standardisation of mir milk spectra, a step to build an ... · grelet c., fernandez pierna j.a.,...
TRANSCRIPT
Grelet C., Fernandez Pierna J.A., Dehareng F. & Dardenne P.
Centre Wallon de Recherches Agronomiques (CRA-W), Gembloux, Belgique
0 50 100 150 200 250 300-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
Standardisation of MIR milk spectra,
a step to build an international spectral database
PARTNERS
11 MILK CONTROL ORGANISATIONS
Name Country
Institut de l’Elevage FR
Gembloux Agro-Bio Tech
(Université de Liège)
BE
Centre wallon de Recherches
agronomiques (Département
Qualité Productions Animales)
BE
Comité du Lait BE
Teagasc IR
Scottish Agricultural College UK
Name Country
Association Wallonne de l’Elevage BE
Chambre régionale Agriculture Alsace FR
ADECL62 (Pas-de-Calais) FR
CLASEL (Sarthe & Mayenne) FR
SCL du Doubs et du territoire de
Belfort
FR
France Conseil Elevage FR
LKV Baden-Württemberg DE
LKV Nordrhein-Westfalen DE
National Milk Recording UK
Irish Cattle Breeding Federation IR
CONVIS LU
5 Research units
+ 1 Laboratoire
•Position of the peaks Qualitative analysis
•Intensity of the peaks Quantitative analysis
C=O
Fat, organic acids
C=O, N-H,
C-N
protein
CH
Fat + protein
CH
Fat
C-O
Fat
C-O, OH
Lactose
MIR technology
MIR spectra Reference analysis
Saturated FA (g/100ml) = 0.0985 + 4.6191 *X1 +1.1659*X2 +1.4827*X3
+4.1684*X4 +7.3294*X5 +9.8991*X6 +11.183*X7 +8.0711*X8
+2.1599*X9 -0.4619*X10 -1.7876*X11 -2.5708*X12 -2.8941*X13 -
2.9217*X14 -2.7392*X15 -2.2543*X16 -1.2677*X17 +0.0676*X18
+1.0762*X19 +1.3228*X20 +1.0241*X21 +0.536*X22 + 0.0177*X23 -
0.5265*X24 -1.1445*X25 -1.8178*X26 +-2.212*X27 -2.0766*X28 +-
8.3083*X29 -3.703*X30 +-1.1999*X31 +-0.5698*X32 -0.1674*X33
+0.246*X34 +0.666*X35 +1.2938*X36 +2.0946*X37 -0.0689*X38 -
1.4774*X39 -1.7984*X40 -2.0553*X41 -2.9338*X42 -4.644*X43 -6.764*X44
-8.1475*X45 -5.6904*X46 +2.6657*X47 +10.9883*X48+14.4346*X49
+13.8878*X50 +10.2135*X51+4.8464*X52 -1.2081*X53-7.4854*X54-
11.6799*X55 -12.6849*X56 -10.7724*X57 -4.8936*X58 + 0.4425*X59
+3.583*X60+………………………………………………………….…….+2
.9636*X310 +6.4566*X311
MIR predictions
Saturated FA (g/100ml) =ax + b
y = 0.5087x - 0.0089 R² = 0.9945
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3
Ab
so
rban
ce v
alu
e
FA %
Milk control MIR
MIR spectra
of each cow
Milk
composition Fat
Proteins
Urea
FA
...
© Bentley
Classical use of MIR spectra :
Fast
Cheap
MIR technology
OptiMize this
source
of informations
Spectra MIR Cow state
Prediction tools fast, cheap, via milk control
organisations
Informations on : - fertility
- feeding
- health
- environmental impact
Innovative view of OptiMIR:
-1.00E-01
-5.00E-02
0.00E+00
5.00E-02
1.00E-01
1.50E-01
2.00E-01
2.50E-01
3.00E-01
93
099
910
68
11
38
12
07
12
77
13
46
14
16
14
85
15
55
16
24
16
93
17
63
18
32
19
02
19
71
20
41
21
10
21
79
22
49
23
18
23
88
24
57
25
27
25
96
26
65
27
35
28
04
28
74
29
43
30
13
Cm-1
Hypothetic spectra of healthy cow and cow suffering of SARA
Healthy cow
SARA ?
Concrete example: SARA
Impact of SARA on fat rate,
fatty acid profile… (Sauvant et al.,
1999; Colman et al,, 2010)
Ab
so
rban
ce
Combine data from each MRO :
STEP N°1
Transnational
database
Spectra MIR + Fertility, health, feeding and environmental
data
Prediction tools of
cows states
No standard format !!
Issue
0 200 400 600 800 1000 1200-0.5
0
0.5
1
1.5
2
2.5
3
3.5original data
Bentley.
Foss
10 spectra from identical milks :
N= 10y= -0.019098+0.61228xR square= 0.99995biais= 1.0452RMSE= 1.2073
0 1 2 3 4 5 6 70.5
1
1.5
2
2.5
3
3.5
4Ymeasured vs. Ypredicted
Predicted Values
Refe
rence V
alu
es
Predictions fat with raw spectra :
Prédictions Bentley
Pré
dic
tio
ns Fo
ss
Prédiction Foss : 2.6%
Prédiction Bentley : 4.3%
Need of spectral
standardisation
Absorbance in the area of r1 (master)
correlated to R2 (slaves)
« Master »
« Slave »
r1
R2
PIECE-WISE DIRECT STANDARDIZATION (PDS)
r1j = R2j bj + b0j
Method
Method
0
1
2
3
4
5
0 2 4 6
Pro
tein
s %
Fat %
Ring test samples composition
30 Labs
69 Apparatus
0 200 400 600 800 1000 1200-0.5
0
0.5
1
1.5
2
2.5
3
3.5original data
Bentley – Slave
889 wav.
Foss – Master
1060 wav.
0 200 400 600 800 1000 1200-0.5
0
0.5
1
1.5
2
2.5
3
3.5slave after interpolation
Interpolation
Method
0 200 400 600 800 1000 1200-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6slave after shift transmittance/absorbance
0 200 400 600 800 1000 1200-0.5
0
0.5
1
1.5
2
2.5
3
3.5slave after interpolation
Foss:
Transmittance
Bentley: Absorbance
Logarithmic transformation
Method
PDS
0 100 200 300 400 500
-0.1
0
0.1
0.2
0.3
0.4
0.5
Spectra after log
wavelength (cm-1)
A
0 100 200 300 400 500
-0.1
0
0.1
0.2
0.3
0.4
0.5
Spectra after slave standardization
wavelength (cm-1)
A
Method
Possibility to merge spectra
into a common database
N= 10y= -0.01999+1.0076xR square= 0.99993biais= 9.4369e-016RMSE= 0.017542
0 1 2 3 4 5 6 70
1
2
3
4
5
6
7Ymeasured vs. Ypredicted
Predicted Values
Refe
rence V
alu
es
N= 10y= -0.019098+0.61228xR square= 0.99995biais= 1.0452RMSE= 1.2073
0 1 2 3 4 5 6 70.5
1
1.5
2
2.5
3
3.5
4Ymeasured vs. Ypredicted
Predicted Values
Refe
rence V
alu
es
Regression of fat prediction Master/Slave before PDS
Regression of fat prediction Master/Slave after PDS
Test: application of fat prediction
Results
Prédiction Foss : 4.3%
Prédiction Bentley : 4.3%
Possibility to create and use
models on all apparatus
Example of first tools:
84 Delta spectra
368 Foss spectra
Reference
values (SF6)
+
+
Methane prediction equation
(A.Vanlierde, 2013)
= 452 spectra in a
common format PDS
Application
0
100
200
300
400
500
600
700
800
900
0 100 200 300 400 500 600 700 800
Me
su
red
g C
H4 /
day
Prediction g CH4 / day
N = 452
R²c = 0,76
SECV=69
RPD=1.83
PDS allows to group spectra from all apparatus in a
common database
Conclusion
0 50 100 150 200 250 300-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
2015
Allows to create and use universal
equation on all apparatus
Thanks for your
attention !
0 50 100 150 200 250 300-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
WITH SUPPORT OF
http://www.optimir.eu