J. Dairy Sci.

97:1735
http://dx.doi.org/10.3168/jds.2013-6648

American Dairy Science Association , 2014.

Prediction of fatty acid profiles in cow, ewe, and goat milk

by mid-infrared spectrometry
1

M. Ferrand-Calmels,* I. Palhire, M. Brochard,* O. Leray, J. M. Astruc,* M. R. Aurel, S. Barbey,#

F. Bouvier," P. Brunschwig,* H. Caillat, M. Douguet,* F. Faucon-Lahalle,*{ M. Gel,* G. Thomas,*
J. M. Trommenschlager,** and H. Larroque
*Institut de lElevage, 149 rue de Bercy, F-75595 Paris cedex 12, France
INRA, UR0631, Station dAmlioration Gntique des Animaux, F-31326 Castanet-Tolosan cedex, France
Actilait, Rue de Versailles, F-39800 Poligny, France
INRA, UE 0321, Domaine exprimental de La Fage, F-12250 Roquefort-sur-Soulzon, France
#INRA, UE 0326, Domaine exprimental du Pin-au Haras, F-61310 Exmes, France
"INRA, UE 0332, Domaine exprimental de Bourges-La Sapinire, 18390 Osmoy, France
CNIEL, 42 rue de Chteaudun, 75314 Paris cedex 09, France
**INRA, UR 0055, Agro-Systmes Territoires Ressources, 662 avenue Louis Buffet, F-88500 Mirecourt, France

ABSTRACT

Mid-infrared (MIR) spectrometry was used to

esti- mate the fatty acid (FA) composition in
cow, ewe, and goat milk. The objectives were
to compare different statistical approaches
with wavelength selection to predict the milk
FA composition from MIR spectra, and to
develop equations for FA in cow, goat, and
ewe milk. In total, a set of 349 cow milk
samples, 200 ewe milk samples, and 332 goat
milk samples were both analyzed by MIR and
by gas chromatography, the ref- erence
method. A broad FA variability was ensured by
using milk from different breeds and feeding
systems. The methods studied were partial
least squares regres- sion (PLS), first-derivative
pretreatment + PLS, ge- netic algorithm +
PLS, wavelets + PLS, least absolute shrinkage
and selection operator method (LASSO), and
elastic net. The best results were obtained with
PLS, genetic algorithm + PLS
and
first
derivative + PLS. The residual standard
deviation
and
the coefficient of
determination in external validation were used
to characterize the equations and to retain the
best for each FA in each species. In all cases,
the predictions were of better quality for FA
found at medium to high concentrations (i.e.,
for saturated FA and some mono- unsaturated
FA with a coefficient of determination in
external validation >0.90). The conversion of
the FA expressed in grams per 100 mL of
milk to grams per 100 g of FA was possible
with a small loss of accuracy for some FA.
Key words: milk, fatty acid, mid-infrared
spectrom- etry

Received January 31, 2013.

Accepted September 30, 2013.

Corresponding author: marion.ferrand@idele.fr

INTRODUCTI
ON

Over the last few years, the FA content

and quality of human foods has become a
major nutritional topic. Evidence of this can
be found in the latest French nutri- tional
guidelines that indicate the recommended
intake of individual FA in a diversified diet
(ANSES, 2011a). The dairy industry has to
face 2 major issues regarding these new
recommendations: (1) identifying methods to
adapt milk FA composition to the consumers
demands and (2) finding tools to precisely
characterize the FA composition of milk.
With regard to the first issue, previous
studies
have
shown
that
feeding
management (Chilliard et al., 2007; RaynalLjutovac et al., 2008; Coppa et al., 2013)
and genetic selection (Arnould et Soyeurt.,
2009; Schennink et al., 2009; Stoop et al.,
2009) can be efficient tools to alter the FA
composition of milk. However,
such
strategies can only effectively be organized 17

for entire populations if reliable large-scale

measuring techniques are available.
As to finding tools to determine the precise
FA com- position of milk, mid-infrared (MIR)
spectrometry can be classically and efficiently
used to analyze food prod- ucts. This
technology is already used in dairy farming to
measure the fat and protein contents for
purposes of milk payment, herd management,
and genetic selec- tion. In France, these
analyses are performed frequently (once per
month for all animals of a herd in dairy cattle
or goats managed with the usual milk
recording design) at a large scale by
laboratories located throughout the country.
More recently, this process has been extended
to include the analysis of milk FA (Soyeurt et
al., 2006, 2011), especially in dairy cattle. Most
often, FA com- position is predicted using MIR
spectra with partial least squares regression
(PLS). Although it has been reported that
preprocessing before PLS regression im-

18

FERRAND-CALMELS ET AL.

proves calibration equations (Soyeurt et al.,

2011), the possible benefit of wavelength
selection before applying PLS regression has
not been addressed. It has been suggested
that wavelength selection before PLS regression might improve calibration equations and
provide good results in various situations
(Leardi et al., 1992; Spiegelman et al., 1998).
It can be performed using different methods:
genetic algorithms (Leardi et al., 1992),
wavelet decomposition (Mallat, 2008), or
penal- ization methods such as the elastic
net, which is often used on genomic data
(Croiseau et al., 2011). However, no
comparison of these methods is currently
available. Equations for goat milk were
developed using near- infrared spectroscopy
(NIRS; Andueza et al., 2013), but no
calibration equations for predicting milk FA
using MIR spectrometry have been published
so far for ewe and goat milk. Within the
framework of the PhnoFinlait program
(Faucon-Lahalle et al., 2009), a large-scale
French research and development project
aiming at monitoring the composition of
cattle, sheep, and goat milk, calibration
equations were developed using different
mathematical approaches to estimate milk FA
composition in French herds using MIR spectrometry.
The objectives of this study were (1) to
compare different statistical approaches with
wavelength se- lection for the prediction of
milk composition using MIR spectra, (2) to
establish whether the use of MIR spectrometry
for predicting the FA composition of cow milk
could be put into practice in French breeding
and feeding systems, and (3) to develop the
first referenced equations for FA in goat and
ewe milk.
MATERIALS AND METHODS
Experimental Design

Samples were mainly collected from the

Institut
National
de
la
Recherche
Agronomique (INRA) experi- mental farms
(Table 1). These experimental farms breed
ruminants of the most widespread breeds in
France in conditions representative of French
breeding systems, with at least 2 different
feeding systems depending on the season
(winter and summer) for each breed. Besides
providing more practical technical conditions
for col- lecting samples, these farms apply
controlled breeding conditions and, in some
cases, shelter populations with strong genetic

variability. When this genetic or feed- ing

variability was not sufficiently representative of
the national situation, additional samples were
collected from commercial herds. Two samples
were collected per animal at each milking.
Bronopol was added to one of the samples,
which was then analyzed by MIR spectrometry;
the other was frozen at 20C immedi-

ately after collection for future analysis using

GC, the reference method.
Cow Milk Samples. A first set of 249 milk
samples was collected from 127 Holstein
Normande crossbred dairy cows in 2008 and
2009 at the INRA Domaine exprimental du
Pin experimental farm. The cows were a
part of a QTL detection experiment
(Larroque et al., 2002), and were produced
after 2 generations of crosses (F2) between
Normande and Holstein breeds that display
numerous differences, in particular those
pertaining to milk fat and protein content.
Milk samples were collected twice during
the first lactation, in the winter and summer.
The average stage of lactation was 160 DIM
in the winter, and 209 DIM in the summer.
During the winter (NovemberApril), all cows
were given the same diet, formulated to
cover their requirements (INRA, 1989) and
based on corn silage given ad libitum and
completed with grass silage (4 kg of DM/d),
beet pulp (1 kg of DM/d), and soybean meal
(2 kg of DM/d). During the summer, cows
grazed herbage with high nutritive value in a
2
rotational graz- ing system (35 hm per cow)
and were supplemented with corn silage (1.5
kg/d). When the daily milk yield exceeded 21
kg/d in the winter and 23 kg/d in the summer,
cows received 1 kg of energy concentrate
per
2.5 kg of milk produced over the limit
(maximum 7 kg of concentrate in winter and
5 kg in summer).
Journal of Dairy Science Vol. 97 No. 1, 2014

A second set of 153 milk samples were

collected from
42
Montbliardes
and
35
Holsteins
(primiparous or multiparous) in 2009 at the
INRA experimental farm in Mirecourt (France),
which implements an organic dairy farming
system. Depending on calving dates, the
cows belonged either to a grazing system or
to
a mixed-crop dairy system without
concentrate. Milk samples were collected twice
during their lactation: in the winter (March
2009) at 29 and 184 DIM, or in the summer
(June 2009) at 120 and 274 DIM, respectively,
for the grazing and mixed-crop dairy systems.
In the winter, cows from the grazing system
received ad li- bitum hay from the first cut of
permanent grassland, which was completed
with a second cut of permanent grassland (5.6
kg of DM/d) and a mineral complement. At the
same period, cows from the mixed-crop dairy
system received a diet dominated by hay
from a first cut of alfalfa and orchard grass
offered ad libitum and completed by flattened
triticale (1.7 kg of DM/d), a second cut of
temporary grassland (6.7 kg of DM/d), and a
mineral complement. In the summer, cows
from the grazing system grazed a new
enclosed pasture of permanent grassland
(orchard grass, white clover, and fescue). Cows
from the mixed-crop dairy system grazed
alternatively (night and day) 2 enclosed
pastures of permanent grassland (based on
ryegrass, white clover, and orchard grass). Milk
samples were collected
using

Thank you for using www.freepdfconvert.com service!

Only two pages are converted. Please Sign Up to convert all pages.
https://www.freepdfconvert.com/membership