Post on 01-Apr-2015
GxE in commercial pig breedingreaction norms
selection for the response environment
Pieter KnapGenus-PIC
Selection of genotypes for a particular production environment
Between linesrelatively straightforward
Within-linemuch more interesting
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Selection between linesrelatively straightforward: usually few lines to choose from
Selection of genotypes for a particular production environment
Within-line selectionmuch more interesting: continuous variation to choose from
Rischkowsky & Pilling (2007)
Anderson (2004) after Haldane (1946)
0.60
0.62
0.64
0.66
0.68
0.70
aver
age
daily
gai
n (k
g /
d)
very highinfectiousness
very lowhigh low
Schinckel et al. (1999)
Poster: Antti Kause
Anderson (2004) after Haldane (1946)
Within-line selectionmuch more interesting:
continuous variation to choose from
Within-line selectionmuch more interesting:
continuous variation to choose from
Anderson (2004) after Haldane (1946)
0.60
0.62
0.64
0.66
0.68
0.70
aver
age
daily
gai
n (k
g /
d)
very highinfectiousness
very lowhigh low 0.62 0.64 0.66 0.68 0.70
treatment mean: average daily gain (kg / d)
0.60
0.62
0.64
0.66
0.68
0.70
average daily gain (kg / d)
y = 0.30 + 0.57 x
y = –0.30 + 1.43 x
Schinckel et al. (1999)
Within-line selectionmuch more interesting:
continuous variation to choose from
Anderson (2004) after Haldane (1946)
E > I : incentive to improve the environment
I > E : incentive to match genotype to environment
• Select in the response envrmnt
• Select on data from theresponse environment
Knap & Su (2008)
Knap & Su (2008)
Individual reaction norms
intercept : the conventional EBVfor productivity(when they differ, the trait is heritable)
slope :the EBV for environmental sensitivity of productivity(when they differ, the trait shows GxE)
two breeding goal traits
environment
phen
otyp
e
EN
PN
EH
PH
b
PC
EC
EL
PL
EN
PN
PC = PN – b × ( EN – EC )
selection environment
response environment
EH
PH
b
PC
EC
EL
PL
EN
PN
PC = PN – b × ( EN – EC )
average performance in commercial conditions:= the breeding goal trait
genetic potential
environmental sensitivity
how far away is the nucleus from the
commercial level ?
P = WT × KO × [Vcarcass+ LEAN × Vlean]
– DAYS120 × [Cday + ADF × Cfeed ]
P = WT × KO × [Vcarcass+ LEAN × Vlean]
– [ PN, DAYS – bDAYS × (DAYSN – DAYSC) ] × [Cday + ADF × Cfeed ]
Set up the profit equation to derive economic values
Two breeding goal traits
Differentiate to derive marginal economic values
MEV(PN, DAYS) = dP / dPN, DAYS = – [Cday + ADF × Cfeed ]
P = WT × KO × [Vcarcass+ LEAN × Vlean]
– [ PN, DAYS – bDAYS × (DAYSN – DAYSC) ] × [Cday + ADF × Cfeed ]
MEV(bDAYS) = dP / dbDAYS = (DAYSN – DAYSC) × [Cday + ADF × Cfeed ]
= – (DAYSN – DAYSC) × MEV(PN, DAYS)
Differentiate to derive marginal economic values
MEV(bDAYS) = dP / dbDAYS = (DAYSN – DAYSC) × [Cday + ADF × Cfeed ] =
= – (DAYSN – DAYSC) × MEV(PN, DAYS)
The MEV of the environmental sensitivity depends on
• the MEV of the trait as such• the distance selection environment response environment
Differentiate to derive marginal economic values
MEV(PN, DAYS) = – [Cday + ADF × Cfeed ] =
= – [0.24 + 2.3 × 0.29 ] = –0.16 € per d
MEV(bDAYS) = – (DAYSN – DAYSC) × MEV(PN, DAYS) =
= –(163 – 179) × –0.16 = –2.56 € per d/d
Negative MEV : a reduction of DAYS120 means faster growth
Negative MEV : a reduction of the slope brings commercial performance closer to the potential
Individual reaction norms
intercept : the conventional EBVfor productivity(when they differ, the trait is heritable)
slope :the EBV for environmental sensitivity of productivity(when they differ, the trait shows G×E)
two breeding goal traits
An elegant option to deal with G×E on the individual level:
Calculate sensitivity EBVs, and include them in the index, weighted by the MEV as usual.
is that feasible?
Line B; parity 1 only
66 farms with 33.641 records of33.641 daughters of 792 sires
Line B; all parities
93 farms with 73.352 records of52.120 daughters of 1091 sires
Lines A, B and AB; all parities
144 farms with 346.030 records of121.104 daughters of 2040 sires
Litter size: daughter group reaction norms
sires sires sires
Line B; parity 1 only
66 farms with 33.641 records of33.641 daughters of 792 sires
Line B; all parities
93 farms with 73.352 records of52.120 daughters of 1091 sires
Lines A, B and AB; all parities
144 farms with 346.030 records of121104 daughters of 2040 sires
Litter size reaction norms of sires: standard error of slope vs. HYS environmental range
Line B; parity 1 only
66 farms with 33.641 records of33.641 daughters of 792 sires
Line B; all parities
93 farms with 73.352 records of52.120 daughters of 1091 sires
Lines A, B and AB; all parities
144 farms with 346.030 records of121104 daughters of 2040 sires
sires sires sires
Litter size reaction norms of sires: standard error of slope vs. number of daughters
sires siressiressiressires siressiressiressires
Line B; parity 1 only
66 farms with 33.641 records of33.641 daughters of 792 sires
Line B; all parities
93 farms with 73.352 records of52.120 daughters of 1091 sires
Lines A, B and AB; all parities
144 farms with 346.030 records of121104 daughters of 2040 sires
Litter size reaction norms of sires: standard error of slope vs. slope
h2 rG
intcpt 10 26±7
slope 8±3
h2 rG
intcpt 9 69±5
slope 2±0.4
h2 rG
intcpt 10 –9±15
slope 15±8
Knap & Su (2008)
Line B; parity 1 only
66 farms with 33.641 records of33.641 daughters of 792 sires
Line B; all parities
93 farms with 73.352 records of52.120 daughters of 1091 sires
Lines A, B and AB; all parities
144 farms with 346.030 records of121.104 daughters of 2040 sires
Litter size: daughter group reaction norms
E > I > G
I > E > G
?Same data (Line B; all parities) analyzed with SAS
E > I : incentive to improve the environment
I > E : incentive to match genotype to environment
• Select in the response envrmnt
• Select on data from theresponse environment
?
Individual reaction norms
intercept : the conventional EBVfor productivity(when they differ, the trait is heritable)
slope :the EBV for environmental sensitivity of productivity(when they differ, the trait shows G×E)
two breeding goal traits
An elegant option to deal with G×E on the individual level:
Calculate sensitivity EBVs, and include them in the index, weighted by the MEV as usual.
is that feasible?
Not for pigs, today
The individual reaction norm approach is notfeasible for commercial pig breeding, today
Simplify
Most extreme:
E as a continuous variable (= reaction norms)
two E classes (e.g. nucleus & commercial)
…or anything in between
Poster: Ann McLaren et al.Poster: Anna-Maria Tyrisevä et al.
Van Sambeek (2010)
Reciprocal Recurrent Selection
Commercial Sibling Test
Combined Crossbred & Purebred Selection
• Standal (1968)
• McNew & Bell (1971)
• Biswas et al. (1971)
• Wei Ming & Van der Werf (1994)
• Baumung et al. (1997)
• Bijma & Van Arendonk (1998)
• Spilke et al. (1998)
• Misztal et al. (1998)
• Dekkers & Chakraborty (2004)
Theory:
… grown on commercial farms
An example: PIC's GN-Xbred program
• after that, semen is
used for GN
matings
• semen of GN boars is first used
on crossbred sows
multiplication
commercial crossbred sows
GN
commercial crossbred slaughter pigs
crossbred progeny
purebred progeny
An example: PIC's GN-Xbred program
multiplication
commercial breeding stock
GN
commercial crossbred slaughter pigs
PICTraqDatabase
selection decisions
CBVs
GN progeny performance data
Commercial progeny performance data
Commercial sowperformance data
• crossbred halfsib performance
CBVs of GN selection
candidates
crossbred halfsibs of purebred
GN selection candidates
• Xbred sow performance
CBVs of GN selection
candidates
GN-Xbred logistics
sire lines
dam lines
Reciprocal Recurrent Selection
Commercial Sibling Test
Combined Crossbred & Purebred Selection
Is this useful?
Depends on the coheritability
• ΔGC|N ~ hC × rG (C,N) × hN
• ΔGC|C ~ hC × hC
• is hC > rG (C,N) × hN ?
is rG (C,N) low enough ?
what about hN vs hC ?
• !! effective heritabilities !!
The crucial aspects :Can the trait be recorded at all in nucleus conditions ?
And on how many animals ?
• Cecchinato et al. (2010): stillbirth rate rG = 0.25 ± 0.34
• Bosch et al. (2000): litter size 0.40 < rG < 0.59
• Zumbach et al. (2007): ADG 0.53 < rG < 0.80; BFT and LMD 0.78 < rG < 0.89
• Ibáñez-Escriche et al. (2011): lean percentage 0.81 < rEBV < 0.96
• Brandt & Täubert (1998): ADG and BFT 0.87 < rG < 1.0
• Standal (1968)
• McNew & Bell (1971)
• Biswas et al. (1971)
• Wei Ming & Van der Werf (1994)
• Baumung et al. (1997)
• Bijma & Van Arendonk (1998)
• Spilke et al. (1998)
• Misztal et al. (1998)
• Dekkers & Chakraborty (2004)
Theory:
ADG
ADG
BFD
BFD
DFI
DFI
RFI
RFI
cros
sbre
d co
mm
erci
al p
erfo
rman
ce
rEBV = 0.55
rEBV = 0.54
rEBV = –0.06
rEBV = 0.06
rEBV = 0.85
rEBV = 0.78
rEBV = 0.85
rEBV = 0.80
crossbred comm
ercial performance
purebred nucleus performanceKnap & Wang (2012)
Poster: Helene Gilbert et al.
cros
sbre
d co
mm
erci
al p
erfo
rman
ce
purebred nucleus performance
crossbred comm
ercial performance
rEBV = 0.33 rEBV = 0.24
grower-finisher mortality rate
Poster: Geir Steinheim et al.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
With xbred data
EB
V A
ccu
racy
• low rG (C,N)
• many more data from C than from N
• much more variation in C :
σ2 = p × (1 – p) and p is much higher
E > I : incentive to improve the environment
I > E : incentive to match genotype to environment
• Select in the response envrmnt
• Select on data from theresponse environment
This is the actual worldwide situation in technified pig production,according to the evidence that I have
E > I : incentive to improve the environment
I > E : incentive to match genotype to environment
• Select in the response envrmnt
• Select on data from theresponse environment
This is what we are targeting,in terms of genetic evaluation:~ "better safe than sorry"
E > I : incentive to improve the environment
I > E : incentive to match genotype to environment
• Select in the response envrmnt
• Select on data from theresponse environment
In better conditions,the better animalsare more better
Genetic variation can be• detected more easily• exploited and valuated
more easily
Incentive for the breeder: more diversity in better conditions improve them
E > I : incentive to improve the environment
Genetic Services: live consultancy at the customer level
Genetic Services:
manuals & documentation
Genetic Services:
manuals & documentation
Genetic Services:
manuals & documentation
Conclusions
• in technified pig production, G×E is probably not dramatic
• individual reaction norms are the perfect way to deal with it
• but statistically very demanding and too data-hungry
• CCPS is a feasible compromise, and it works very well
• improving production conditions (i) improves performance
and (ii) makes the better animals more better
GxE in commercial pig breedingreaction norms
selection for the response environment
Pieter KnapGenus-PIC