Fresh milk has a shelf life of a few hours at
refrigerated temperature of less than 4?, due to increase the shelf life of
milk and saving the costs of cooling, milk adulterated with chemicals such as
penicillin, formalin and hydrogen peroxide. These additives can increase the
shelf life of milk without cooling and these additions have the potential to
cause serious health-related problems for human and also affect on processing
of milk in dairy industry. Several analytical techniques can be used to detect
adulteration but they often require time-consuming, sample preparation,
expensive laboratory equipment, and highly skilled personnel. We employed a
portable Near Infrared spectroscopy (NIR) combined with multivariate analysis
was developed to detect adulteration as well as to quantify the level of
hydrogen peroxide in cow milk adulterated. In this study fresh cow milk samples
were collected from three farms surrounding Beijing district and five farms in
Hebei province in China were investigated. Those fresh cow milk samples were
then adulterated with hydrogen peroxide at ten different percentage levels: 0.5%,
0.7%, 0.9%, 1%, 1.5%, 1.7%, 1.9%, 2%, 2.5%, and 2.7% of hydrogen peroxide. All
samples were scanned using NIR spectroscopy (JDSU, California, USA). The
MicroNIR 1700 spectrometer analyses the wavelength region is 870 – 1660 nm. The
chemometric tools like Principle component analysis (PCA), partial least
discriminant analysis (PLS-DA) and partial least squares regression (PLS) was
applied for statistical analysis of the obtained NIR spectral data. To check
the discrimination between the pure and hydrogen peroxide adulterated milk
samples, PLS-DA model was used. For PLSDA model the R-square value obtained was
0.989 with 0.081 RMSE (Root mean square error). Furthermore, PLS regression
model was also built to quantify the levels hydrogen peroxide adulterant in cow
milk samples. The PLS regression model was obtained with the R-square 93% and
with 1.38 RMSECV(Root mean square error of cross validation) value having good
prediction with RMSEP (Root mean square error of prediction) value 1.50 and
correlation of 0.95. This recently developed method is non-destructive, low-cost,
no need of much sample preparation and having sensitivity level less than 1%
level of hydrogen peroxide adulteration. This recently developed method is fast
and simple, being suitable for the control of raw milk in a dairy industry or
for the quality inspection of commercialized milk.

Keywords: NIR-spectroscopy; hydrogen peroxide;
Milk adulteration; PCA; PLS-DA; PLS regression

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

 

 

 

 

1.    
 Introduction

Adulteration in food is a biger problem faces
the world, which has required more investigation, because it has attracted more
attention worldwide recently. (Spink and Moyer 2011).

Milk is the fluid secreted by the female of
all mammalian species, to meet the essential nutritional requirement of the
neonate, and it is very important component in human diets in the worldwide.

Milk is most important source for essential
nutrients and health maintenance. Milk is perfect food. It is valuable source
of fat, protein, carbohydrates, vitamins and minerals. (Faraz et al., 2013).

Milk is a balanced mixture and a perishable
food. It is one of few foods consumed in the natural form throughout the world
.Milk contain 87% water, 3.3% protein , 3.9% fats , 5% lactose and 0.7% ash .
Milk provides us building protein for body, giving vitamins, minerals (calcium)
for bone forming and energy supplying (lactose and milk fat). Besides providing
certain essential fatty acids it includes all essential amino acid. All the
properties of milk make it an important food for all ages. (Khan et al., 2005).

The low costs and high nutritional value of
milk has made it form a significant part of the human diet. However, global
increased demand made this foodstuff pone for more potential adulteration (Salih et al., 2017).

 In 2016, more than 816 million
tons of milk was produced around the world (including cow, buffalo, sheep, goat
and camel milk). India was the largest producer, with 155.2 millions of tons, USA
was the second major producer (96.3 millions of tons) and China was 43.4
millions of tons, table 1 and figure 1 (FAO, 2016). However, as fast as the
production and demand for milk have grown, a larger frequency of sophisticated
milk adulterations has been reported in different countries.

 year

2014

2015

2016

WORLD

771

803

816

Table 1.
World production milk

                                                                        
figure1. World production milk

Adulteration of milk and other dairy products
has existed from old times (Astrid et al., 2010), and it is done either for
profit margin or lack of hygienic status of storage, processing, transportation
and marketing (Faraz et al., 2013).

Cleanliness and purity of milk are indicating
for good quality milk (Kanwal et al., 2004). For
quality dairy items and better health of consumer, quality control of milk is
required (Nirwal et al., 2013).

Unfortunately, adulteration is very common in
developing countries (Salih et al., 2017). Increase population, the urbanization of rural, and
scattered colonization are the few main factors affected on increasing the
demand of milk production (Awan et al., 2014). In order to meet the shortage between demand and
supply of milk the unscrupulous producers are often found to involve in milk
adulteration by adding adulterating substances in milk (Mustafa et al., 2014).

In fresh milk, chemical used as preservatives
like formalin, hydrogen peroxide, boric acid and antibiotics are added to
increase the shelf life (Chanda et al., 2012).The other kind of adulteration of
milk by the additions of starch, rice flour, skim milk powder, reconstituted
milk, urea, melamine, salt, glucose, vegetable oil, animal fat and whey powder.
These additions is to increase the thickness and viscosity of the milk, and to
maintain the composition of fat, carbohydrate and protein (Campos Motta et
al., 2014; Singuluri and Sukumaran 2014; Soomro et al., 2014).

So to increase the shelf life, addition of
chemical preservatives in branded milk (Mohanan et al., 2002) is a very common
practice. Sometimes hydrogen peroxide (H2O2) is used as a
preservative (Paixao and Bertotti.2009). These additions have the potential to
cause serious health-related problems.

2.    
Materials and methods

The Figure
2 show a Technical frame diagram of experiment design.

2.1.        
Cow milk samples preparation

In this study fresh cow milk samples were
collected from three farms surrounding Beijing district and five farms in Hebei
province in China were investigated. Those fresh cow milk samples were then
adulterated with hydrogen peroxide at ten different percentage levels: 0.5%,
0.7%, 0.9%, 1%, 1.5%, 1.7%, 1.9%, 2%, 2.5%, and 2.7% of hydrogen peroxide. The
total number of samples used was 114: 34 pure cow milk samples, 80 adulterated
with hydrogen peroxide. For PLS regression all the samples were joined together
and split into two sets, a training set (70% of the samples) and a test set for
validation (30% of the samples).

Technical frame diagram:

Milk samples

Farms

Milk directly from cow

Adulteration milk samples

Control

H2O2

Unknown samples

 NIRS scan

 

 

 

 

 

 

 

 

 

 

Scatter
effect remove

                                                                                                                                                 

Calibration

                                                                                                                                                 

Robust NIR method with procedure
and optimal models
 

Validation

Linear /non-linear analysis

 

 

 

 

Pilot farms/ markets
 

                                                                                                    

                     Application

2.2.        
NIR spectroscopic analysis

All samples were scanned using NIR spectroscopy
(JDSU, California/ USA). The MicroNIR 1700 spectrometer analyses the wavelength
region is 870 – 1660 nm. Prominent absorption peaks were appeared in the region
from 930 to 1350 nm wavelength.

2.3.        
Statistical analysis

Microsoft Excel 2010 and The Unscrambler
version 9.0 by Camo were used for statistical analysis. The PCA, PLS-DA and PLS
regression models were built for both pure and adulterated cow milk samples.
Spectral pretreatments, such as standard normal variate (SNV) and 1st
derivative with Savitzky-Golay smoothing 5 points were carried. Full cross
validation was used to validate the PLS-DA models. For PLS regression all the
samples were joined together and split into two sets, a training set (70% of
the samples) and a test set for validation (30% of the samples). External cross
validation was used to validate the PLS regression models built with the
training set. The Root Mean Square Error of Cross Validation (RMSECV) was used
as an internal indicator of the predictive ability of the models. RMSECV is
calculated using equation (1)

      

      ——————————————–
(1)

 

Where (yi ) is the measured value
(actual % of adulteration), (?i) is the % of adulteration predicted
by the model, and n is the number of segments left-out in the cross-validation
procedure, which is equal to the number of samples of the training set. Smaller
values of RMSECV are indicative of a better prediction ability of the model.

The RMSEP is a statistical measure how well
the model predicts new samples (not used when building the model). It is
calculated using equation (2)

 

     
——————————————– (2)

 

             

Where yt,i is the measured value
(actual % of adulteration), ?t,i is the % of adulteration predicted
by the model, and nt is the number of samples in the test set. RMSEP
expresses the average error to be expected in future predictions when the
calibration model is applied to unknown samples.

 

3.    
Results and Discussion

3.1.        
Near infrared spectra

Figure 3 shows the NIR spectra of all the
samples ranging from 870 – 1660 nm in term of wavelength while in term of wave
numbers ranging from 11494.25 -6024.10 cm-1.

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3 NIR spectra both for
pure and hydrogen peroxide adulterated milk

 

 

The spectra in Figure 3 show a scattering
effect due to milk and hydrogen peroxide solid particles and white colors.
Spectral pretreatments, such as SNV were used to remove the scattering effect
as shown in Figure 4.

 

 

Figure
4 NIR spectra both for pure and hydrogen peroxide adulterated milk samples
after SNV pre-processing.

 

Although the spectra appear to be very
similar, the application of a 1st derivative function with
Savitzky-Golay smoothing 5 points were applied at 2 polynomial order shows that
there are clear differences in the spectral absorption regions as shown in
Figure 5.

 

Figure
5: 1st derivative NIR spectra both for pure and hydrogen peroxide adulterated
milk samples.

 

It can be seen from the spectra in Figures 5 that there are prominent
absorption peaks at wavelength 980 nm and 1200 nm for both pure and hydrogen
peroxide adulterated milk samples.

 

 

 Figure 6:
PCA score plot for four different types of cow milks samples

 

 

In order to visualize the effect of variation
among the tow different places (Beijing & Hebei) of cow milk an alternative
approach of principal components analysis (PCA), was applied in that a PCA
model was built as shown in Figure 6. PCA is a standard multivariate data
analysis exploratory tool. It is used to reduce the dimensionality of a complex
data set without much loss of information, to extract the most important
information from the data table, to identify noise and outlier in the data set.
It is a way of identifying the underlying patterns in data for further analysis
using other techniques. The procedure of PCA is like that it converts a set of
correlated variables into a new set of uncorrelated variables called principal
components. PCA redistributes the total variance of the data set in such a way
that the first principal component has maximum variance, followed by second
component and so on.

Variance
PC1>Variance PC2>… Variance PCk

Total
variance=Variance PC1+Variance PC2+ … Variance PCk

The covariance of any of the principal component
with any other principal component is zero (uncorrelated) and they are
orthogonal to each other.

 It can
be seen from the PCA score plot that there is complete differentiation and
separation among those two different places of milk samples. They are spaced
and grouped in the specific regions of the PCA score plot. The milk samples
those are similar are clustered in one group than the different one. Use
existing PCA models to build a SIMCA classification model, and then classify new samples.

This table (2) shows the classification of each sample. Classes that are significant for a sample are marked with a star.
The outcome of the classification depends on the significance limit; 0 .5%. Look
for samples that are not recognized by any of the classes, or those that are
allocated to more than one class.

 

 

Table 2.
The classification of pure and hydrogen peroxide adulterated
milk samples.

Sample

Control
PCA

Adulterated
PCA

1H hp3

*       

*

1H hp4

*       

*

1H hp5

*       

*

1H hp7

*       

*

1H hp8

*       

*

1H hp9

*       

*

1B
hpl2

*       

*

1B
hpl3

*       

*

1B
hpl5

*       

*

1B
hpl6

*       

*

1B
hpl7

*       

*

1B
hpl8

*       

*

1B
hpl9

*       

*

2B
col1

*       

 

2B
col2

*       

 

2H
col1

*       

 

2H
col2

*       

 

2H
col3

*       

 

2H
col5

*       

 

2H
hp10

 

 

2H hp2

*       

*

2H hp3

*       

*

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Similarly, Partial least-squares discriminant
analysis (PLS-DA) model was built for the spectral data between pure and with 2%
hydrogen peroxide milk adulteration as shown in Figure 7. PLS-DA model can be
used as an identification tool to check hydrogen peroxide adulteration in cow milk.
If there is any amount of hydrogen peroxide in cow milk they will occupy the
space in between the pure and adulterated samples of the Figure 7.

 

 

 

Figure 7: PLS-DA model for the pure cow milk
and with 2%
hydrogen peroxide   adulteration

 

It can be seen from Figure 7 that there is a clear discrimination
between the pure cow milk samples as well as with 2% hydrogen peroxide milk adulteration.
The RMSECV value for PLSDA model was found 0.081 with R square value of 0.989.
The PLSDA model with minimum error that is RMSECV and with highest correlationship
value i.e. R is the best one.

 

PLS regression results

To predict the level of hydrogen peroxide
adulteration in cow milk samples PLS regression model was built by using 70% of
the samples as a training set with hydrogen peroxide at ten different
percentage levels: 0.5%, 0.7%, 0.9%, 1%, 1.5%, 1.7%, 1.9%, 2%, 2.5%, and 2.7%
of hydrogen peroxide. PLS regression models are shown in Figures 8 and 9. PLS
finds a set of orthogonal components that maximize the level of explanation of
both X and Y provides a predictive equation for Y in terms of the X’s