 4000 to 700 cm-1 with resolution of 8 cm-1, co-

43
adding 64 scans to increase signal-to-noise ratio. Measurements were performed on an

attenuated total reflectance (ATR) accessory, equipped with a diamond crystal, with

refractive index of 2.5 and an incidence angle of 45o, yielding three internal reflections.

The instrument was calibrated before measurements using the crystal alone (air) as

background. The powder samples were pressed onto the ATR-IR crystal using a high-

pressure clamp (PIKE Tech, Madison, WI). Spectral collection was done in duplicate for

each powder sample.

The portable spectrometer was equipped with a temperature-stabilized DTGS

detector, ZnSe beam splitter and spectrum was collected using a single-bounce diamond

ATR accessory. The powder samples were pressed onto diamond crystal using a pressure

clamp with a slip clutch press providing consistent pressure. The IR spectra were

collected using MicroLab software (Agilent Technologies, Santa Clara, CA) operating in

the wavenumber ranges from 4000 to 700 cm-1 with resolution of 4 cm-1, and 64 scans

were co-added to increase signal to noise ratio. Replicate spectra were also obtained for

each powder sample.

2.3.3 Multivariate classification analysis

The spectra collected were analyzed using the supervised pattern recognition

discriminant analysis: soft independent modeling of class analogy (SIMCA) (Wold,

1976). SIMCA is a classification algorithm based on principal component (PC) analysis

that is applied to each category of interest separately generating a “training set” to build

the classification model (Branden and Hubert, 2005). Spectral data were vector-length

normalized and transformed to their second derivative using a 25-point Savitzky-Golay

44
polynomial filter. A cross validation algorithm was used to determine the number of

principal components describing the systematic variation from spectral data (Wold 1978;

Wold et al. 1983), thus, avoiding over-parametrization or modeling of the noise (Bro et

al., 2008). Sample residuals and Mahalanobis distances were used to determine outliers

(Hruschka, 2001). SIMCA models were evaluated in terms of the matrix of scores and

loadings, discriminating power and interclass distances (Albano et al., 1978; Branden and

Hubert, 2005). Scores-plot projections of the original data onto principal components

axes were used to visualize sample clustering (patterns, groupings, or outliers) while the

loadings plot shows the impact of the variables (wavenumbers) on each vector. Interclass

distance is a critical parameter to determine the separation between classes, which is

measured as the distance by taking along a line connecting the their geometric centers

(Albano et al., 1978), and this is conducted by the comparison of the F-statistic with a

given confidence interval, which is normally 9 % ( eza- rquez et al., 1 ). When F

value is larger than the critical F value, which means the probability is smaller that

critical value (α), the interclass distance can be identified as significantly large and hence

the un nown spectrum does not belong to the nown class, either it is an outlier, or it

comes from a class not represented in the data set ( eza- rquez et al., 1 ). It is also

well known to use external validation samples sets for the evaluation of established

calibration model. The final step in modeling was testing the predictive ability of the

SI CA model by using “validation” dataset. SIMCA model enables positive

identification of new samples to the class only if it falls within the class borders, which

are 95% probability clouds defined around the samples of each class. If it falls outside the

45
class border, it is considered as an outlier (Infometrix, 2008). A chemometrics modeling

software, Pirouette (v4.0, Infometrix Inc., Woodville, WA), was used to perform the

multivariate classification. For analysis, spectra were imported into Pirouette as .dx files

and then analyzed by SIMCA.

2.4 Results and discussion

The spectrum of commercial whey protein powders was collected by attenuated

total reflectance using a benchtop (Figure 2.1A) and portable (Figure 2.1B) infrared

spectrometer, allowing optimization of the contact of the powders with the ATR crystal

surface and minimizing noise interference. Whey protein powders (WPI, WPC, and

WPH) showed fairly narrow bands arising from functional group vibrations in the

information-rich region of 1800 to 900 cm-1, with very similar spectral patterns obtained

by using the bench-top and portable FT-IR units (Figure 2.1). 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)), contributing to the peptide bond group

vibrations of proteins (Kong 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 associated with the broad peak band ~ 1250 cm-1

corresponding to the combined signal from the amide III groups and P=O stretching in

phosphodiesters (>PO2-) (Coates, 2000; Kong and Yu, 2007). The bands between 1200

46
and 900 cm-1 were associated with vibrations of C-O-C and C-O, attributed to sugars,

acids and/or polysaccharide components of the samples (Shiroma and Rodriguez-Saona,

2009).

CH groups of lipids
A

Amide II
Amide I
υC-O-C of

acyl chain of TAG


C-H stretching of

C=O groups of lipids


polysaccharides
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
CH groups of lipids
Amide II
Amide I

υC-O-C of
acyl chain of TAG
C-H stretching of

C=O groups of lipids

polysaccharides
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)

Figure 2.1. Mid-Infrared spectra of Whey protein powder from bench-top FT-IR
Spectroscopy (A) and portable FT-IR Spectroscopy (B) with Attenuated Total
Reflectance (ATR)
47
The spectral complexity required the application of mathematical filter functions

(normalization and second derivative), as well as, selection and combination of certain

spectral regions (1800-1000 cm-1) to enhance spectral resolution and model performance.

Prior to data analysis, each spectrum was vector-length normalized to minimize the

effects of potentially variable concentrations in the samples, and to reduce the systematic

error in the data (Zeaiter et al., 2005; Rinnan et al., 2009a). Second derivative

transformation of the spectra was performed to resolve spectral bands from the raw

spectrum (Zeaiter et al., 2005).

The SIMCA classification plots (Fig. 2.2) showed well separated clustering

among the powder samples, whose orientation in the 3D space correlated with whey

protein type (i.e. WPC, WPI and WPH). However, Figure 2.2 showed the existence of 7

clusters defined by pH and protein level of the powders associated with WPI (3 classes)

and WPH (3 classes) while all WPC samples were classified in a single cluster. A similar

clustering pattern was obtained using the bench-top (Figure 2.2A) and portable (Figure

2.2B) infrared units. High protein concentration (>65%) WPC are produced by

ultrafiltration and diafiltration while WPI and WPH are manufactured using various

technologies such as ion exchange, electrodialysis, nanofiltration, controlled enzymatic

hydrolyzation (Jelen, 2011), having a strong effect on the composition of the final

powder. van der Ven and others (2002) reported the ability of FTIR spectra to distinguish

protein hydrolysates prepared from different protein (casein or whey) sources and

proteolytic enzymes (acidic or neutral/alkaline).

48
A WPI (II) B WPI (III) WPI (II)
WPI (III)
PC2
PC2

WPC
WPI (I) WPI (I) WPC
PC3
PC1 PC1
PC3
WPH (I)
WPH (II) WPH (I)
WPH (II)
WPH (III)
WPH (III)

C Soy protein D WPI (III) Soy protein


WPI (II)
WPI (II)
WPI (III) PC2
PC2
WPI (I)
WPC

Gelatin WPC
PC3
WPI (I) Gelatin
PC1
PC1
PC3

WPH (II) WPH (I) WPH (I)


WPH (II)

Figure 2.2. Soft independent modeling of class analogy (SIMCA) classification plots for
calibration models using a benchtop (A) and portable (B) FTIR spectrometers. SIMCA
validation performance for calibration models using a benchtop (C) and portable (D)
FTIR spectrometers based on the infrared spectra of whey protein powders. WPI (I): pH
of protein powder ranging from 5.8-7.5, WPI (II): pH ranging from 2.8-3.5, WPI (III): pH
ranging from 3.8-4.0, WPH (I): protein levels were 80%-85% and pH ranging from 6.0-
9.0, WPH (II): 90% and pH ranging from 7.2-7.8, WPH (III): CGMP. pH was measured
by dissolving 5g of whey powder in 8 oz of deionized water.

49
SIMCA showed that WPI grouping was based on their powder pH level with

classes discriminated having pH ranges of 5.8 to 7.5 (WPI - I), 2.8 to 3.5 (WPI – II) and

3.8 to 4.0 (WPI – III), most likely associated with the type of technology used to

manufacture the WPI powder. A different situation was observed for WPH, with samples

clustering based on protein levels of 80% to 85% (WPH – I) and 90% (WPH – II). A

unique case was WPH - III samples that clustered furthest from the rest and were

associated with containing casein glycomacromolecule protein (CGMP).

Examination of the discriminating power plot provides information on spectral

frequencies that were most important for separating the different classes. Figure 2.3A

showed that the SIMCA model, generated using spectra collected using the benchtop

FTIR unit, had a major discriminating band 1390 and 1570 cm-1 corresponding to the

deprotonated carboxylate group for the symmetric and antisymmetric stretching vibration

(Barth, 2000). This band shifts upon cation chelation, and the effects depend upon the

mode of chelation and have been used to study several Ca2+ binding proteins (Fabian et

al., 1996; Mizuguchi et al., 1997). Thus, the band observed at 1339 cm-1 most likely can

be attributed to the carboxylate vibrations of Asp groups due to protonation change

(Gerwert et al., 1989).

50
1578 1636
A
1339
Discriminating Power
300
1389

200
1446
1200

100 1523

0
1080 1273 1466 1659
600
B 1351
Discriminating Power

1387 1570
1627
400

1426 1510

200

0
1000 1200 1400 1600 1800
Wavenumber (cm-1)

Figure 2.3. Soft independent modeling of class analogy (SIMCA) discriminating power
based on the infrared spectra of whey protein powder from bench-top FT-IR
Spectroscopy (A) and Portable FT-IR Spectroscopy (B).

The amide I region of the spectrum (~1630 cm-1) in figure 2.3 was also associated

with the discrimination of whey protein types. The absorbance in amide I band region is

predominantly due to the stretching vibration of the carbonyl peptide bond, whose

frequency is highly sensitive to hydrogen bonding and thus to protein secondary structure

51
(Kong, 2007). The amide I band with a maximum wavenumber centered near 1633 cm-1,

is characteristic of β-sheet (Farrell Jr., 2001). The aliphatic moieties of amino acid side

chains give rise to vibration absorbance bands at 1465 and 1450 cm-1 associated to δ

(CH3) and δ(CH2) group frequencies while the ν(CN) band near 1 3 cm-1 has been

related to backbone conformation (Barth, 2000). The discriminating power plot for the

model created using spectra from the portable FT-IR unit (Figure 2.3B) also showed the

importance of the carboxylate groups of acidic amino acids (1350, 1387 and 1570 cm-1)

in the classification of the powders.

SIMCA classification models showed large interclass distance values (Table 2.1)

for discrimination among WPI, WPC and WPH samples, evidencing the ability of

infrared spectroscopy combined with chemometrics to achieve reliable resolution of

unique spectral markers. Interclass distance is the distance between the geometric centers

of two clusters; it represents the performance of separation with larger interclass distance

resulting in better separation (Albano et al., 1978). Class distances greater than 3 between

clusters are considered significant for identification of data points as members of a group

(Kvalheim and Kratang, 1992). All of our interclass distances obtained using spectra

from both bench-top and portable FT-IR spectrometers had values larger than 3,

evidencing the excellent discrimination among all classes from the view of statistical

significance (p < 0.05). Interestingly class 5 (WPH-III) achieved large interclass

distances from the rest of the classes (9.42-49.92), which indicated that the CGMP

powder had marked different compositional characteristics from others. Most whey

powder samples fall in classes WPC, WPI (I) and WPH (I), showing interclass distances

52
of 3 (WPC vs WPI (I)), 6 (WPI (I) vs WPH (I)) and 5.5 (WPC vs WPH (I)). The protein

levels of WPH (I) and WPH (II) groups averaged 81.82±2.13% and 90%, respectively,

exhibiting an interclass distance of 4 reflecting differences in composition associated

with levels of protein isolation. Overall, the interclass distances showed excellent

predictive ability of the model for discriminating among the different commercial whey

protein powders.

A
CS1@3 CS2@3 CS3@3 CS4@3 CS5@3 CS6@3 CS7@3
CS11 0.00
CS2 3.33 0.00
CS3 6.3 . 8 0.00
CS4 7.9 9.3 . 9 0.00
CS5 1 .83 1 . 1 .1 16.1 0.00
CS6 7.3 6.78 1 .77 19. 3 .8
CS7 . 7 6. 11.7 1 .3 6. 8 5.31 0.00

CS1@3 CS2@3 CS3@3 CS4@3 CS5@3 CS6@3 CS7@3


CS1 0.00
CS2 3. 0.00
CS3 6. .69 0.00
CS4 1 . 1 . 8 . 8 0.00
CS5 16.18 16.37 9. 3 .1 0.00
CS6 7. 7 8. 1 .3 8. 9 9.9 0.00
CS7 .6 7.73 11. 9 31.3 3 . 8.02 0.00
Table 2.1. Interclass Distances generated from whey protein powder calibration model
using bench-top (A) and portable (B) FT-IR Spectroscopy.
1
CS: CS indicated class assigned to individual category of whey powder products.
Classes were assigned as: Class 1: WPI (I), pH of protein powder ranging from 5.8-7.5,
class 2: WPC, class 3: WPH (I), protein levels were 80%-85% and pH ranging from 6.0-
9.0, class 4: WPH (II), 90% and pH ranging from 7.2-7.8, class 5: WPH (III), CGMP,
class 6: WPI (II), pH ranging from 2.8-3.5, class 7: WPI (III), pH ranging from 3.8-4.0.
pH was measured by dissolving 5g of whey powder in 8 oz of deionized water.
CS1@3: this indicated the number of factor used in SI CA model.
53
Finally, the predictive ability of the model was tested using a validation set of

samples that were not included in the calibration set. The number of test samples used

was 30% of the total samples evaluated for each class (Table 2.2). However, WPI (II) and

WPH (II and III) samples were not included in the validation because of its limited

numbers (3-5 samples). We obtained 100% accurate predictions for all whey protein

powders (Figure 2.2C and 2.2D) that were assigned to its correct class. Furthermore, we

included a soy and gelatin protein to test the model performance in predicting proteins

from different sources. The results showed that these “confounding” proteins were

reliably predicted as not belonging to any of the classes by using the models generated

from spectra by bench-top (Figure 2.2C) and portable (2.2D) spectrometers.

Number of Sample Involved


Protein Powder Type FT-IR Benchtop Portable FT-IR
Spectroscopy Spectroscopy
Training Validation Training Validation
WPC WPC 24 8 10 5
WPI (I) 34 16
WPI (II) 12 7
WPI
WPI (III) 10 4
Total 44 12 20 7
WPH (I) 16 16
WPH (II) 5 2
WPH
WPH (III) 3 3
Total 19 5 16 5
SOY PROTEIN 1 1
GELATIN 1 1
Table 2.2. Number of protein samples used for spectral acquisition using the bench-top
and portable FT-IR Spectrometers
54
2.5 Conclusion

FT-IR spectroscopy combined with pattern recognition chemometrics had the ability to

authenticate commercial whey protein products, clustering powders according to different

manufacturing processing, reflected by differences in powder pH (WPI) and protein

levels (WPH). Portable FT-IR spectroscopy proved its robustness by producing models

with high spectral resolution and large interclass distances resulting in accurate prediction

ability, which made this technology a perfect choice for “in-field” authentication.

55
Chapter 3: A Novel Application of FT-IR Technique Combined with Chemometrics
in Analyzing and Predicting the Astringency of Acidic Whey Protein Beverages

3.1 Abstract

Formulating whey protein beverages at an acidic pH, in order to obtain better

clarity, normally develops an unpleasant and astringent flavor. ur ob ective was to

explore the application of infrared spectroscopy in predicting the astringency of acidic

whey protein beverages. Three types of whey protein powders including whey protein

isolate (WPI), whey protein concentrate (WPC), and whey protein hydrolysate (WPH)

from different manufacturers were used to formulate beverages at pH range . - 3.9.

Each sample was sensory tested through trained panelists by the SpectrumT method of

descriptive analysis for generation of quantitative astringency data. Three Fourier

transform infrared spectroscopy (FTIR) techniques, including FTIR- attenuated total

reflectance (ATR) icro-spectroscopy, FTIR-Reflectance icro-spectroscopy and

Portable FTIR-ATR spectrometer were used for spectra collection. Spectra were analyzed

by multivariate regression analysis (Partial least squares regression, PLSR) to build

calibration models with the combination of sensory astringency scores obtained from

trained panelists. The PLSR analysis showed a strong relationship between the reference

astringency values and the IR predicted values. PLSR models based on different protein

types (WPI, WPC, and WPH beverages) were generated for astringency determination

with standard error of cross validation (SECV) less than . 7 and correlation coefficient
56
larger than .7 . a or characteristic absorption bands related to astringency were
-
associated with the C groups and amide I & II regions. A rapid and reliable technique

to accurately and precisely predict the astringency of acidic whey protein beverages was

developed, which can contribute to improve sensory characteristics of whey-containing

products and provide a novel tool for timely corrective actions during processing, thus

decreasing manufacturing cost.

3.2 Introduction

The use of whey proteins in beverages in food industry is attractive because of

their high nutritional value and wide functional versatility. Also the consumption of whey

protein beverages is appealing due to weight control, muscle building and health benefit.

Accompanying the increasing popularity of whey protein beverages, the sensory

challenge of astringency has developed (Sano et al., eecher et al., 8) when

they are formulated at low pH to improve beverage clarity and heat stability ( iller,

7). This astringency limits consumer acceptance of these products.

Astringency is described as a drying, puc ering sensation which is typically

caused by four ma or types including vegetable tannins, dehydrating agents (alcohol),

salts of multivalent metallic cations (alum), and acids ( ate-Smith, 19 , Siebert and

Chassy, ). Astringent foods such as wines, teas, fruits, and soy-based products are

one of the most widely studied categories, and the mechanism of astringency has been

attributed to polyphenolic compounds (Haslam et al., 1988) that bind and precipitate

specific salivary proteins ( axter et al., 1997 Kallithra a et al., 1998), diminishing

57
saliva’s ability to lubricate the mouth (Clifford, 1996). Further studies included whey

protein beverages as a new member in astringency food family (Sano et al., ),

however, the exact cause of astringency in acidic whey protein beverages is un nown.

Recent studies that have investigated the astringency of whey protein beverages have

proposed that the whey proteins are directly responsible for the astringency. eecher and

others ( 8) suggested that at low pH, positively charged whey proteins are capable of

binding with and aggregating salivary proteins, causing astringency. Whereas Lee and

Vic ers ( 8) suggested that the astringency of acidic whey protein beverages was

possibly caused by their high acidity.

Currently, astringency of whey protein beverages is determined by trained sensory

panelists, a benchmar in qualifying and quantifying sensory properties, however, it is

laborious, expensive and time consuming. With the demand for protein fortified foods

and beverages increasing steadily, a quic and reliable method for predicting astringency

to improve overall flavor is important. Time and money can be saved through analysis of

a few variables whereby quality control of the product could be established early in the

manufacturing process. Infrared spectroscopy ( -7 cm-1) is an attractive technology

for the rapid, sensitive, and high-throughput analysis of food components, producing a

molecular "fingerprint" spectrum that represents the absorption bands corresponding to

frequencies of vibrations between the bonds of the atoms ma ing up the material.

Furthermore, specific fairly narrow bands arising from group vibrations may be assigned

to nown specific chemical entities in most cases. Developments in the field of

multivariate techniques have been prompted by the need of reliable, accurate, robust and

58
simple methods for routine analysis of spectroscopic data (Udelhoven et al., ).

Principal component regression (PCR) and partial least-squares regression (PLSR) are

factor analysis methods that have the potential to estimate the component concentration

from the infrared spectra (Haaland and Thomas, 1988). ur ob ective was to develop an

astringency predictive model for acidic whey protein beverages based on the relationship

between the perceived intensity of the selected sensory stimuli and unique infrared

spectral fingerprints.

3.3 Materials and methods

3.3.1 Whey protein sample

Whey protein beverages were made from whey protein isolate (WPI), whey

protein concentrate (WPC) or whey protein hydrolysate (WPH) powder obtained from

different manufacturers. Sucrose was added in each formulation, and phosphoric acid

( %) or Potassium carbonate ( %) was used to ad ust the pH of individual beverage

sample to target value. Two batches of beverage samples were prepared separately. The

first batch was controlled under uniform pH value (pH=3. ) after formulation from

different powders, including WPI (3 ), WPC (18), and WPH (1 ) powders. In the second

batch, beverage formulated by each whey protein powder was further ad usted to

different pH values (ranging from . to 3.9). WPI (9), WPC (6), and WPH ( ) powder

samples from different manufacturers were used in the second batch. Each sample was

coded with 3 digit number and vacuum pac ed in a glass bottle. everage samples were

maintained at ºC until spectra collection using FT-IR.

59
A separate batch of samples was made from pre-selected powders, including

whey protein powders and 1 gelatin powder to evaluate the model performance. Each

whey protein solution was formulated from 1 type of powder along with sucrose, and

dissolved in water. After measuring pH of each fresh solution, phosphoric acid ( %) or

potassium carbonate ( %) was used to ad ust the pH to , 3, , , 6 and 7. Solids

measurement was ta en before and after pH ad ustment to verify how the concentration

may have changed.

3.3.2 Astringency prediction

Quantitative sensory test on all whey protein samples mentioned above was

conducted at the Sensory Service Center at North Carolina State University, which is an

internationally recognized program in Dairy products. Sensory evaluation was conducted

by panelists trained in the Spectrum method of descriptive analysis for generation of

quantitative data. A to 1 -point universal intensity scale was utilized to describe

astringency score, where meant no intensity of astringency perception and 1 indicated

highest intensity of astringency possible.

3.3.3 FTIR Micro-spectroscopy

To perform spectra collection from each beverage sample, a Varian 6 U A IR

microscope (Varian, Randolph, A) interfaced with an Excalibur series 31 GX FT-IR

spectrometer (Varian, Walnut Cree , CA) was used for rapid data scan. The FTIR

microscope was equipped with a dynamically aligned ichelson interferometer, a

motorized x-y stage, a potassium bromide beam splitter, and a broadband mercury-

cadmium-telluride ( CT) detector. Spectra were observed using Resolution Pro Software

60
(Version . , Varian). IR spectra were collected from to 7 cm-1 with co-adding

1 8 scans to increase signal to noise ratio. easurements were performed by two

approaches: using a slide-on attenuated total reflectance (ATR) germanium ob ective

(Varian 6 U A, Palo Alto, CA) or FT-IR reflectance. In both methods, aliquots ( . ul)

of the beverage samples were placed on a microscope slide, and vacuum dried for

minutes to form a thin film. The slide was placed directly onto the stage of Varian 6

U A IR microscope. Resolution of 8 cm-1 was used in the ATR approach, and Resolution

of cm-1 was used in the reflectance approach.

3.3.4 Portable FTIR Spectroscopy

easurements were also performed in a portable system (Cary 63 FTIR

Spectrometer, Agilent Technologies, Santa Clara, CA) coupled with a diamond ATR

accessory, a temperature-stabilized DTGS detector, and ZnSe beam splitter. In every

spectrum collection step, . ul of the beverage sample was placed on the center of

diamond crystal then water vacuum was applied for 1 minute to get rid of the ma ority of

the water in the beverage. Finally, a nice film was formed on top of diamond crystal and

it was the moment to ta e spectrum measurement. The IR spectra were collected using

icroLab software (Agilent Technologies, Santa Clara, CA) operating in the

wavenumber ranges from to 7 cm-1 with resolution of cm-1, and 6 scans were

co-added to increase signal to noise ratio.

3.3.5 Multivariate regression analysis

The astringency scores of whey protein samples were used to test the ability to

predict the sensory attributes based on their IR spectra in a multivariate model.

61
ultivariate regression analysis (Partial least squares regression, PLSR) can analyze data

with strongly correlated, noisy, and numerous X-variables, and also simultaneously

model responser variables (Wold et al., 1). In this experiment, PLSR was performed

to build calibration models to simultaneously correlate the reference sensory astringency

score to the infrared spectral information. PLSR is a bi-linear regression analysis method

that extracts a small number of orthogonal factors that are linear combinations of the

spectral (X) variables, and uses these factors as regressors for the analyte’s concentration

(Y-variable). These orthogonal factors (latent variables) explain as much as possible of

the covariance of the X and Y variables. PLSR has the potential to estimate the

component concentration and chemical and physical properties (loading vectors, vector

of final calibration regression coefficients, and spectral residuals) from the spectra. PLSR

has been particularly successful in developing multivariate calibration models for the

spectroscopy field because it uses the concentration information (Y-variable) actively in

determining how the regression factors are computed from the spectral data matrix (X),

reducing the impact of irrelevant X-variations in the calibration model. This capability

provides a more information-rich data set of reduced dimensionality and eliminates data

noise which results in more accurate and reproducible calibration models. (Wold et al.,

1). efore analysis, pre-processing methods such as standard normal variate

transformation (SNV) and Savitz y–Golayand second derivative were used to remove the

slope variation from spectra caused by scatter and variation of particle size ( arnes et al,

1989 Candolfi et al., 1999), to resolve overlapping bands, and to correct for baseline drift

(Rinnan et al., 9b). PLSR models were evaluated in terms of loading and score

62
vectors, standard error of calibration (SEC), standard error of cross validation (SECV),

coefficient of determination (r-value) and outlier diagnostics. The PLSR calibration

model was cross validated (leave-one-out approach) to validate the prediction ability of

new data and estimate the chance to obtain a good fit of random sample (Wold et al.,

1).

3.4 Results and discussion

3.4.1 Perceived astringency scores for whey protein beverages

Whey protein beverages manufactured with a fixed pH (3. ) ranged in astringency

scores from 1.9 to 3.7 for WPC, . to . for WPI and .1 to .3 for WPH, indicating the

importance of protein source in the astringency sensory response. Furthermore,

astringency scores of whey protein beverage samples manufactured at various pH levels

(Table 3.1) showed that whey protein beverage samples exhibited different astringency

patterns when their pH values decreased from 3.9 to . : Statistical analysis ( ultiple

Linear Regression, Table 3. ) showed that astringency scores were explained by pH and

protein source (type and processing) (p-value < . ) in acidic whey protein beverages,

revealing the complexity of astringency development in acidic whey protein beverages in

the pH ranges of . -3.9. eecher and others ( 8) also evaluated the effect of pH on

astringency development in whey protein solutions and found that whey protein solutions

exhibited higher astringency at acidic pH (~3. ) than neutral pH (6.8), indicating the

importance the pH levels to explain the astringency. They also found a decrease in

astringency between pH 3. and .6 that coincided with an increase in sourness ( eecher

63
et al., 8). ur results indicated that differences in astringency of whey protein

beverages at acidic pH might also be associated to the type of processing of the whey

proteins during manufacture of the powders.

Whey Protein Type


Whey Protein Concentrate Whey Protein Isolate Whey protein Hydrolyzate
Sample pH Astringency RSD Sample pH Astringency RSD Sample pH Astringency RSD
2.59 3.8 5.69 2.40 2.6 2.41 2.5 16.84
2.89 3.7 7.49 2.48 3.5 2.74 2.4 12.52
WPC1 3.53 2.9 6.75 WPI1 2.49 2.3 WPH1 3.19 2.1 13.53
3.34 3.4 7.87 3.29 3.0 3.43 2.0 22.07
3.87 3.3 8.36 3.70 2.6 3.78 1.9 20.21
2.56 3.7 5.96 2.43 2.9 2.52 3.8 7.91
2.86 3.6 6.86 2.65 3.1 2.82 3.6 6.41
WPC2 3.21 3.6 6.55 WPI2 3.08 2.6 WPH2 3.34 3.8 6.14
3.57 3.0 7.96 3.45 3.0 11.68 3.58 3.3 8.08
3.91 3.7 6.38 3.72 3.3 10 3.95 2.9 8.54
2.58 3.4 6.82 2.25 2.8 2.51 3.5 5.99
2.85 3.7 5.55 2.53 2.9 2.77 3.7 6.2
WPH3
WPC3 3.30 3 7.47 WPI3 2.86 2.9 3.24 3.0 9.89
3.58 3.1 6.78 3.29 3.2 3.65 2.9 6.51
3.85 3.5 5.52 3.56 2.8 3.96 2.5 9
2.56 3.7 6.46 2.55 3.4 2.44 2.9 6.32
2.86 3.2 7 2.82 3.4 2.76 3.6 5.26
WPC4 3.28 2.7 14.05 WPI4 3.24 2.9 WPH4 3.27 3.0 7.12
3.56 2.9 4.14 3.50 2.7 3.47 3.0 5.01
3.89 2.5 8.08 3.77 2.3 3.87 3.0 6.3
2.61 3.6 7.17 2.59 3.2
2.89 3.5 5.47 2.87 2.9
WPC5 3.26 3 9.1 WPI5 3.31 2.9
3.53 2.9 5.89 3.56 2.4
3.83 3.1 7.47 3.87 2.8
2.63 3.8 4.64
2.91 3.5 8.06 2.62 3.6 7.51
WPC6 3.34 3 6.73 WPI6 2.93 3.2 9.47
3.63 3.2 4.48 3.31 3.1 9.58
3.92 3.1 5.86 2.38 3.0 9.93
2.60 3.2 11.05
WPI7 3.04 2.8 10.43
3.28 2.7 10.39
3.57 3
2.59 3.0 7.25
WPI 8 2.87 3.1 8.6
3.08 3.1 6.8
2.47 3.8 8.67
2.81 3.2 8.73
WPI9 3.11 3.7 8.61
3.46 3.2 8.39
3.80 3 12.03
Table 3.1. Sensory astringency scores perceived from whey protein beverage samples at
various pH values ranging from . to 3.9.

64
Protein Soures pH Levels1 Protein Types R (%)
P(regression)
WPI 9 . .
WPC 6 6 .3 . 1
WPH 9 . .
Table 3.2. The multiple linear regression using pH levels and protein types as predictor
and astringency as response.
1
For each protein types from different protein sources, 5 different pH levels ranging from
2.2 to 3.9 were used as explanatory variable.

3.4.2 Comparing spectra information from whey protein powder and beverage

We developed a novel protocol to analyze whey protein beverages by IR icro-

spectroscopy by using a SpectRI (Tienta Sciences) slide, which provided a high

infrared reflectance surface and the ability to concentrate the samples up to 1 times due

to the presence of a hydrophobic coating. This technique permitted collection of high-

quality and reproducible spectral data from small amounts of samples (~ . - μL)

through formation of thin homogeneous films after vacuum-drying. This technique was

selected because it provided high-throughput capabilities allowing data collection in <1

min.

The infrared spectrum of whey protein isolate powder and beverage (Figure 3.1)

in the information-rich region of 18 -7 cm-1 showed characteristic bands arising from

functional group vibrations. The large broad bands in the raw spectrum at 169 ad 1637

cm-1 in the raw spectrum were associated with amide I (α-helix and β-sheet structures of

proteins) while the signal at and 1 cm-1 corresponded to amide II (combined δ(N-H)

and ν(C-N) contributions of the peptide bond) group vibrations of proteins. The band at

65
approximately 1 cm-1 resulted from the deformation bending of C-H in the >CH

groups and the band at 139 cm-1 corresponded to the stretching of C= in the C -

groups (Figure 3.1). A distinct feature in the spectrum was associated with the broad pea

band ~ 1 cm-1 corresponding to the combined signal from the amide III groups and
-
P= stretching in phosphodiesters (>P ). The bands between 1 and 9 cm-1

resulted from the vibration of C- -C and C- were attributed to sugars, acids and/or

polysaccharide components of the samples. This latter region was prevalent in the

beverage samples (Figure 3.1) as they contained added sugars and acids in the matrix.

Whey Protein powder


Whey Protein beverage uC-O-C of

Amide II
polysaccharides
of acyl chain of triacylglycerols

CH groups of lipids
Amide I
CH and carbonyl stretching
O-H groups
water
Absorbance

C=O groups of lipids

3600 3200 2800 2400 2000 1600 1200 800

Wavenumber (cm-1)

Figure 3.1. Attenuated Total Reflectance (ATR) id-Infrared spectra of Whey protein
powder and beverage.

66
The spectral complexity required the application of mathematical filter functions

(SNV and second derivative) and selection and combination of certain spectral regions to

enhance spectral resolution. athematical (second derivative) transformations of the

spectra highlighted distinct features between the samples (Figure 3. ) and reduced

spectral noise helping to minimize the variability between replicates.

Powder
Absorbance

Beverage

3600 3200 2800 2400 2000 1600 1200 800


Wavenumber

Figure 3. . Second derivative spectral transformation collected using an Attenuated Total


Reflectance Infrared (ATR-IR) accessory equipped either with a ZnSe crystal plate
(bench-top) or Germanium crystal ( icro-spectroscopy).

3.4.3 SIMCA classification of whey protein beverages with fixed pH

Evaluation of whey protein beverages with fixed pH (3. ) value was examined

using FTIR-ATR icro-spectroscopy and SI CA classification plot (Figure 3.3) was

67
generated to visualize well-separated clustering among the beverage samples. Evaluation

of the SI CA classification plots showed that most samples formed tight clusters, and

were far away from others with interclass distance values greater than 1 for

discrimination among WPI, WPC and WPH samples, evidencing the ability of infrared

spectroscopy combined with chemometrics to achieve reliable resolution of unique

spectral mar ers. Usually interclass distances greater than 3 between clusters were

considered significant for identification of data points as members of a group (Kvalheim

and Kratang, 199 ).

Figure 3.3. Soft independent modeling of class analogy (SI CA) classification plot and
discriminating power based on the infrared spectra of whey protein beverage samples.

The SI CA discriminating power algorithm showed that the 16 1 cm-1 band was

most important for separating the different classes, corresponding to disordered β-sheet

!
68
bands. In peptides, the frequency of the C-amide I band is a sensitive local probe of

secondary structure and the alignment of residues within a β-sheet. As the peptides align,

the C-amide I band at 16 1 cm-1 (due to the misaligned form) decreases in intensity and

the band at 1 91 cm-1 (due to the aligned form) increases (Petty et al., ). These

irreversible changes in protein structure are possibly due to heating or allowing the

sample to stand at room temperature for extended periods of time (Petty et al., ). The

band at 16 1 cm-1 has also been assigned to the characteristic N-H bending vibrations of

Asp. Although the original Asp band can be observed at 1 97 cm-1, with increasing

concentration the band shifts toward 16 1 cm-1.

3.4.4 PLSR calibration models of whey protein beverages with fixed pH

The PLSR plots and loadings for whey protein beverages with fixed pH value

(WPI, WPC, and WPH beverages) were generated separately in Figure 3. . Among all

three types of whey protein products, WPH generated the best calibration models with the

highest R ( .98) and lowest standard error of cross validation (SECV= .11) shown in

Table 3.3. WPI and WPC beverages also performed strong correlations between IR

spectra and sensory astringency score (R> .81, SECV< . ).

69
3.8 A. WPI B. WPI
0.2

3.4
-0.0
1582

3.0 -0.2

1601
2.6 -0.4
IR Predicted Value

2.6 3.0 3.4 1543 1620 1697 1775


C. WPC D. WPC
3.4 0.2
1535

Loadings
3.0 -0.0

1504
-0.2 1582
2.6
1647

-0.4
2.2

2.4 2.8 3.2 3.6 1466 1543 1620 1697

E. WPH F. WPH
4.0 0.2

3.5 -0.0
1582

3.0 -0.2
1574 1690

-0.4
2.5

-0.6
2.0 2.5 3.0 3.5 4.0 1543 1620 1697
Wavenumber (cm-1)
Sensory Astringency Score

Figure 3.4. Partial least squares regression (PLSR) plots and loadings based on the
infrared spectra of whey protein WPI, WPC and WPH beverage samples with fixed pH
(3.5) using FTIR-ATR germanium Micro-spectroscopy.

70
WPI WPC WPH
1
Sample Number 16 13
-1
Wavenumber Range (cm ) 1 -18 1 -1697 1 97-1736
Factor 6
Cumulative 1 . 99.99 1 .
SECV3 .17 . .11
R .89 .81 .98
SEC .16 . 3 .1
R Cal6 .91 .8 .99
7
RPD .18 1.7 .

Table 3.3. Calibration and cross-validation results of multivariate models developed by


whey protein beverages with fixed pH (3.5) using FTIR-ATR Micro-spectroscopy
spectra. 1Sample Number: Whey protein beverage samples used in individual model.
Cumulative: Cumulative variances explained by the calibration model. 3SECV: Standard
error of cross-validated calibration model. R: Coefficient of cross-validated calibration
model. SEC: Standard error of calibration model. 6R Cal: Coefficient of calibration
model. 7RPD: relative percent deviation

The model producing the minimum SECV, standard error of prediction

(magnitude of error expected when independent samples are predicted using the model),

was selected as the best model for the spectral data set. Examination of the loading

indicated bands explaining the regression between infrared spectra and astringency scores

in all three models was 1 8 cm-1, which was associated with anti-symmetric stretching

of Asp C groups ( izuguchi et al., 1997).

3.4.5 PLSR calibration models of whey protein beverages with various pH

The effect of various pH values of whey protein beverages were modeled for

predicting the astringency scores by trained sensory panel. The performance of models

generated from spectra collected by using FTIR icro-spectroscopy (ATR and

71
reflectance modes) and Portable FTIR-ATR spectrometer were evaluated for predicting

whey protein astringency. As illustrated in both Figure 3. and Table 3. , good regression

models were obtained for all three whey protein products using FTIR-ATR icro-

spectroscopy, with the smallest R equal to .8 . The most important bands for Factor 1

(solid line) in the loading plot for WPI, WPC, and WPH beverages were centered at 1

and 16 7 cm-1 that were related to the structural information on proteins. The large broad

bands in the raw spectrum at 16 7 cm-1 were associated with amide I (α-helix structures

of proteins) while the signal at and 1 cm-1 corresponds to amide II (combined δ(N-H)

and ν(C-N) contributions of the peptide bond) group vibrations of proteins (Kong and Yu,

7). While Factor 1 explained a large percentage of the model variance, Factor

(dashed line) also played an important role in revealing functional groups in protein

contributing to the establishment of astringency model. The bands 1 8 cm-1 observed in



Factor of WPC beverage models indicated the importance of C groups ( izuguchi

et al., 1997).

72
3.5 A. WPI B. WPI

0.2

3.0 1539
-0.0

1562
2.5
-0.2
1647 1690

2.5 3.0 3.5 1543 1620 1697 1775

C. WPC D. WPC
0.2
IR Predicted Value

0.1
3.5

Loadings
-0.0 1431
1585
1539
3.0 -0.1
1443
-0.2
1686
1647
2.5
2.5 3.0 3.5 1466 1659

E. WPH 0.3
F. WPH
3.6

3.2
0.1

2.8

1543
2.4 -0.1 1458
1427
1620
2.0
1647
-0.3
2.2 2.6 3.0 3.4 3.8 1466 1659
Sensory Astringency Score Wavenumber (cm-1)

Figure 3.5. Partial least squares regression (PLSR) plots and loadings based on the
infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH
(ranging from 2.2 to 3.9) using FTIR-ATR germanium Micro-spectroscopy.

PLSR calibration models of WPI, WPC, and WPH based on FTIR-Reflectance icro-

spectroscopy and Portable FTIR-ATR spectrometer also formed good relationship

between infrared spectra and sensory astringency scores, as illustrated in Table 3. . With

all three applications of IR techniques, very robust calibration models (R> .93) were

73
constructed based on WPH beverage samples. However WPI beverages showed the

lowest performance possibly due to the complexity of the matrix that increases the signal

interference and decreases the ability to resolve the target signal.

FTIR-ATR icro- FTIR-Reflectance Portable FTIR-ATR


spectroscopy icro-spectroscopy Spectrometer
WPI WPC WPH WPI WPC WPH WPI WPC WPH
Sample number 3 36 8 9 7 1
Wavenumber 1 1 1 1 1 37 13 1 1 1 1
range (cm-1) -1798 -1798 -1798 -1699 -18 -18 -17 8 -176 -1698
Factor 6 6 6 6 6 6 6
Cumulative 99.99 99.98 99.9 99.89 99.98 99.7 99.9 99.99 99.9
SECV .16 .1 . .19 .1 . .18 .17 .1
R .8 .9 .93 .77 .91 .9 .7 .8 .93
SEC .16 .1 . .18 .13 .18 .1 .16 .1
R Cal .86 .96 .9 .8 .93 .96 .83 .88 .96
RPD 1.8 .98 .67 1. 7 .39 .91 1. 8 1.91 .8
Table 3.4. Calibration and cross-validation results of multivariate models developed by
whey protein beverages with various pH using FTIR-ATR Micro-spectroscopy, FTIR-
Reflectance Micro-spectroscopy, or Portable FTIR-ATR Spectrometer spectra.

Comparing the loading of both FTIR-Reflectance icro-spectroscopy (Figure

3.6) and Portable FTIR-ATR spectrometer (Figure 3.7) with FTIR-ATR icro-

spectroscopy (Figure 3. ), similar bands arising from Factor 1 in every loading plot were

confirmed to be related to amide I (centered at 16 cm-1) and amide II (centered at 1

cm-1) regions, which indicated that when pH was involved as a variable in product

manufacturing, protein secondary structure played a very important role in regulating

astringency generated from whey protein beverages. eanwhile it should not be

74
negligible that the loading plot disclosed important bands including 1333, 1 ,1 9 ,

1 96, and 1 99 associated with functional groups rather than protein secondary structure,

which were meaningful in understanding the regression models. IR bands at 1 99 and

1 9 cm-1 exhibited in WPI and WPC calibration models by FTIR-Reflectance icro-

spectroscopy can be assigned to the characteristic N-H bending vibrations of Asp

( izuguchi et al., 1997). Calibration model based on the WPH spectra by FTIR-

Reflectance icro-spectroscopy also gave an important band at 1333 cm-1, which most

li ely can be attributed to the carboxylate vibrations of Asp groups due to protonation

change (Gerwert et al., 1989). Examining the loadings of WPI, WPC, and WPH beverage

calibration models constructed by Portable FTIR-ATR spectrometer, IR bands at 1 96

cm-1 arising from WPI model, while IR bands at 1 cm-1 existed in all three models.

The absorbance bands at 1 cm-1 was possibly related to the δ(CH ) group frequencies

of aliphatic side chains of amino acid ( arth, ).

75
A. WPI 0.2 B. WPI
3.4
0.1

3.0
-0.0
1329
2.6 1256 1352
-0.1 1541 1599
1454
1649
2.2 -0.2
2.5 3.0 3.5 1275 1468 1661

C. WPC D. WPC
3.8
0.2
IR Predicted Value

3.4 0.1

Loadings
-0.0
3.0
-0.1 1595
1541
2.6 -0.2 1647 1715

2.5 3.0 3.5 1564 1661 1757

E. WPH 0.2 F. WPH


3.5

0.1
3.0

-0.0
2.5
1333 1362

2.0 -0.1 1541


1456
1653
1639
2.2 2.6 3.0 3.4 3.8 1468 1661
Sensory Astringency Score Wavenumber (cm-1)

Figure 3.6. Partial least squares regression (PLSR) plots and loadings based on the
infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH
(ranging from 2.2 to 3.9) using FTIR-reflectance Micro-spectroscopy.

76
3.5 A. WPI 0.2 B. WPI

0.1

3.0
-0.0

-0.1 1542 1454


2.5 1596

-0.2 1655 1639

2.5 3.0 3.5 1672 1579 1486

C. WPC 0.2 D. WPC


3.8
IR Predicted Value

0.1

Loadings
3.4 -0.0

-0.1 1542 1454


3.0 1491

-0.2 1639

3.0 3.4 3.8 1765 1672 1579 1486

E. WPH F. WPH
0.2

3.4
0.1

-0.0
3.0

1534 1454
-0.1 1620
1486 1430
2.6 1640

2.5 3.0 3.5 1672 1579 1486


Sensory Astringency Score Wavenumber (cm-1)

Figure 3.7. Partial least squares regression (PLSR) plots and loadings based on the
infrared spectra of whey protein WPI, WPC and WPH beverage samples with various pH
(ranging from 2.2 to 3.9) using portable FTIR-ATR diamond equipment.

Interestingly, three models (WPI, WPC and WPH) generated with the spectra

collected for beverage samples with fixed and various pH showed good performance

statistics (Table 3. ) and correlation with astringency sensory scores (Figure 3.8) by a

descriptive sensory analysis. The PLSR models based on FTIT-ATR spectra collected

77
from whey protein-containing beverages (R> .79, and SECV< . 7) gave satisfactory

performance statistics.

A. WPI B. WPI
3.5
0.2

3.0 -0.0

1539
1562
-0.2
2.5 1686
1647

2.5 3.0 3.5 1543 1620 1697 1775


4.0 C. WPC D. WPC
0.2
IR Predicted Value

3.6
Loadings

3.2 -0.0

1539 1593
2.8 1686
-0.2 1431
1647
2.6 3.0 3.4 3.8 1466 1659
4.0
E. WPH F. WPH
0.2
3.5

3.0
-0.0
2.5 1543
1431
2.0 -0.2 1620
1647

2.0 2.5 3.0 3.5 4.0 1466 1543 1620 1697


Sensory Astringency Score Wavenumber (cm-1)

Figure 3.8. Partial least squares regression (PLSR) plots and loadings based on the
infrared spectra of whey protein WPI, WPC and WPH beverage samples with fixed pH
(3.5) and various pH (ranging from 2.2 to 3.9) using FTIR-ATR germanium Micro-
spectroscopy

78
WPI WPC WPH
Sample Number 39 31
-1
Wavenumber Range (cm ) 1 93-1798 1 -1798 1 -17 8
Factor 6
Cumulative 99.9 99.99 99.67
SECV .18 . . 7
R .8 .79 .88
SEC .17 . 3 .
R Cal .83 .83 .9
RPD 1.77 1.6 .13

Table 3.5. Calibration and cross-validation results of multivariate models developed by


whey protein beverages with fixed and various pH using FTIR-ATR Micro-spectroscopy
spectra.

The regions that gave the lowest SECV were chosen for developing cross-

validated PLSR calibration models given in Table 3.3, 3. , and 3. , which were different

for each individual model. The factor selected for each model should be able to represent

enough information to build the calibration model, and at the same time, should not be

too large to include the structural information (irrelevant variables). The factors selected

in all models were ranging from to 6, explaining larger than 99% variables for all of our

calibration models. Ta ing into consideration that the standard deviation among panelist

for evaluating astringency scores ranged from . - . units with a %RSD ranging from

. to .1 (Table 3. ), the predictive ability of our infrared models with SECV < . 7

units showed good reliability, since the %RSD will also be a source of variability for the

regression models. y maximizing the variance from whey protein sources, pH values,

processing techniques, and sensory evaluation, robust calibration models to predict

79
astringency generated from acidic whey protein beverages based on different whey

protein types were constructed by multivariate regression combined with infrared spectra.

During the investigation of loadings, infrared bands responsible in constructing the PLSR

astringency models were also revealed, which could help us to understand the reasons

behind. It pointed out that whey protein itself, and pH were possible reasons to address

astringency from acidic whey protein beverages.

3.4.6 PLSR prediction of a separate set of whey protein beverages

Table 3.6 showed the astringency prediction value of a separate set of beverages

made from specific whey protein powders and ad usted to various pH levels. Astringency

score predictions were done by collecting sample infrared spectra and using the

previously cross-validated WPI, WPC and WPH models by FTIR-reflectance icro-

spectroscopy. The predicted scores showed good precision associated with low standard

deviation calculated from replicates. verall, as the pH level of whey protein beverage

increased (from to 7), the astringency associated with each sample showed a decreasing

tendency.

The impact of modifying the pH on the predicted astringency scores on beverage

samples was higher when formulated with WPC and WPH powders, while the

astringency scores of beverage samples formulated from WPI powders showed a slight

declined in astringency scores as pH levels were raised. The astringency scores of GPH1

showed the smallest changes among all whey protein powders, probably because the

ma or protein in GPH1 was actually gelatin. Total solids measurement indicated similar

solid concentrations among all samples. ur results showed that the pH values had

80
apparent impact on the induced astringency of whey protein beverages, while the pH

ranges were evaluated from acidic conditions to neutral conditions. This result agreed

with previous wor done by eecher and others ( 8), who also emphasized the ma or

effect of pH on astringency from whey protein solutions.

Assigned ID Actua Prediction SD °Brix Assigned ID Actual Prediction SD °Brix


l pH Score pH Score
WPC2_01 6.516 1.90 0.06 10.8 WPC6_0 6.572 1.00 0.17 10.8
WPC2_2 2.066 3.73 0.01 11 WPC6_2 2.079 3.40 0.13 11
WPC2_3 3.108 3.04 0.03 10.6 WPC6_3 3.085 2.78 0.03 10.6
WPC2_4 4.046 2.45 0.17 10.4 WPC6_4 4.102 2.16 0.11 10.2
WPC2_5 4.993 2.19 0.04 10.4 WPC6_5 4.988 1.48 0.04 10.5
WPC2_6 5.907 1.95 0.01 10.8 WPC6_6 6.183 0.96 0.03 10.5
WPC2_7 7.349 1.65 0.00 11.2 WPC6_7 6.911 0.79 0.01 10.4

GPH1_0 5.511 1.72 0.02 10.2 WPH3_0 6.654 2.71 0.15 11


GPH1_2 2.099 2.00 0.04 10.4 WPH3_2 2.08 4.08 0.08 11.4
GPH1_3 3.066 1.77 0.01 10.2 WPH3_3 3.079 3.23 0.02 11
GPH1_4 4.011 1.79 0.03 10.2 WPH3_4 3.962 2.99 0.01 10.6
GPH1_5 4.756 1.73 0.02 10 WPH3_5 5.077 2.89 0.02 10.6
GPH1_6 6.022 1.71 0.02 10 WPH3_6 6.072 2.81 0.17 10.8
GPH1_7 6.959 1.63 0.01 10 WPH3_7 7.008 2.76 0.05 10.8

WPI2_0 3.671 2.57 0.02 10.2


WPI2_2 2.036 2.96 0.01 10.4 WPI1_2 2.086 2.91 0.02 10.4
WPI2_3 2.959 2.72 0.00 10.2 WPI1_3 3.094 2.73 0.01 10.2
WPI2_4 4 2.55 0.02 10.1 WPI1_4 4 2.61 0.00 10.2
WPI2_5 5.276 2.35 0.01 10 WPI1_5 5.051 2.47 0.02 10
WPI2_6 5.992 2.29 0.01 10.2 WPI1_6 5.64 2.41 0.00 10.2
WPI2_7 7 2.18 0.01 10.2 WPI1_72 7.027 2.24 0.01 10

Table 3.6. The predicted values of astringency in formulated whey protein beverages
using IR spectra and calibration models by FTIR-reflectance Micro-spectroscopy.
1
Sample codes including indicate that they are beverages before pH ad ustment and the
actual pH levels are natural pH levels when whey protein beverages were formulated
from powder
The pH level of WPI1 beverage was 7. 7 before any pH ad ustment hence WPI1_7
represents both WPI1_ and WPI1_7

81
3.5 Conclusion

FTIR spectroscopy has allowed for the rapid (~ min analysis time) and accurate

analysis of WPI, WPC, and WPH beverages and contributed to the development of

simple and rapid protocols for predicting astringency and the final quality of the

beverages, which finally could enhance the flavor of whey protein products and

increasing consumer li ing of the whey protein food products.

82
Chapter 4 Understanding Factors Influencing Development of Astringency in Whey
Protein Products

4.1 Abstract

Whey proteins beverages are often formulated to be acidic in order to improve

their clarity and microbial stability but as a result develop an unpleasant and astringent

flavor. ur ob ective was to evaluate the factors regulating astringency in whey protein

products by the statistical and instrumental methods. Whey protein isolate (WPI), whey

protein concentrate (WPC) and whey protein hydrolyzate (WPH) from different

manufacturers were used to formulate beverages at acidic pH ranges. Powders were

pressed onto a three-reflection ATR-IR diamond crystal with a high-pressure clamp for

spectra. everages were measured by FTIR Reflectance icro-Spectroscopy with the aid

of vacuum drying. Trained panelists from Sensory Service Center at North Carolina State

University using descriptive analysis were involved to determine astringency scores.

Whey protein powder samples were acid hydrolyzed to free amino acids and analyzed by

Gas Chromatography method to quantify the concentrations of acidic amino acids.

ultiple Linear Regression ( LR), Partial Least Square Regression (PLSR) and Pearson

Correlation were used to evaluate the factors regulating astringency in whey protein.

LR using pH and protein types as explanatory variables explained high levels of

variance ( . % - 9 . %), illustrating their importance in regulating astringency in whey

protein beverages. Good PLSR models based on whey protein powders were generated
83
for astringency determination (SECV< .3 and correlation coefficient (r> .7 ). a or

characteristic absorption bands detected for astringency, in the range of 1396 to 17

cm1, were also associated with the C -


groups and amide I region. Understanding the

factors affecting astringency of whey proteins beverages manufactured at low pH can

help to develop strategies to improve consumer acceptance of these products.

4.2 Introduction

Astringency is the combination of complex sensations due to the perception of

shrin ing, drawing or puc ering of oral mucosa and muscles around the mouth (Lee and

Lawless, 1991 AST , ). Among foods that are astringent, Tannins (polyphenol

compounds) were found responsible in causing astringency most of the time in wine,

grape seeds, teas, fruits and vegetables. Tannins were able to interact and bind proline-

rich proteins and histidine-rich proteins in salivary proteins, followed by the aggregation

and precipitation of protein-tannin complex, and finally wea en the salivary’s ability

(Luc et al, 199 axter et al., 1997 Dinnella et al., 9).

Whey protein, a complex mixture of different proteins including α-lactalbumin, β-

lactoglobulin, bovine serum albumin, immunoglobulins, lactoferrin, lactoperoxidase,

glycomacropeptides and others (Laetitia . onnaillie and Peggy . Tomasula, 8),

was also recognized as an astringent agent but could not simply share the same

mechanism as tannins. ne comparable suggestion stated that the interactions between

the positively charged whey proteins (pH < pI) and negatively charged saliva proteins

could contribute to the astringency in acidic whey protein solutions ( eecher et al., 8

84
Vardhanabhuti et al., 1 ). Although efforts to explore the reasons behind astringency

triggered by acidic whey protein solutions have been made, the mechanism remains

unclear. ur ob ective was to evaluate the factors regulating astringency in whey protein

products by the statistical and instrumental methods.

4.3 Materials and methods

4.3.1 Whey protein sample

Two batches of beverage samples were prepared separately. ne batch included

different types of whey protein powder, including whey protein isolate (WPI, 3 different

samples), whey protein concentrate (WPC, 18) and whey protein hydrolysate (WPH, 1 )

were obtained from a leading beverage industry. After formulated from these powder

samples, they were controlled under uniform pH value (pH=3. ). The other batch

included 9 WPI powder, 6 WPC powder and WPH powder samples from different

manufacturers. After formulated from these powder samples, the beverages from every

powder sample were further ad usted to pH values ranging from . to 3.9. Each sample

was coded with 3 digit number and vacuum pac ed in glass bottle. everage samples

were maintained at ºC until spectra collection using FT-IR.

4.3.2 Astringency prediction

Quantitative sensory test on relevant whey protein beverage samples was

conducted at the Sensory Service Center in North Carolina State University, which is an

internationally recognized program in Dairy products. Sensory analysis was conducted in

compliance with institutional review board human sub ect regulations.

85
4.3.3 Fourier-Transform Infrared Spectroscopy (FT-IR)

Whey protein power samples were measured using an IR spectrometer (Excalibur

FTS 3 GX Varian, Palo Alto, CA) coupled with an attenuated total reflectance (ATR)

accessory. The optical bench included a dynamically aligned ichelson interferometer,

potassium bromide beam splitter, and a deuterated tri-glycine sulfate (DTGS) detector. IR

spectra were observed using commercial software (Win-IR Pro, Version . Varian Inc.)

from to 7 cm-1 with resolution of 8 cm-1, co-adding 6 scans to increase signal to

noise ratio. easurements were performed on a diamond crystal, with refractive index of
o
. and an incidence angle of yielding three internal reflections. The instrument was

calibrated before measurements using the crystal alone as bac ground. The powder

samples were pressed onto the ATR-IR crystal using a high-pressure clamp (PIKE Tech,

adison, WI), and replicate spectra were obtained.

Whey protein beverage samples were measured using FTIR-reflectance micro-

spectroscopy, the methodology was the same as mentioned in Chapter 3.

4.3.4 Statistical analysis

ultiple Linear Regression ( LR) attempts to model the relationship between

two or more explanatory variables (pH and whey protein types) and a response variable

(astringency) by fitting a linear equation to observed data. This line describes how the

mean response changes with the explanatory variables.

The astringency scores of whey protein powder samples were used to test the

ability to predict the sensory attributes based on their IR spectra in multivariate model.

ultivariate regression analysis (Partial least squares regression, PLSR) was performed

86
to build calibration models to simultaneously correlate the reference sensory score data to

the infrared spectral information. Partial Least Squares Regression (PLSR) was cross-

validated (leave-one-out approach) to generate calibration models. PLSR models were

evaluated in terms of loading vectors, standard error of calibration (SEC), standard error

of cross validation (SECV), coefficient of determination (r-value) and outlier diagnostics.

Pearson correlation was used to discover the correlation coefficient between the

concentrations of specific amino acids in whey protein powder and the astringency score

at fitted pH value of corresponding whey protein powder.

4.3.5 Gas Chromatography (GC) analysis

Whey protein powder was firstly acid hydrolyzed to yield free amino acids. mg

of each powder was mixed with 1ml of HCl (6N) in a sealed glass tube and placed in

oven at 11 ºC overnight to fully hydrolyze the protein. The resulting solution was blow-

dried with nitrogen gas and dissolved in 1ml of water (HPLC grade). This water solution

contained free amino acids yielding from whey protein powder and was analyzed by

following GC method with hours.

Amino acids hydrolyzed from whey protein powder samples were analyzed by

EZ:Faast™ amino acid analysis it (Phenomenex, Torrance, CA), which provided rapid

clean-up, derivatization and GC analysis of amino acids from complex mixtures. The

method was based on solid phase extraction (SPE), followed by a rapid derivatization

reaction and GC-Flame ionization detector (FID) analysis. The clean-up step using SPE

cartridges could save time on prior removal of lengthy proteins and interfering

substances, the next steps involved production of chloroformate derivatives of both the

87
amino and carboxylic acid groups, and the final step was GC-FID analysis. An Agilent

689 series (Santa Clara, CA) gas chromatograph equipped with an FID was used for

sample analysis. The chromatographic column with the dimensions 1 m* . mm i.d. of

a proprietary phase was provided with the it. The temperature programme of oven

started at 11 ºC, followed by a linear ramp of 3 ºC/min up to 3 ºC, and a hold time of

1 min. The carrier gas was helium at a rate of 1. ml/min at 1 ºC. The in ection using an

auto sampler was . ul and in ector temperature was ºC, while the detector was ept

at 3 ºC.

4.4 Results and discussion

The results of LR analysis used powder type and pH values as explanatory

variables to predict the corresponding astringency score was shown in table .1. While

simple linear regression using pH of whey protein beverages, as single explanatory

variable cannot sufficiently explain variance of beverage astringency (R ranging from

1 . % to 6. %), the incorporation of both variables by LR largely enhanced the

variance explained (R ranging from 71.8% to 9 . %). As indicated in table .1, the p

values associated with pH variable were all smaller than . . The statistical analysis

indicated that both pH and protein types played important roles in the development of

whey protein astringency. While protein types were under consideration, it was possibly

implying that astringency generated from acidic whey protein beverages could be

attributed to the protein secondary structure, amino acids compositions and all other

factors related to the whey protein types. oth figures in Figure .1 showed the normal

88
distribution of data set (WPI was used as an example), examining it was appropriate to

use LR to analyze the data.

(%)
Protein Types R P(regression) P (pH)
WPI 9 7 .3 . . 3
WPC 6 71.8 . 6 . 3
WPH 9 . . .
Table 4.1. Multiple Linear Regression using pH and whey protein types as predictor and
astringency scores as response for WPI, WPC, and WPH.

Figure 4.1. The examination of validity using MLR to analyze WPI samples. Residuals vs
Fits for Astringency (A) and Normal plot of Residuals for Astringency (B).
89
The relationship between pH values and astringency in acidic whey protein

beverages was further proved by PLSR analysis based on the IR spectra. As shown in

Figure . , strong correlation between pH values and IR spectra of acidic whey protein

beverages with various pH levels was observed in WPI, WPC and WPH models, which

could be confirmed by high R (> .97) and low standard errors (< .11) shown in Table

. . The wavenumber ranges used in all three models were from 1 to 18 cm-1, which

basically covered the ma or bands involved in protein structure. The most outstanding

bands in all loadings plots were amide I (~1 cm-1) and amide II (~16 cm-1) bands,

which indicated that pH was closely related to the alteration of protein’s secondary

structure, hence implied the relationship between pH and astringency.

PLSR model
WPI WPC WPH
Wavenumber range (cm-1) 1 1 -1796 1 -18 1 -18
Factor 6
Cumulative1 99.83 99.8 99.76
SECV .11 . 8 .1
R3 .97 .99 .98
SEC .1 . 8 . 9
Table 4.2. Calibration and cross-validation results of PLSR models developed by pH of
whey protein beverages using FTIR-Reflectance Micro-spectroscopy spectra.
1
Cumulative: Cumulative variances explained by the calibration model.
SECV: Standard error of cross-validated calibration model.
3
R: Coefficient of cross-validated calibration model.
SEC: Standard error of calibration model.

90
3.8 A. WPI B. WPI
0.2

3.4 0.1

3.0 -0.0

1454 1522
-0.1 1541 1578
2.6
1684
1649
-0.2
2.2
2.2 2.6 3.0 3.4 3.8 1468 1661
4.0
C. WPC D. WPC
0.2

3.6 0.1
IR Predicted pH

Loadings
3.2 -0.0

-0.1 1454
2.8 1541
1520
1647
-0.2 1626

2.5 3.0 3.5 1468 1661

4.0 E. WPH 0.2 F. WPH

0.1
3.5

-0.0
3.0
1499
-0.1 1539
1456 1518
2.5 1603 1657
-0.2 1638
2.6 3.0 3.4 3.8 1468 1661
Measured pH Wavenumber (cm-1)

Figure 4.2. PLSR plots and loadings based on the infrared spectra of whey protein WPI,
WPC and WPH beverage samples and measured pH using FTIR-Reflectance Micro-
spectroscopy.

To explore the effect of whey protein type on the development of astringency, a

regression analysis of each individual whey protein powder samples, correlating their

infrared spectra using FTIR-ATR with sensory astringency scores of corresponding

beverages at fitted pH (3. ) was done for WPI, WPC, and WPH separately. y

91
controlling one factor (pH) at a certain level, it might be possible to unmas other

mechanisms behind astringency. The regression plots and loadings were showing in

Figure .3, and the statistics performance of these data were also generated and showed

in Table .3.

Figure 4.3. PLSR plots and loadings based on the infrared spectra of whey protein
powder samples (WPI, WPC, and WPH).
92
PLSR model
WPI WPC WPH
Factor 6 7 8
Cumulative 99.3 93. 99.
SECV .19 .3 . 8
R .87 .7 .99
SEC .17 . .
Table 4.3. Calibration and cross-validation statistic performance of PLSR models
developed by using whey protein powder samples.

The regions that gave the lowest SECV were chosen for developing cross-

validated PLSR calibration models illustrated in Figure .3 and Table .3, which included

the frequency ranges of 18 -9 cm-1 and provided the best predictive ability. SECV is

an estimate of the standard error of prediction (magnitude of error expected when

independent samples are predicted using the model) and the model producing the

minimum SECV is selected as the best model for the spectral data set. Within this

experiment, WPH had the lowest SECV value, whereas WPC showed the highest error

for prediction.

The cumulative variance gives the proportion of variability of the property that is

described by the model. ost of the variance were explained in calibration models

(>93. %), while reasonable factor numbers were associated (<8). The numbers of factors

should not be too high, in that way too much structural noise will be incorporated in

building the model, and generated unreliable data.

Among all of the whey protein output, WPH generated the best calibration

models, while WPC did not show results as good as others, it had relatively low

cumulative variance (93. ), high SECV ( .3 ), and also a low R value ( .7 ). The reason
93
why this happened was probably that WPC may contain various protein levels, with all

other component such as high amount of lactose not removed. The complexity of the

matrix were possibly bringing signal interference and decreasing the ability to resolve the

target signal. Nevertheless, the ability of predicting astringency by direct analysis of the

whey protein powders provided unique opportunity for rapid assessment of the quality of

the incoming material as related to their impact on astringency.

Figure .3 also showed the PLSR loadings for WPI, WPC and WPH powder

samples. The most important bands explaining the astringency scores were in the 1396,

1 66, 1 97 and 16 -17 cm-1 region. 1396 cm-1 was associated the stretching of C=

in the C - groups (Coates, ), 1 97 cm-1 was assigned to the vibrations of C -

+ -
bound with Ca and the C antisymmetric stretching vibrations of the glutamyl side-

chain give rise to a band at 1 67 cm-1 ( izuguchi et al., 1997). The broad bands between

16 -17 cm-1 were associated with amide I region (Kong and Yu, 7). These signals

were related to modifications occurring in the secondary structure of the native protein

possibly attributed to calcium binding with acidic amino acids in the β-sheet

conformation.

Since carbonyl groups in amino acids side chains were suggested to be important

in the construction of correlation between whey protein powder’s IR spectra and their

corresponding astringency, a GC analysis to quantify the amino acids with carbonyl side

chains in whey protein powders and Pearson correlation was used to exam the connection

between the amino acids concentration with carbonyl side chains in whey protein

powders and the corresponding astringency of beverages at pH 3. . The results of

94
Pearson correlation were shown in Table . .

Pearson correlation P
GLU .8 . 3
ASP .6 8 . 9
ASP+GLU .78 . 13
Table 4.4. Pearson correlation between concentrations of glutamic acid (GLU) and
aspartic acid (ASP) in whey protein powders and the corresponding astringency of
beverages at pH 3.5.

In GC analysis, the selected whey protein powders, soy protein powder, and

gelatin powder were first acid hydrolyzed to yield free amino acids. Soy protein and

gelatin were added to improve the sampling in Pearson correlation, because soy protein

was nown to be very astringent and gelatin was recognized as low astringent. After

digestion, both glutamine and glutamic acid in protein were determined together because

glutamine was deaminated to form glutamic acid similarly asparagine and aspartic acid

were determined together. The resulting concentrations of GLU and ASP were analyzed

by Pearson correlation. The highest correlation coefficient was generated from GLU

concentration and astringency ( .8 ), and the lowest one was from ASP and astringency

( .6 8). However all three coefficients suggested the possibility of lin age between

amino acids with carbonyl side chains in whey protein powders and their corresponding

astringency. The GC analysis and Pearson correlation supported that whey protein itself

was responsible for the astringency and amino acids with carbonyl side chains were

possibly contributing to the development of astringency.


95
4.5 Conclusion

verall, both pH and protein types played important roles in the development of

whey protein astringency. FT-IR was able to monitor protein properties as related to

astringency development, acidic amino acids (Glu and Asp) and the β–sheet protein

conformation were pointed out as important explanation of astringency prediction.

Pearson correlation and PLSR analysis illustrated that carbonyl groups and secondary

conformation of denatured whey proteins were associated with the astringency, which

was match the results from our LR analysis.

4.6 Significance and future work

y the combination of infrared spectroscopy with pattern recognition technique,

we found the unique bands from commercial whey protein powder to authenticate whey

protein, not only based on the protein levels but also the possible processing techniques.

The authentication of raw protein ingredient can ensure desired functionalities in various

food applications and finally benefit food industry with a better quality.

With the comprehensive information obtained from food samples infrared spectra,

the PLSR analysis showed relationship between the reference value and the predicted

value, indicating the high possibility to predict the perceived astringency induced by

whey protein compounds in acidic beverages based on the detection of specific functional

groups.

ased on 6 whey protein beverage samples with fixed pH, our results showed

that the protein types had obvious impact on the perceived astringency. Then statistical

96
analysis ( LR and Pearson correlation) further indicated that both pH and protein types

were important explanatory variables to predict astringency in acidic whey protein

beverages (variance explained was larger than 71.8%), which matched the results of

important bands resolved from PLSR models of beverage samples with various pH.

However introducing other food ingredients such as citric acid into acidic whey

protein beverages might interfere with the bands associated to astringency. And the

complex matrix of whey protein, including minerals, lactose, lactic acid, and others might

also influence the signal perceived as infrared spectra. Future wor can be done to

explore the mechanism of astringency triggered by whey protein. The change of protein’s

secondary conformation and the location of amino acids with carbonyl side chains in

whey protein can be possible factors related to astringency. We can also evaluate the

effect of mas ing agents to decrease the perceived astringency in acidic whey protein

beverages. Possible mas ing agents can be tested are hydrocolloid, sugar, salt and the

combination of them.

97
List of References

Adler-Nissen, J. 1986. Enzymatic Hydrolysis of Food Proteins. Elsevier Applied Science


Publishers, Barking, Essex, London,

Albano, C., W. Dunn III, U. Edlund, E. Johansson, B. Norde´n, M. Sjöström, and S.


Wold. 1978. Four levels of pattern recognition. Anal. Chim. Acta. 103:429– 443.

Allain, A. F., P. Paquin, and M. Subirade. 1999. Relationships between conformation of


β-lactoglobulin in solution and gel states as revealed by attenuated total reflection
Fourier transform infrared spectroscopy. International journal of biological
macromolecules. 26:337-344.

Arnold, R. A., A. C. Noble, and V. L. Singleton. 1980. Bitterness and astringency of


phenolic fractions in wine. J. Agric. Food Chem. 28:675–678.

Asselin, J., J. Amiot, S. F. Gauthier, W. Mourad, and J. Hebert. 1988. Immunogenicity


and allergenicity of whey protein hydrolysates. J. Food Sci. 53:1208-1211.

ASTM. 2004. Standard definitions of terms relating to sensory evaluation of materials


and products. Annual book of ASTM standards. American Society for Testing
and Materials, Philadelphia.

Atra, R., G. Vatai, E. Bekassy-Molnar, and A. Balint. 2005. Investigation of ultra-and


nanofiltration for utilization of whey protein and lactose. Journal of food
engineering. 67:325-332.

Baer, R. J., J. F. Frank, M. Loewenstein, and G. S. Birth. 1983. Compositional analysis of


whey powders using near infrared diffuse reflectance spectroscopy. J. Food Sci.
48:959-961.

98
Bajec, M.R., and G. Pickering. 2008. Astringency: Mechanisms and perception. Crit.
Rev. Food Sci. Nutr. 48:858-875.

Bakker, J. 1998. Astringency: A matter of taste. Biologist. 45:104–107.

Baldasso, C., T. C. Barros, and I. C. Tessaro. 2011. Concentration and purification of


whey proteins by ultrafiltration. Desalination. 278:381-386.

Barnes, R. J., M. S. Dhanoa, and S. J. Lister. 1989. Standard normal variate


transformation and de-trending of near-infrared diffuse reflectance spectra.
Applied spectroscopy. 43:772-777.

Barth, A. 2000. The infrared absorption of amino acid side chains. Progress in Biophysics
& Molecular Biology. 74:141–173.

ate-Smith, E. C. 19 . Astringency in food. Food Processing and Pac aging. 3: 1 -


13 .

Bate-Smith, E. C. 1973. Haemanalysis of tannins: the concept of relative astringency.


Phytochem. 12:907–912.

axter, N. ., T. H. Lilley, E. Haslam, and . P. Williamson. 1997. ultiple interactions


between polyphenols and a salivary proline-rich protein repeat result in
complexation and precipitation. iochemistry. 36: 66- 77.

eecher, . W., . A. Dra e, P. . Luc , and E. A. Foegeding. 8. Factors regulating


astringency of whey protein beverages. . Dairy Sci. 91: 3- 6 .

Bennick, A. 2002. Interaction of plant polyphenols with salivary proteins. Crit. Rev. Oral
Biol. Med. 13:184–196.

Bonnaillie L. M. and P. M. Tomasula. 2008. Whey Protein Fractionation. Pages 15-38 in


Whey Processing, Functionality and Health Benefits. C. Onwulata and P. Huth,
ed. Wiley-Blackwell, Ames, Iowa.
99
Booij, L., W. Merens, C Markus, and A. W. van der Does. 6. Diet rich in α-
lactalbumin improves memory in unmedicated recovered depressed patients and
matched controls. Journal of Psychopharmacology. 20:526–535.

Bouaouina, H., A. Desraumax, C. Loisel, and J. Legrand. 2006. Functional properties of


whey proteins as affected by dynamic high-pressure treatment. International Dairy
Journal. 16:275–284.

Boye, J. I., I. Alli, and A. A. Ismail. 1996. Interactions involved in the gelation of bovine
serum albumin. Journal of Agricultural and Food Chemistry. 44:996-1004.

Branden, K. V., and M. Hubert. 2005. Robust classification in high dimensions based on
the SIMCA Method. Chemometrics and Intelligent Laboratory Systems. 79:10-
21.

Brandenburg, A. H., C. V. Morr, and C. L. Welter. 1992. Gelation of commercial whey


protein concentrates: effect of removal of low-molecular-weight components. J.
Food Sci. 57:427-432.

Breslin, P. A., M. Gilmore, G. K. Beauchamp, and B. G. Green. 1993. Psychophysical


evidence that oral astringency is a tactile sensation. Chem. Senses. 18:405-417.

Bro, R., K. Kjeldahl, A. K. Smilde, and H. A. L. Kiers. 2008. Cross-validation of


component models: A critical look at current methods. Anal Bioanal Chem.
390:1241–1251.

Brody, E. P. 2000. Biological activities of bovine glycomacropeptide. British Journal of


Nutrition. 84:39-46.

Cala, O., N. Pinaud, C. Simon, E. Fouquet, M. Laguerre, E. J. dufourc, and I. Pianet.


2010. NMR and molecular modeling of wine tannins binding to saliva proteins:
revisiting astringency from molecular and colloidal prospects. FASEB J. 24:1-10.

100
Candolfi, A., R. De Maesschalck, D. Jouan-Rimbaud, P. A. Hailey, and D. L. Massart.
1999. The influence of data pre-processing in the pattern recognition of excipients
near-infrared spectra. J. Pharm. Biomed. Anal. 21:115–132.

Canon, F., F. Pat , E. eudec, T. arlin, V. Cheynier, A. Giuliani, and P. Sarni-


Manchado. 2009. Characterization, stoichiometry, and stability of salivary
protein-tannin complexes by ESI-MS and ESI-MS/MS. Anal. Bioanal. Chem.
395:2535-2545.

Charlton, A. J., N. J., Baxter, T. H. Lilley, E. Haslam, C. J. McDonald, and M. P.


Williamson. 1996. Tannin interactions with a full-length human salivary proline-
rich protein display a stronger affinity than with single proline-rich repeats. FEBS
Lett. 382:289-292.

Charlton, A., N. Baxter, M. Khan, A. Moir, E. Haslam, A. Davies, and M. Williamson.


2002. Polyphenol/peptide binding and precipitation. J. Agric. Food Chem.
50:1593-1601.

Chegini, G., and M. Taheri. 2013. Whey Powder: Process Technology and Physical
Properties: A Review. Middle East Journal of Scientific Research. 13: 1377-1387.

Clemente, A. 2000. Enzymatic protein hydrolysates in human nutrition. Trends Food Sci.
Technol. 11:254-262.

Clifford, . N. 1996. Astringency. Proceedings of the Phytochemical Society of Europe.


1: 87–1 7.

Coates, J. 2000. Interpretation of infrared spectra, a practical approach. Pages 10815–


10837 in Encyclopedia of Analytical Chemistry. R.A. Meyers, ed. John Wiley &
Sons Ltd, Chichester, NY.

Condelli, N., C. Dinnella, A. Cerone, E. Monteleone, and M. Bertuccioli. 2006.


Prediction of perceived astringency induced by phenolic compounds II: Criteria
for panel selection and preliminary application on wine samples. Food Qual.
Prefer. 17:96-107.

101
Cordle, C. T. 1994. Control of food allergies using protein hydrolysates. Food Technol.
48:72-76.

Creamer, L.K., D. A. D. Parry, and G. N. Malcolm. 1983. Secondary structure of bovine


β-lactoglobulin B. Archives of Biochemistry and Biophysics. 227:98–105.

Croft, A. K. and M. K. Foley. 2008. Proline-rich proteins-deriving a basis for residue-


based selectivity in polyphenolic binding. Org. Biomol. Chem. 6:1594-1600.

Curley, D.; Kumosinski, T.; Unrah, J.; Farrell, H. Changes in the secondary structure of
bovine casein by Fourier transform infrared spectroscopy: effects of calcium and
temperature. J. Dairy Sci. 1998, 81, 3154-3162.

Damodaran, S. 1989. Interrelationship of molecular and functional properties of food


proteins. Pages 6-43 in Food proteins. J. E. Kinsella and W. Soucie, ed. Am. Oil
Chem. Soc.

Dangles, O., and C. Dufour. 2006. Flavonoid-protein interactions. Pages 443-470 in


Flavonoids-Chemistry, Biochemistry and Applications; O. M. Andersen, K. R.
Markham, ed. Taylor & Francis, Boca Raton, FL, USA,

de Freitas, O., G. J. Padovan, L. Vilela, J. E. Dos Santos, J. E. Dutra de Oliviera, and L. J.


Greene. 1993. Characterization of protein hydrolysates prepared for enteral
nutrition. J. Agric. Food Chem. 41:1432-1438.

de Freitas, V., and N. Mateus. 2001. Structural features of procyanidin interactions with
salivary proteins. J Agric. Food Chem. 49:940-945.

De Marchi, M., V. Bonfatti, A. Cecchinato, G. Di Martino, and P. Carnier. 2010.


Prediction of protein composition of individual cow milk using mid-infrared
spectroscopy. Italian Journal of Animal Science. 8:399-401.

Demiglio, P., and G. J. Pickering, 2008. The influence of ethanol and pH on the taste and
mouthfeel sensations elicited by red wine. J. Food Agric. Environ. 6:143-150.

102
Dinnella, C., A. Recchia, G. Fia, M. Bertuccioli, and E. Monteleone. 2009. Saliva
characteristics and individual sensitivity to phenolic astringent stimuli. Chem.
Senses. 34:295-304.

Dionysius, D. A., P. A. Grieve, and J. M. Milne. 1993. Forms of lactoferrin: their


antibacterial effect on enterotoxigenic Escherichia coli. J. Dairy Sci. 76:2597-
2606.

Dissanayake, M., A.L. Kelly, and T. Vasiljevic. 2010. Gelling properties of


microparticulated whey proteins. J. Agric. Food Chem. 58:6825–6832.

Doultani, S., K. N. Turhan, and M. R. Etzel. 2003. Whey protein isolate and
glycomacropeptide recovery from whey using ion exchange chromatography.
Journal of food science 68:1389-1395.

Dupuy, N., L. Duponchel, J. P. Huvenne, B. Sombret, and P. Legrand. 1996.


Classification of edible fats and oils by principal component analysis of Fourier
transform infrared spectra. Food Chem. 57:245-251.

Elgar, D. F., C. S. Norris, J. S. Ayers, M. Pritchard, D. E. Otter, and K. P. Palmanoa.


2000. Simultaneous separation and quantitation of the major bovine whey
proteins including proteose peptone and caseinomacropeptide by reversed-phase
high-performance liquid chromatography on polystyrene-divinylbenzene. Journal
of Chromatography A. 878:183–196.

Etzel, M. R. 2004. Manufacture and use of dairy protein fractions. J.Nutr. 134:996S–
1002S.

Etzion, Y., R. Linker, U. Cogan, and I. Shmulevich. 2004. Determination of protein


concentration in raw milk by mid-infrared Fourier transform infrared/attenuated
total reflectance spectroscopy. Journal of dairy science. 87:2779-2788.

Fabian,H., T. Yuan, H.J. Vogel, and H.H. Mantsch. 1996. Comparative analysis of the
amino- and carboxy-terminal domains of calmodulin by Fourier transform
infrared spectroscopy. Eur. Biophys. J. 24:195–201.

103
Farrell, P. M., E. H. Mischler, S. A. Sondel, and M. Palta. 1987. Predigested formula for
infants with cystic fibrosis. J. Am. Dietetic Assoc. 87:321-328.

Farrell Jr., H.M., E. D. Wickham, J. J. Unruh, P.X. Qi, and P.D. Hoagland. 2001.
Secondary structural studies of bovine caseins: temperature dependence of b-
casein structure as analyzed by circular dichroism
 and FT-IR spectroscopy and
correlation with micellization. Food Hydrocolloids. 15:341-354.

Finch, J. N., and E. J. Lippincott. 1956. Hydrogen bond systems: Temperature


dependence of OH frequency shifts and OH band intensities. Chem. Phys.,
24:908-909.

Foegeding, E. A. 1989. Molecular properties and functionality of proteins in food gels.


Pages 94-185 in Food proteins, J. E. Kinsella and W. Soucie, ed. Am. Oil Chem.
Soc.

Foegeding, E. A., J. P. Davis, D. Doucet, and M. K. McGuffey. 2002. Advances in


modifying and understanding whey protein functionality. Trends in Food Science
& Technology. 13:151–159.

Foegeding, E. A., and P. J. Luck. 2011. Whey protein products. Pages 873-878 in
Encyclopedia of Dairy Sciences. H. Roginski, P.F. Fox, and J.W. Fuquay, ed.
Academic Press, London, UK.

Fontoin, H., C. Saucier, P. -L. Teissedre, and Y. Glories. 2008. Effect of pH, ethanol and
acidity on astringency and bitterness of grape seed tannin oligomers in model
wine solution. Food Qual. Prefer. 19:286-291.

Fox, P.F. 2003. The major constituents of milk. Pages 5-38 in Dairy processing, G.
Smith, ed. CRC Press, Boca Raton, FL.

Frokjaer, S. 1994. Use of hydrolysates for protein supplementation. Food Technol. 48:86-
88.

104
Fu, F. N., D. B. DeOliveira, W. R. Trumble, H. K. Sarkar, and B. R. Singh. 1994.
Secondary Structure estimation of proteins using the amide III region of fourier
transform infrared spectroscopy: application to analyze calcium-binding-induced
structural changes in calsequestrin. Applied Spectroscopy. 48:1432-1441.

Galani, D., and R. K. Owusu Apenten. 1999. Heat-induced denaturation and aggregation
of β-lactoglobulin: kinetics of formation of hydrophobic and disulphide-linked
aggregates. Int. J. Food Sci. and Techn. 34:467-476.

Gawel, R., A. Oberholster, and I. L. Francis. 2000. A mouth-feel wheel: terminology for
communicating the mouth-feel properties of red wine. AJGWR. 6:203–207.

Gerwert, K., B. Hess, J. Soppa, and D. Oesterhelt. 1989. Role of aspartate-96 in proton
translocation by bacteriorhodopsin. Proceedings of the National Academy of
Sciences, 86:4943-4947.

Gillies, M. T. 1974. Whey Processing and Utilization. Pages 24-31 in Economic and
technical aspects. No. 19. Noyes Data Corp., Park Ridge, NJ.

Green, B. 1993. Oral astringency: A tactile component of flavor. Acta Psychologia.


84:119–125.

Griffiths, P. R., and J. A. de Haseth. 1986. Fourier Transform Infrared Spectroscopy.


John Wiley & Sons, New York.

Haaland, D. M., and E. V. Thomas. 1988. Partial least-squares methods for spectral
analyses. 1. Relation to other quantitative calibration methods and the extraction
of qualitative information. Analytical Chemistry. 60:1193-1202.

Hagerman A. E, and L. G. Butler. 1978. Protein precipitation method for the quantitative
determination of tannins. J. Agric. Food Chem. 26:809-812.

Hagerman, A. E., M. E. Rice, and N. T. Ritchard, 1998. Mechanisms of protein


precipitation for two tannins, pentagalloyl glucose and epicatechin16 (4 → 8)
catechin (procyanidin) J. Agric. Food Chem. 46:2590-2595.
105
Hartwig, P., and M. R. McDaniel. 1995. Flavor characteristics of lactic, malic, citric, and
acetic acids at various pH levels. J. Food Sci. 60:384–388.

Hashimoto, A., and T. Kameoka. 2008. Applications of infrared spectroscopy to


biochemical, food, and agricultural processes. Appl. Spectrosc. ReV., 43:416.

Haslam, E., T. H. Lilley, and L. G. utler. 1988. Natural astringency in food stuffs-a
molecular interpretation. Critical Reviews in Food Science & Nutrition. 7: 1- .

Haslam E, T. H. Lilley, E. Warminski, H. Liao, Y. Cai, and R. Martin. 1991. Polyphenol


complexation. A study in molecular recognition. Pages 8-49 in Phenolic
compounds in food and their effects on health. I. Analysis, occurrence and
chemistry. C. T. Ho, C. Y. Lee, M.T. Huang, ed. Am Chem Soc., Washington,
DC.

Heine, W. E., P. D. Klein and P. J. Reeds. 1991. The Importance of α-Lactalbumin in


Infant Nutrition. The Journal of Nutrition. 121:227-283.

Hill, R. L. and K. Brew. 1975. Lactose synthetase. Adv. Enzymol. Relat. Areas Mol.
Biol. 43:411-490.

Hinrichs, J. 2001. Incorporation of whey proteins in cheese. International Dairy Journal.


11:495-503.

Hruschka, W. R. 2001. Data analysis: wavelength selection methods. Pages 39–58 in


Near-infrared technology in the agricultural and food industries. P. Williams and
K. Norris, ed. American Association of Cereal Chemists, Inc., St. Paul, MN.

Huang, Y., B. A. Rasco, and A. G. Cavinato, 2008. Enzymic Technique: Enzyme-Linked


Immunosorbent Assay (ELISA). Pages 477- 510 in Modern Techniques for Food
Authentication. D. W. Sun, ed. Elsevier, New York.

Huffman, L.M. 1996. Processing whey protein for use as a food ingredient. Food
Technology. 50:49–52.

106
Hyde, R. J., and R. M. Pangborn. 1978. Parotid salivation in response to tasting wine.
Am. J. Enol. Vitic. 29:87–91.

Infometrix, Inc. 2008. Classification methods. Pirouette, multivariate data analysis


version . , user’s manual. Infometrix, Inc., othell, WA.

Jelen, P. 1979. Industrial whey processing technology: An overview. Journal of


Agricultural and Food Chemistry, 27:658-661.

Jelen, P. 2011. Whey processing. Pages 731-737 in Encyclopedia of Dairy Sciences. H.


Roginski, P.F. Fox, and J.W. Fuquay, ed. Academic Press, London, UK.

Jellinek, G. 1985. Sensory Evaluation of Food: Theory and Practice. Ellis Harwood,
Chichester.

bstl, E., . ’Connell, . P. A. Fairclough, and . P. Williamson. 2004. Molecular


model for astringency produced by polyphenol/protein interactions.
Biomacromolecules. 5:942-949.

Joslyn, M. A. and J. L. Goldstein. 1964. Astringency of fruits and fruit products in


relation to phenolic content. Adv. Food Res. 13:179–217.

ovanovi , S., . ara , and . a e . . Whey proteins-Properties and Possibility of


Application. Mljekarstvo. 55:215-233.

Kallithraka, S., J. Bakker, and M. N. Clifford. 1997. Effect of pH on astringency in model


solutions and wines. J. Agric. Food Chem. 45:2211-2216.

Kallithra a, S., . a er, and . N. Clifford. 1998. Evidence that salivary proteins are
involved in astringency. ournal of Sensory Studies. 13: 9- 3.

107
Karoui, R., G. Downey, and C. Blecker. 2010. Mid-Infrared Spectroscopy Coupled with
Chemometrics: A Tool for the Analysis of Intact Food Systems and the
Exploration of Their Molecular Structure-Quality Relationships-A Review.
Chemical reviews. 110:6144-6168.

Kartheek, M., A. A. Smith, A. K. Muthu, and R. Manavalan. 2011. Determination of


adulterants in food: A review. Journal of Chemical and Pharmaceutical Research.
3:629–636.
Keast, R. S. J. 2003. The effect of zinc on human taste perception. J. Food Sci. 68:1871-
1877.

Kelly, M., B. Vardhanabhuti, P. Luck, M. A. Drake, J. Osborne, and E. A. Foegeding.


2010. Role of protein concentration and protein–saliva interactions in the
astringency of whey proteins at low pH. Journal of dairy science. 93:1900-1909.

Kessler, H. G. 2002. Food and bioprocess engineering. Pages 56–96 in Dairy


Technology. Verlag A. Kessler, München. German.

Khalil-Manesh, R., C. Agness, and H. C. Gonick. 1989. Aluminum- binding protein in


dialysis dementia. I. Characterization in plasma by gel chromatography and
electrophoresis. Nephron. 52:323-328.

Kilara, A., and M.N. Vaghela. 2004. Whey proteins. Pages 72–99 in Proteins in Food
Processing, R.Y. Yada ed. Woodhead Publishing, Cambridge, England.

Kilara, A. 2008. Whey and whey products. Pages 337-355 in Dairy Processing and
Quality Assurance. R. C. Chandan, A. Kilara, and N. P. Shah, ed. Wiley-
Blackwell, Ames, Iowa.

Kinghorn, N. M., C. S. Norris, G. R. Paterson, and D. E. Otter. 1995. Comparison of


capillary electrophoresis with traditional methods to analyse bovine whey
proteins. Journal of Chromatography A. 700:111-123.

Kinsella, J. E., P. F. Fox, and L. B. Rockland. 1986. Water sorption by proteins: milk and
whey proteins. Critical Reviews in Food Science & Nutrition. 24:91-139.

108
Kinsella, J. E., and D. M. Whitehead. 1989. Proteins in whey: chemical, physical, and
functional properties. Adv. Food Nutr. Res. 33:437-438.

Kong, J., and S. Yu. 2007. Fourier transform infrared spectroscopic analysis of protein
secondary structures. Acta Biochim Biophys Sin. 39:549–559.

Krimm S., and J. Bandekar. 1986. Vibrational spectroscopy and conformation of


peptides, polypeptides, and proteins. Pages 181-364 in Advances in Protein
Chemistry. Vol. 38. C. B. Anfinsen, J. T. Edsall, and F. M. Richards, ed.
Academic Press, Orlando, FL.

Kuehl, D., and R. Crocombe. 1984. The quantitative analysis of a model fermentation
broth. Appl. Spectrosc., 38:907-909.

Kuhn, P. R., and E. A. Foegeding. 1991. Mineral salt effects on whey protein gelation.
Journal of Agricultural and Food Chemistry. 39:1013-1016.

Kuntz, I. D., 1971. Hydration of macromolecules. III. Hydration of polypeptides. J. Am.


Soc, 93, 514-516.

Kuntz, I. D. and W. Kauzman, 1974. Hydration of proteins and polypeptides. Adv.


Protein Chem. 28:239-345.

Kvalheim, O. M., and T.V. Kratang. 1992. SIMCA - Classification by means of disjoint
cross validates principal components models. In Multivariate pattern recognition
in chemometrics. R. G. Brereton, ed. Elsevier, Amsterdam, NY.

Lawless, H.T., C. J. Corrigan, and C. B. Lee. 1994. Interactions of astringent substances.


Chem. Senses. 19:141-154.

Lawless, H. T., J. Horne, and P. Giasi. 1996. Astringency of organic acids is related to
pH. Chem. Senses. 21:397–403.

109
Lawless, H. T., S. Schlake, J. Smythe, , J. Lim, H. Yang, K. Chapman, and B. Bolton.
2004. Metallic taste and retronasal smell. Chem. Senses. 29:25– 33.

Lee, C. T., and H. T. Lawless. 1991. Time-course of astringent sensations. Chem. Senses
16:225–238.

Lee, S., C.V. Morr, and E.Y.W. Ha. 1992. Structural and functional properties of
caseinate and whey protein isolate as affected by temperature and pH. J. Food Sci.
57:1210–1214.

Lee, C. A., and Z. . Vic ers. 8. The astringency of whey protein beverages is caused
by their acidity. International dairy ournal. 18:11 3-11 6.

Lesschaeve, I. and A. C. Noble. 2005. Polyphenols: factors influencing their sensory


properties and their effects on food and beverage preferences. Am. J. Clin. Nutr.
81: 330S-335S.

Libnau, F. O., O. M. Kvalheim, A. A. Christy, and J. Toft. 1994. Spectra of water in the
near-and mid-infrared region. Vibrational spectroscopy. 7:243-254.

Lim, J., and H. T. Lawless. 2005. Qualitative differences of divalent salts:


multidimensional scaling and cluster analysis. Chem. Senses. 30:719–726.

Lin, J., and C. W. Brown. 1992. Near-IR spectroscopic determination of NaCl in aqueous
solution. Appl. Spectrosc. 46:1809.

Luck, G., H. Liao, N. J. Murray, H. R. Grimmer, E. E. Warminski, M. P. Williamson, T.


H. Lilley, E. Haslam. 1994. Polyphenols, astringency and proline-rich proteins.
Phytochemistry. 37:357-371.

Mangino, M. E. 1984. Physicochemical aspects of whey protein functionality. Journal of


dairy science 67:2711-2722.

Mangino, M. E. 1992. Gelation of whey protein concentrates. Food Technol. 46:114.


110
Marchi, M. D., V. Bonfatti, A. Cecchinato, G. D. Martino, and P. Carnier. 2009.
Prediction of protein composition of individual cow milk using mid-infrared
spectroscopy. Ital. J. anim. Sci. 8:399-401.

ar us, C. R., . livier, and E. HF. de Haan. . Whey protein rich in α-lactalbumin
increases the ratio of plasma tryptophan to the sum of the other large neutral
amino acids and improves cognitive performance in stress-vulnerable subjects.
The American journal of clinical nutrition 75:1051-1056.

Martin, S., and R. M. Pangborn. 1971. Human parotid secretion in response to ethyl
alcohol. J. Dent. Res. 50:485–490.

Martín-Diana, A. B., M. C. Gomez-Guill´en, P. Montero, and J. Fontecha. 2006. Vis-


coelastic properties of caseinmacropeptide isolated from cow, ewe and goat
cheesewhey. J. Sci. Food Agric. 86:1340-1349.

Mascarenhas, M., J. Dighton, and G. A. Arbuckle. 2000. Characterization of plant


carbohydrates and changes in leaf carbohydrate chemistry due to chemical and
enzymatic degradation measured by microscopic ATR FT-IR spectroscopy. Appl.
Spectrosc. 54:681-686.

McClure, W. F., and D. L. Standfield. 2002. Pages 212 in Handbook of Vibrational


Spectroscopy. John Wiley & Sons Ltd., Chichester.

McRae, J. M., and J. A. Kennedy. 2011. Wine and grape tannin interactions with salivary
proteins and their impact on astringency: a review of current research. Molecules.
16:2348-2364.

Mendenhall, I. V., and R. J. Brown. 1991. Fourier transform infrared determination of


whey powder in nonfat dry milk. Journal of dairy science. 74:2896-2900.

eza- rquez, . G., T. Gallardo-Vel zquez, G. sorio-Revilla. 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. 86:511–519.

111
Michell, A. J., and L. R. Schimleck, 1996. NIR spectroscopy of woods from Eucalyptus
globulus. Appita. J. 49:23-26.

iller, G. 7. A sporting attitude. Prepared Foods. 176:91–9 .

Mizuguchi, M., M. Nara, K. Kawano, and K. Nitta,1997. FT-IR study of the calcium-
binding to bovine a-lactalbumin. Relationships between the type of coordination
and characteristics of the bands due to the Asp carboxylate groups in the calcium
binding site. FEBS Lett. 417:153–156.

Morr, C.V. 1989. Whey proteins: Manufacture. Pages 245–284 in Developments in Dairy
Chemistry-4: Functional Milk Proteins, P. F. Fox, ed. Elsevier Applied Science
publisher, New York.

Morr, C. V., and E. Y. W. Ha. 1993. Whey protein concentrates and isolates: Processing
and functional properties. Critical Reviews in Food Science and Nutrition.
33:431-476.

Mulvihill, D. M., and J. E. Kinsella, 1988. Gelation of l β-actoglobulin: effects of sodium


chloride and calcium chloride on the rheological and structural properties of gels.
J. Food Sci. 53:231- 236.

Murayama, K., and M. Tomida. 2004. Heat-induced secondary structure and


conformation change of bovine serum albumin investigated by Fourier transform
infrared spectroscopy. Biochemistry. 43:11526-11532.

Nakano, T., N. Ikawa, and L. Ozimek. 2006. Separation of glycomacropeptide from


sweet whey by using chitosan and a centrifugal filter. Milchwissenschaft, 61:191-
193.

Ngarize, S., H. Herman, A. Adams, and N. Howell. 2004. Comparison of changes in the
secondary structure of unheated, heated, and high-pressure-treated β-lactoglobulin
and ovalbumin proteins using Fourier transform Raman spectroscopy and self-
deconvolution. Journal of agricultural and food chemistry. 52:6470-6477.

112
Nicorescu, I., C. Loisel, A. Riaublanc, C. Vial, G. Djelveh, G. Cuvelier, and J. Legrand.
2009. Effect of dynamic heat treatment on the physical properties of whey protein
foams. Food hydrocolloids, 23:1209-1219.

Nielsen, P. M. 1997. Functionality of protein hydrolysates. Pages 443-472 in Food


proteins and their applications. S. Damadoran, and A. Paraf, ed. Marcel Dekker,
New York.

Noguchi, H. 1981. Hydration around hydrophobic groups. Pages 281-293 in Water


Activity: Influence on Food Quality, L. Rockland, and G. Stewart, ed. Academic
Press, New York.

Peleg, H., K. K. Bodine, and A. C. Noble. 1998. The influence of acid on astringency of
alum and phenolic compounds. Chemical senses. 23:371-378.

Peleg, H., and A. C. Noble. 1999a. Effect of viscosity, temperature and pH on


astringency in cranberry juice. Food quality and preference. 10:343-347.

Peleg, H., K. Gacon, P. Schlich, and A. Noble. 1999b. Bitterness and astringency of
flavon-3-ol monomers, dimers and trimers. J. Sci. Food Agric. 79:1123–1128.

Permya ov, E. A., and L. . erliner. . α-Lactalbumin: structure and function. FEBS
letters. 473:269-274.

Petty, S. A., T. Adalsteinsson, and S. M. Decatur. 2005. Correlations among morphology,


β-sheet stability, and molecular structure in prion peptide aggregates.
Biochemistry. 44:4720-4726.

Phillips, L. G., W. Schulman, and J. E. Kinsella. 1990. pH and heat treatment effects on
foaming of whey protein isolate. Journal of Food Science 55:1116-1119.

Phillips, L. G., S. T. Yang, and J. E. Kinsella. 1991. Neutral salt effects on stability of
whey protein isolate foams. J. Food Sci. 56:588-589.

113
Pout-El, A. 1981. Protein functionality: Classification, definition and methodology.
Protein functionality in foods. J. R. Cherry, ed. Am. Chem. Soc., Washington,
DC.

Richert, S. H., C. Morr, and C. M. Cooney. 1974. Effect of heat and other factors upon
foaming properties of whey protein concentrates. Journal of Food Science. 39:42-
48.
Rinnan, A., F. van den Berg, and S. B. Engelsen. 2009a. Review of the most common
pre-processing techniques for near-infrared spectra. Trends in Analytical
Chemistry. 28:1201-1222.

Rinnan, Å, L. Nørgaard, F. van den erg, .s Thygesen, R. ro, and S. . Engelsen.
9b. Data Pre-processing. Pages 9- in Infrared Spectroscopy for Food
Quality Analysis and Control. D. W. Sun, ed. Elsevier Inc.

Ripoche, A., and A. S. Guillard. 2001. Determination of fatty acid composition of pork
fat by Fourier transform infrared spectroscopy. Meat Sci., 58:299-304.

Román, A., G. Vatai, A. Ittzés, Z. Kovács, and P. Czermak. 2012. Modeling of


diafiltration processes for demineralization of acid whey: an empirical approach.
Journal of Food Process Engineering, 35:708-714.

Rubico, S. M., and M. R. McDaniel. 1992. Sensory evaluation of acids by free-choice


profiling. Chem. Senses. 17:273–289.

Rutten, M. J. M., H. Bovenhuis, J. M. L. Heck, and J. A. M. van Arendonk. 2011.


Predicting bovine milk protein composition based on Fourier transform infrared
spectra. Journal of dairy science. 94:5683-5690.

Safar, M., D. Bertrand, P. Robert, M. F. Devaux, and C. Genot. 1994. Characterization of


edible oils, butters and margarines by Fourier transform infrared spectroscopy
with attenuated total reflectance. J. Am. Oil Chem. Soc. 71:371-377.

Sano, H., T. Egashira, Y. Kine awa, and N. Kitabata e. . Astringency of bovine mil
whey protein. . Dairy Sci. 88: 31 – 317.

114
Santos, P. M., E. R. Pereira-Filho, and L. E. Rodriguez-Saona. 2013. Rapid detection and
quantification of milk adulteration using infrared microspectroscopy and
chemometrics analysis. Food chemistry. 138:19-24.

Sawyer, L., G. Kontopidis, and S. -Y. Wu. 1999. β-lactoglobulin- a three-dimensional


perspective. Int. J. Food Sci. and Techn. 34:409-418.

Schmidt, R. H., B. L. Illingworth, E. M. Ahmed, R. L. Richter. 1978. The effect of


dialysis on heat-induced gelation of whey protein concentrates. J. Food
Processing and Preservation. 2:111-121.

Schmidt, R. H., B. L. Illingworth, J. C. Deng, J. A. Cornell, 1979. Multiple regression


and response surface analysis of the effects of calcium chloride and cysteine on
heat-induced whey protein gelation. J. Agric. Food Chem. 27:529-532.

Schmidt, R. H. 1981. Gelation and coagulation, Page 48-131 in Protein Functionality in


Foods. J. P. Cherry, ed. Am. Chem. Soc., Washington, D.C.

Shiroma, C., and L. Rodriguez-Saona. 2009. Application of NIR and MIR spectroscopy
in quality control of potato chips. J. Food Composition and Analysis. 22:596–605.

Siebert, K. . and A. W. Chassy. . An alternate mechanism for the astringent


sensation of acids. Food Quality and Preference. 1 :13‐ 18.

Simon, C., K. Barathieu, M. Laguerre, J. -M. Schmitter, E. Fouquet, I. Pianet, and E. J.


Dufourc. 2003. Three-dimensional structure and dynamics of wine tannin-saliva
protein complexes, a multitechnique approach. Biochemistry. 42:10385-10395.

Singh, A. M., and D. G. Dalgleish. 1998. The emulsifying properties of hydrolyzates of


whey proteins. Journal of dairy science. 81:918-924.

Sinha, R., C. Radha, J. Prakash, and P. Kaul. 2007. Whey protein hydrolysate: Functional
properties, nutritional quality and utilization in beverage formulation. Food
Chemistry. 101:1484-1491.
115
Smith, A. K., H. June, and A. C. Noble, 1996. Effects of viscosity on the bitterness and
astringency of grape seed tannin. Food Qual. Prefer. 7:161-166.

Smith, A. K., and A. C. Noble. 1998. Effects of increased viscosity on the sourness and
astringency of aluminum sulfate and citric acid. Food quality and preference, 9:
139-144.

Smithers, G. W. 2008. Whey and whey proteins-From ‘gutter-to-gold’. International


Dairy Journal. 18:695-704.

Sørensen, L. K., M. Lund, and B. Juul. . 2003. Accuracy of Fourier transform infrared
spectrometry in determination of casein in dairy cows' milk. J. Dairy Res. 70:445-
452.

Sowalsky, R. A., and A. C. Noble. 1998. Comparison of the effects of concentration, pH


and anion species on astringency and sourness of organic acids. Chemical senses.
23:343-349.

Subramanian, A., W. J. Harper, and L. E. Rodriguez-Saona, 2009. Rapid prediction of


composition and flavor quality of cheddar cheese using ATR–FTIR spectroscopy.
J. Food Sci. 74: C292 - C297.

Swaisgood, H. E. 1986. Chemistry of milk protein. Pages 1-60 in Developments in Dairy


Chemistry- 1. P. F. Fox, ed., Elsevier Applied Science Publishers Ltd, London
and New York.

Taira, S., M. Ono, and N. Matsumoto. 1997. Reduction of persimmon astringency by


complex formation between pectin and tannins. Postharvest Biology and
Technology. 12:265-271.

Thoma-Worringer, C., J. Sorensen, and R. Lopez-Fandino. 2006. Health effects and


technological features of caseinomacropeptide. Int. Dairy J. 16:1324–1333.

116
Trapp, G. A. 1985. Aluminum binding to organic acids and plasma proteins. Implications
for dialysis encephalopathy. J . Environ. Path. Toxicol. Oncol. 6:15-20.

Tunick, M. H. 2008. Whey Protein Production and Utilization: A Brief History. Pages 1-
13 in Whey Processing, Functionality and Health Benefits. C. Onwulata and P.
Huth, ed. Wiley-Blackwell, Ames, Iowa.

USDEC. 2004. Reference manual for U.S. whey and lactose products.
http://www.usdec.org

Udelhoven, T., D. Naumann, and J. Schmitt. 2000. Development of a hierarchical


classification system with artificial neural networks and FT-IR spectra for the
identification of bacteria. Applied Spectroscopy. 54:1471-1479.

Van de Voort, F. R., J. Sedaman, G. Emo, and A. A. Ismail. 1992. Assessment of Fourier
transform infrared analysis of milk. J AOAC Int. 75:780.

van der Ven, C., S. Muresan, H. Gruppen, D. B. A. D. Bont, K. B. Merck, and A. G. J.


Voragen. 2002. FTIR spectra of whey and casein hydrolysates in relation to their
functional properties. . Agric. Food Chem. :69 3−69 .

Vardhanabhuti, B., M. A. Kelly, P. J. Luck, M. A. Drake, and E. A. Foegeding. 2010.


Roles of charge interactions on astringency of whey proteins at low pH. Journal of
dairy science. 93:1890-1899.

Vidal, S., P. Courcoux, L. Francis, M. Kwiatkowski, R. Gawel, P. Williams, E. Waters,


and V. Cheynier. 2004. Use of an experimental design approach for evaluation of
key wine components on mouth-feel perception. Food Qual. Prefer. 15:209-217.

Willard, H. H.; L. L. Merritt, J. A. Dean, and F. A. Settle Jr. 1981. Infrared Spectroscopy.
Pages 177- 216 in Instrumental methods of analysis. Wadsworth Publishing
Company, Belmont, CA.

Wit, J.N. 1998. Nutritional and functional characteristics of whey proteins in food
products. J. Dairy Sci. 81:597-608.
117
Wold, S. 1976. Pattern recognition by means of disjoint principal components models.
Pattern Recognition. 8:127–139.

Wold, S. 1978. Cross-Validatory Estimation of the Number of Components in Factor and


Principal Components Models. Technometrics. 20:397-405.

Wold, S., H. Martens, and H. Wold. 1983. The multivariate calibration problem in
chemistry solved by the PLS method. Matrix Pencils. Springer Berlin Heidelberg,
286-293.
Wold, S., . S str m, and L. Eri sson. 1. PLS-regression: a basic tool of
chemometrics. Chemometrics and Intelligent Laboratory Systems. 8:1 9-13 .

Yamauchi, K., M. Shimizu, and T. Kamiya. 1980. Emulsifying properties of whey


protein. Journal of food science. 45:1237-1242.

Yang, H., and J. Irudayaraj. . 2000. Characterization of semisolid fats and edible oils by
Fourier transform infrared photoacoustic spectroscopy. J. Am. Oil Chem. Soc,
77:291-295.

Zeaiter, M., J. -M. Roger, and V. Bellon-Maurel. 2005. Robustness of models developed
by multivariate calibration. Part II: The influence of pre-processing methods.
Trends in Analytical Chemistry. 24:437-445.

Zhang, H., D. Yu, J. Sun, H. Guo, Q. Ding, R. Liu, and F. Ren. 2014. Interaction of milk
whey protein with common phenolic acids. Journal of Molecular Structure.
1058:228-233.

118

You might also like