Milk metabonomics: Impact of sample preparation and preprocessing on model

stagetofuΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

325 εμφανίσεις

Milk metabonomic
s: Impact of sample preparation and

preprocessing on model
robustness
elucidated on
400

milk samples


Sundekilde UK
1
, Clausen MR
1
, Larsen LB
2

& Bertram HC
1

1

Department of Food Science
, Årslev
,
Science and Technology
, Aarhus University,
Denmark

2

Department of Food Science
, Foulum
,
Science and Technology
, Aarhus University, Denmark


The use of NMR metabonomics in assessing milk quality is a
n attractive approach as it is

rapid,
non
-
destructive and
gives reproducible results
. However, w
hole

milk is an emulsion of high and
low molecular weight constituents

some of which give rise to broad NMR peaks, thus care must be
taken in order to
get
high quality data, which in turn shows promising possibilities for
the
prediction
of technological and nu
tritional quality of
the
milk.


In the joint Danish
-
Swedish Milk Genomics Initiative

milk from 1200 cows originating from
different farms in Denmark and Sweden
comprising
three

different breeds
have been

sampled and
screened.

This large
-
scale screening of
milk phenotypes include many different analyses including
metabolic profiling, mineral and fatty acid composition and assessment of functional properties of
the milk.
The present
study encompasses

skimmed milk samples

from

40
0 Danish Holstein
-
Friesian
cows
.

Proton nuclear magnetic resonance (NMR) is applied for metabonomic
profiling

of the
se

individual milk samples using a high
-
resolution 600 MHz NMR spectr
ometer.

The present study
aims at elucidating whether centrifugation or filtration
of samples
selecti
vely removes important
m
etabolites from the NMR profile and if different preprocessing steps are able to highlight
interesting metabolites.


Following NMR measurements the spectra must be phase and baseline corrected and a number of
preprocessing steps mu
st be performed in order to obtain reliable data. Alignment can either be
done by shifting the entire spectrum
(
1
)

or by
interval Correlation Optimized shifting

(
2
)
. Small
shifts can also

be dealt with

by data reduction methods such as binning
.

Thus
, we have evaluated
the use of
the entire resolution

i.e. no binning
, using fixed width bins of varying size (typically
0.01
-
0.1

ppm
;
(
3
)

and by using
adaptive,
intelligent binning

algorithms
(
4
)
. Normalization is
another crucial pr
eprocessing step. In the present study we have tested absence of normalization,
normalization to TSP, integral normalization, and probabilistic quotient normalization
(
5
)
. The final
step before multivariate data analysis is the data scaling, which enables non dominant metaboli
tes to
influence the model behavior.
Hence
, we also test pareto scaling, unit
variance scaling, range
scaling
(
6
)
, VAST scaling
(
7
)
, and log
or power
transformations
(
8
)
.

Furthermore, useful techniques
for variable selection are discussed.



In summary, the present study aims at maximizing t
he
output of a single
set of milk
sample
s

using
different sample preparation and preprocessing techniques

as outlined above
.


Acknowledgements

The present Ph.d.
-
project is part of a joint Swedish/Danish ”Milk Genomics Initiative” funded by FØSU, Danish Cattle
Federation and Faculty of Agricultur
al Sciences, Aarhus University.



Reference List



1.

G. Tomasi, F. van den Berg, C. Andersson,
J. Chemometrics

18
, 231 (2004).


2.

F. Savorani, G. Tomasi, S. B. Engelsen,
Journal of Magnetic Resonance

202
, 190 (2010).


3.

A. Craig, O. Cloareo, E. Holmes, J. K. Nicholson, J. C. Lindon,
Analytical Chemistry

78
,
2262 (2006).


4.

T. De Meyer

et al.
,
Analytical Chemistry

80
, 3783 (2008).


5.

F. Dieterle, A. Ross, G. Schlotterbeck, H. Senn,
Analytical Chemistry

78
, 4281 (2006).


6.

A. K. Smilde, M. t. J. van der Werf, S. Bijlsma, B. van der Werff
-
van der Vat, R. H. Jellema,
Analytical Chemistry

77
, 6729 (2005).


7.

H. C. Keun

et al.
,
Analytica Chimica Acta

490
, 265 (2003).


8.

O. M. Kvalheim, F. Brakstad, Y. Liang,
Analytical

Chemistry

66
, 43 (1994).