wp25- Argyri et all HPLC 6th ICPMF_final

whipmellificiumΒιοτεχνολογία

20 Φεβ 2013 (πριν από 4 χρόνια και 3 μήνες)

200 εμφανίσεις

The potential of end
-
product metabolites on predicting the
shelf life of minced beef stored under aerobic and modified
atmospheres with or without the effect of essential oil

Washington D.C., USA

September 8
-
12, 2009

1

Laboratory of Microbiology and Biotechnology of Foods, Dept of Food Science and
Technology, Agricultural University of Athens, Greece

2

Laboratory of Applied Microbiology, Cranfield Health, Cranfield University, UK

3

Laboratory of Bioanalytical Spectroscopy, School of Chemistry, University of
Manchester, UK

A.A. Argyri
1,2
, E.Z. Panagou
1
, R. Jarvis
3
, R. Goodacre
3
, G.
-
J.E. Nychas
1


The

relationship

between

microbial

growth

and

chemical

changes

occurring

during

meat

storage

has

been

continuously

recognized

as

a

potential

means

to

reveal

indicators

that

may

be

useful

for

quantifying

beef

quality

or

freshness

(Nychas

et

al
.
,

2008
)
.




The

imposed

different

storage

conditions

and

preservatives

could

influence

the

production

of

these

potential

indicators,

through

the

establishment

of

a

transient

microbial

association

defined

as

the

‘Ephemeral

spoilage

micro
-
organisms’

-

ESO

(Nychas

and

Skandamis,

2005
)
.





Background and Rationale

Nychas, G.
-
J.E., Skandamis, P., Tassou, C.C., Koutsoumanis, K. (2008) Meat spoilage during distribution.
Meast Science,
78: 77
-
89.

Nychas, G.
-
J.E., Skandamis, P. (2005) Fresh meat spoilage and modified atmosphere packaging.
In
J.N. Sofos (Ed.),
Improving the Safety of Fresh Meat, CRC/Woodhead Publishing Ltd, Cambridge, UK.

There

is

need

for

a

holistic

approach

in

introducing

shelf
-
life

indicators

that

could

be

applied

irrespective

of

storage

temperature

or

packaging

system

and

be

eligible

to

the

income

of

new

technologies
.



This

approach

is

based

on

the

mining

of

qualitative

and

quantitative

data

of

metabolomics

e
.
g
.

indigenous

or

metabolic

compounds

associated

with

meat

spoilage,

due

to

interaction

of

ESO

with

nutrients

existing

in

meat

(Ellis

at

al
.
,

2001
)
.



The

use

of

HPLC

to

monitor

changes

in

the

organic

acid

profile

from

food

models

systems,

poultry,

fish

stored

under

different

storage

conditions,

has

been

considered

as

a

relatively

simple

and

promising

method
.




Background and Rationale

Ellis, D.I., Goodacre, R. (2001) Rapid and quantitative detection of the microbial spoilage of muscle foods: Current
status and future applications.
Trends in Food Science and Technology,
12: 414
-
424.

Objectives of the work


The

aim

of

the

present

work

was

to

investigate

the

potential

of

HPLC

spectral

data

of

organic

acids,

as

a

quick

analytical

method,

in

combination

with

an

appropriate

data

analysis

strategy

to
:

1.
Discriminate

among

different

quality

classes

of

minced

beef

samples

during

storage

at

different

temperatures

(
0
,

5
,

10
,

15
°
C)

and

packaging

conditions

(aerobic,

MAP,

MAP+EO)
.

2.
Correlate

the

microbial

load

of

different

microbial

groups

at

different

temperatures,

packaging

conditions

and

storage

times

with

spectral

data,

in

an

effort

to

predict

microbial

population

directly

from

HPLC

measurements
.


Materials & Methods

Product

:

Minced beef


Packaging

:

Aerobic,





MAP (
40% CO
2
, 30% O
2
, 30% N
2
),





MAP + oregano essential oil volatile compounds


Storage

temperature

:

0, 5, 10, 15

C


Microbiological analysis

:

Total viable counts, Pseudomonads,

Enterobacteriaceae
, lactic acid bacteria,
Brochotrix

thermosphacta
,

and yeasts and moulds



Organoleptic assessment

:

Spoilage detection based on changes in colour,

odour and taste based on a five member taste panel (Score range 1
-
3;

1=Fresh, 1.5 Semi
-
Fresh, 2
-
3 Spoiled).


HPLC analysis of organic acids

:

Collection of spectral data from the HPLC

(areas under peaks)

to monitor biochemical changes in meat during

storage.




Materials & Methods

HPLC analysis of organic acids


Sample preparation:
2g meat + 4mL dH2O + 1%TFA


Organic Acid Standards:
oxalic, citric, malic, lactic, acetic, formic,







tartaric, succinic and propionic


Apparatus:

Jasco,





Model PU
-
980 Inteligent pump,





Model LG
-
980
-
02 ternary gradient unit,





MD
-
910 multiwavelength detector at 210 nm



Data analysis

Collection of the
HPLC
spectral data
(areas under peaks)

1
st

Principal components analysis (PCA)

(Investigation of the peaks that significantly fluctuate during storage)

2
nd

PCA

Factorial Discriminant Analysis
(FDA)

predict the spoilage status of a sample;
fresh, semi
-
fresh, and spoiled

Regression Models


predict the counts of the different
microbial groups

Partial least squares regression

(PLS
-
R)

Support vector machines regression
(SVR)

Data mean centered and standardized


Multivariate calibration


Building calibration models

We

start

off

with

a

data

matrix,

and

a

corresponding

output

vector

which

indicates

the

value

associated

with

each

sample
.

(Meat)

(HPLC areas under peak)

(Bacterial counts)

We build a calibration model that relates the matrix to the vector.


Using calibration models

(New meat sample)

(HPLC area under peak)

(Predicted bacterial count)

The developed model on known
data, can be then applied to
unknown samples



Multivariate calibration approaches

There are two main

pattern recognition approaches based on:



Multivariate statistics



Multiple Linear Regression (MLR)



Principal Components Regression (PCR)



Partial Least Squares Regression (PLS
-
R)



Machine learning



Artificial neural networks



Support Vector Machines (SVM)


SVM underlying principle
*

Li,

H
.
,

Liang,

Y
.
,

Xu,

Q
.

(
2009
)

Support

vector

machines

and

its

applications

in

chemistry
.

Chemometrics

and

Intelligent

Laboratory

Systems,

95
:

188
-
198
.


The

idea

behind

SVMs

is

to

project

the

original

data

from

a

low

dimensional

input

space

to

a

higher

dimensional

feature

space
.


This

operation

is

called

feature

mapping

and

it

is

a

key

element

in

SVM

building
.


Dimension

superiority

plays

a

vital

role

in

SVMs
.


The

data

contain

more

information

as

the

dimension

increases
.




SVM underlying principle
*

Li,

H
.
,

Liang,

Y
.
,

Xu,

Q
.

(
2009
)

Support

vector

machines

and

its

applications

in

chemistry
.

Chemometrics

and

Intelligent

Laboratory

Systems,

95
:

188
-
198
.


SVM underlying principle

Li,

H
.
,

Liang,

Y
.
,

Xu,

Q
.

(
2009
)

Support

vector

machines

and

its

applications

in

chemistry
.

Chemometrics

and

Intelligent

Laboratory

Systems,

95
:

188
-
198
.


Data

projection

into

a

higher

dimensional

space

is

carried

out

by

a

kernel

function

that

serves

as

a

dimension

increasing

technique

and

further

transforms

the

linearly

inseparable

data

into

linearly

separable

one
.



There

are

number

of

kernels

that

can

be

used

in

Support

Vector

Machines

models
.

These

include

linear,

polynomial,

radial

basis

function

(RBF)

and

sigmoid
.

Pre
-
spoilage

Post
-
spoilage

Post
-
spoilage

Post
-
spoilage

15
°
C Air

15
°
C MAP+OEO

15
°
C MAP


Results

17 pure peaks were selected for analysis ;
RT of 6.2, 6.9 (citric acid), 7.0, 7.9, 8.3, 9.7, 10.9 (lactic acid),

11.9 (formic acid), 12.9 (acetic acid), 14.9, 15.1 (propionic acid), 16.1, 17.8, 18.6, 20.5, 24.6 and 28.1.


0 h

48 h

60 h

54 h




Aerobic storage,


Storage under MAP
,

Storage under MAP + OEO

Lactic acid

0C
0
10
20
30
40
50
60
70
80
90
0
100
200
300
400
500
600
Time (h)
Area (mAU.min)
5C
0
10
20
30
40
50
60
70
80
90
0
100
200
300
400
500
600
Time (h)
Area (mAU.min)
10C
0
10
20
30
40
50
60
70
80
90
0
100
200
300
400
500
600
Time (h)
Area (mAU.min)
15C
0
10
20
30
40
50
60
70
80
90
0
100
200
300
400
500
600
Time (h)
Area (mAU.min)

Results


Results

Qualitative

classification

of

the

samples


Observations (axes F1 and F2: 100.00 %)
0h
0A4
0M4
0M6
0M8
0O4
0O6
0O8
5A3
5A5
5M3
5M5
5M7
5O3
5O5
5O7
5O8
10A4
10M4
10O4
10O6
15A3
15M3
15M6
15O3
15O6
0A10
0A12
0A14
0M12
0M14
0O14
5A7
5A8
5A11
5A14
5M11
5M14
5O14
10A8
10A11
10A14
10A17
10M8
10M11
10M14
10M17
10O11
10O14
10O17
15A11
15A13
15A15
15A17
15M11
15M13
15M15
15M17
15O11
15O13
15O15
15O17
0A6
0A8
0M10
0O10
5M8
10A6
10O8
15A6
15A8
15M8
15O8
-4
-3
-2
-1
0
1
2
3
4
-5
-4
-3
-2
-1
0
1
2
3
4
5
F1 (95.81 %)
F2 (4.19 %)
F
S
SF
After the
end of
shelf life

Before
the end of
shelf life

Discriminant analysis similarity map determined by discriminant factors 1 (F1) and 2 (F2)
for HPLC spectral data of the 3 different beef fillets freshness groups:


Fresh (F),

Semi
-
fresh (SF),

and
Spoiled (S).

Confusion matrix for the cross
-
validation results of DFA


True class

Predicted class


Sensitivity
(%)

Fresh

Semi
-
fresh

Spoiled

Fresh

(n
= 26
)

23

2

1

88.46

Semi
-
fresh

(n
= 11
)

0

10

1

90.91

Spoiled

(n
= 38
)

2

2

34

89.47

Overall correct classification (accuracy): 89.33%


Results

5
6
7
8
9
10
5
6
7
8
9
10
Observed TVC (cfu/g)
Predicted TVC (cfu/g)
5
6
7
8
9
10
5
6
7
8
9
10
Observed TVC (cfu/g)
Predicted TVC (cfu/g)
5
6
7
8
9
10
5
6
7
8
9
10
Observed TVC (cfu/g)
Predicted TVC (cfu/g)
5
6
7
8
9
10
5
6
7
8
9
10
Observed TVC (cfu/g)
Predicted TVC (cfu/g)

Results

Prediction of the microbial population



Performance of regression models

PLS
-
R

Linear SVR

Radial SVR

Sigmoid SVR


Results

Calculation of performance indices (Bias and Accuracy factors)





PLS

-

R



Linear SV

R



Radial basis SV

R



Sigmoid

al


SV

R





Microbial group



B

f



A

f



B

f



A

f



B

f



A

f



B

f



A

f





TVC



0.99



1.10



0.99



1.10



1.00



1.10



1.00



1.10





Pseudomonas

spp



1.02



1.17



1.35



1.39



1.00



1.15



1.02



1.19





Br. thermosphacta



1.00



1.18



0.99



1.18



1.01



1.17



1.00



1.18





LAB



1.00



1.09



1.01



1.09



1.01



1.08



1.02



1.08





Enterobacteriaceae



0.99



1.16



1.00



1.14



1.00



1.14



0.98



1.14





Yeasts & Molds



0.99



1.15



1.00



1.15



0.99



1.11



1.02



1.14



RMSE
0.00
0.50
1.00
1.50
2.00
2.50
TVC
Pseudomonads
Br. thermosphacta
LAB
Enterobacteriaceae
Yeasts-Molds
PLS-R
Linear SVR
Radial SVR
Sigmoid SVR

Results

Prediction of the microbial loads

-

Regression models’ Performance



Good

correlation

of

the

sensorial

evaluation

of

spoilage

with

the

dynamic

changes

of

the

chromatographic

areas

of

organic

acids

at

different

time

intervals
.





In

general

the

PLS
-
R,

radial

basis

SVR

and

sigmoid

SVR

exhibited

slightly

better

performance

than

the

Linear

SVR

whereas

t
he

models

that

described

the

estimates

of

the

TVC
,

as

well

as

the

LAB
,

had

better

performance,

regardless

of

the

type

of

the

model

built
.





HPLC

analysis

of

organic

acids

can

be

proved

as

a

potential

technique

for

meat

analysis

in

predicting

the

spoilage

status

and

the

microbial

load

of

a

meat

sample

regardless

of

the

storage

conditions
.



Concluding remarks

Acknowledgements

This

work

was

supported

by

the

EU

projects

Symbiosis

[
7
th

Framework

Programme

(Con
.

No

21638
)
]

and

ProSafeBeef

[
6
th

Framework

Programme

(ref
.

Food
-
CT
-
2006
-
36241
)
]
.


Thank you for your attention