Strategies for the use of
Molecular Data
for Genetic Improvement
of Livestock
Jack Dekkers
Past and Current
Selection Strategies
selection
Black box of
Genes
Quantitative genetics
h2
Phenotype
B
LU
P
Environment
Estimated
Breeding
Value
Phenotype
of relatives
Molecular Genetics
““In
In Search of the Holy Grail
”
Grail”
M Q
M Q
M Q
m q
m q
m q
Major genes
Q
uantitative
Quantitative
T
rait
Trait
L
oci ((QTL)
QTL)
Loci
= position (locus) on
genome associated
with genetic
differences for a
quantitative trait
Use of Molecular Data in Selection
Unknown
genes
Molec.
Phenotypic
data
?
genetics
Genes
QTL
Phenotype
EBV
Molecular
data
Selection
strategy
Markerbased EBV
Advantage of Molecular
Genetic data for selection
Genes
Molecular
genetics
• Heritability of genotypes = 1
QTL
• Expressed in both sexes
• Expressed at early age
• Requires less phenotypic data
From Molecular Genetics
to
Marker
-Assisted Selection
Marker-Assisted
Four Key Areas of R&D:
1) Molecular genetics
- molecular markers
- genetic maps
2) QTL or major gene detection
3) Use of gene / marker data
in breeding value estimation
4) Use of gene / marker data in selection
Outline
• Strategies and designs for QTL mapping
• QTL mapping in breed crosses
• Within-breed QTL mapping
• Half-sib designs
• Linkage disequilibrium mapping
• candidate genes
• high-density SNP panels
• Marker-assisted genetic evaluation
• Use of markers in selection
• MAS
• Marker-assisted composite line development
• Genomic selection
• MAS for commercial crossbred performance
Suggested Reading
• USDA
-NSIF QTL mapping and MAS conference
USDA-NSIF
2003 NSIF conference proceedings:
http://www.nsif.com/Conferences/2003/contents.html
••Dekkers
Dekkers and van der Werf (2007) Chapter 10 at
http://www.fao.org/docrep/010/a1120e/a1120e00.htm
Principle of QTL detection
Mean
weight
105
(kg)
M Q
M Q
100
M Q
m q
95
m q
m q
1. Genotype pigs for
marker(s
marker(s))
2. Test for association
between marker
genotype and
phenotype
Presence of association
requires not only
linkage but also
linkage disequilibrium
between marker and
QTL
Processes that create LD
Mutation
M Q
Selection
Crossing
M Q
M Q M
M QQ
M Q
M MQ q
M Q
m q
M Q
m q
Inbreeding/drift
m q
m q
X
m q
m q
m q
M Q
m q
M Q
M Q
M Q
m q
m q
M Q
m q
m q
m q
Measure of LD
r = correlation
betw. Alleles
11
00
01
10
00
11
01
r2
LD mapping in outbred populations
requires markers close to QTL
For there to be sufficient LD
1
0.8
1
r=.001
r=.01
r=.05
r=.1
r=.2
0.9
0.8
0.7
0.6
0.6
0.5
0.4
0.3
0.2
0.4
0.1
0
0
5
10
15
20
25
Generation
0.2
0
0
10
20
30
40
50
Distance (cM)
60
70
80
90 100
Example LD in outbred population
1
1
c=.001
c=.01
0.9
0.8
0.7
c=.05
0.6
0.5
0.9
0.4
c=.2
0.3
0.2
c=.1
0.1
0
0.8
0
5
10
15
Generation
0.7
r-squared
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5000
10000
15000
20000
25000
30000
Distance (kb)
35000
40000
45000
50000
20
25
But LD always exists within families
Half-sib family QTL mapping design
r = 0.2
M
Q
Sire
m
Half
-sib
Half-sib
Progeny
M
M
Q
0.4
1/ (1-r)=
2
Freq.
q
m
meiosis
M
q
1/ r
2
=0.1
m
m
q
0.4
1/ (1-r)=
2
Freq.
Q
0.1
1/ r=
2
Prob(Q|M received from sire) = 0.8
Î Marker and QTL are in LD among progeny
Strategies for QTL mapping in livestock
Outbred population
Breed cross
c =.001
c =.01
LD
c =.05
c =.2
0
c =.1
5
10
15
20
25
Generation
LD used
Recomb.
LD extent
Marker map
Coverage
Map resol.
Linkage analysis
LD markers
Linkage analysis
LE markers
F2 / BC
Cand.
families pedigree genes
AIL
HS/FS
Ext.
LD mapping
LD markers
High
density
QTL detection in F2 breed cross
Genome Scan for Pork Quality
Massoud Malek
Hauke Thomsen
Jong-Joo Kim
Hong-hua Zhao
Max Rothschild
Rohan Fernando
Jack Dekkers
Berkshire
x
F2 cross
Yorkshire
Cross creates extensive
Linkage Disequilibrium
M Q
M Q
m q
M Q M
M QQ
M Q
M MQ q
M Q
m q
M Q
1
m q
r=.001
r=.01
r=.05
r=.1
r=.2
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0
5
10
15
Generation
20
25
m q
m q
M Q
M Q
m q
m q
m q
m q
m q
M Q
M Q
0.1
0
X
m q
m q
M Q
m q
10-20 cM
marker
distance
sufficient
F0 2 Berkshire sires
M1 N1
BB
9 Yorkshire dams
YY M2 N2
x
M1 N1
F1
F2
M2 N2
BY
8 sires
BY 26 dams
x
M1 N1
M1 N1
M2 N2
M2 N2
525 BB
Breed origin
probabilities
BY
YB YY
M1 N1
M1 N1
M2 N2
M1 N1
M2 N2
M2 N2
PBB
PBBYY
PYYB
B
PYY
derived for a given position
SSC1 MARBLING
4.5
-logP
4.0
3.5
http://www.animalgenome.org/QTLdb/pig.html
Line-Cross
BB = - 0.13
BY = +0.19
YY = +0.13
1% Chr.w
Malek et al. 2001
3.0
2.5
5% Chr.w
2.0
1.5
1.0
0.5
Detect
QTL
that
differ
in
0.0
0 10 20 30 40 between
50 60 70 80 breeds
90 100 110 120 130
cM frequency
Breed
cross
QTL
scan
F0 2 Berkshire sires
BB
F1
F2
8 sires
525 BB
BY
BY
9 Yorkshire dams
YY
x
x
BY 26 dams
YB YY
QTL that differ
Î in frequency
between breeds
Î Wide QTL region
(20-50 cM)
Æ LD-markers for - introgression
- composite development
(see later)
BUT: Within-breed MAS requires QTL that
segregate within breeds
Follow-up within-breed research in QTL region
- QTL mapping
- candidate genes
Within-breed QTL mapping
Half-sib family design
r = 0.2
M
Q
Sire
m
Half
-sib
Half-sib
Progeny
M
M
Q
0.4
1/ (1-r)=
2
Freq.
q
m
meiosis
M
q
1/ r
2
=0.1
m
m
q
0.4
1/ (1-r)=
2
Freq.
Q
0.1
1/ r=
2
Prob(Q|M received from sire) = 0.8
Î Marker and QTL are in LD among progeny
Halfsib design for QTL detection and MAS
M
m
Mm
M
M
M
M
M
M
M
M
M
m
m
m
m
m
m
m
m
m
Compare production
Problem: within-family LD is not consistent
across families
Sire 1
Sire 2
Sire 3
Sire 4
M
M
M
M
Q
m q
q
m Q
Q
m Q
q
m q
Î Analysis must allow for
different marker
-QTL linkage phases within each family
marker-QTL
Linkage phase = assortment of alleles into haplotypes
Sire 1 has genotypes Mm and Qq
Qq;;
haplotypes MQ and mq
Within breed QTL detection
Î Half-sib analysisÆ confirm w/in breed segregation of QTL
Evans et al. (2003 Genetics:621) - confirmed QTL in 10 commercial lines
- Does not narrow QTL region
Î LD mapping or association analysis
– find markers close enough to the QTL,
such that associations are consistent
across the population (pop-wide LD)
Î Candidate gene approach
Î High density SNP panels
Example LD in outbred population
1
1
c=.001
c=.01
0.9
0.8
0.7
c=.05
0.6
0.5
0.9
0.4
c=.2
0.3
0.2
c=.1
0.1
0
0.8
0
5
10
15
Generation
0.7
r-squared
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5000
10000
15000
20000
25000
30000
Distance (kb)
35000
40000
45000
50000
20
25
A Revolution in Molecular Genetic Technology
2.8 million SNPs
Nature 2004
Intl.Chicken SNP Cons.
High-through-put
SNP genotyping
SNPs
QTL
Genome-wide LD-Analysis within a breed
High-density
SNP map
(1 SNP/<1 cM)
Genotype +
phenotype > 500
animals
LD mapping
Detect QTL / Estimate effects
Statistical methods for LD mapping
• Genotype methods
• LS / MM with fixed marker effects
• Single SNP
y = biMi
+u +e
Mi = # 1 alleles
u~A
σg2
Aσ
(0/1/2)
• Multi-SNP
y = biMi + bi+1Mi+1 + u + e
10011001001100110100
01111001001001011010
j32
00100111001000010111
00111011001101101110
01101000001001100010
00011001010001000111
j33
11101001001011101111
01011000001001101010
j34
• Maximum Likelihood
• Single SNP – mixture distribution
j31
• Multi-SNP – multiply single-SNP likelihoods (Farnir et al. 2002,
Abdallah et al. 2004)
• Haplotype methods
Zhao et al. 2007
Genetics
Slide 27
j31
j32
Composite likelihood
jdekkers, 8/7/2006
Long and Langley 1999
Fan and Xiong 2002
jdekkers, 8/7/2006
j33
Or using any combination of markers, as implemented by Bonnen et al. (Nat. Genet 38 2006)?
Found not to be better by Hong-hua - threshold more stringent because of larger # tests.
jdekkers, 8/7/2006
j34
Mixture distribution for presumed biallelic QTL
jdekkers, 8/7/2006
Statistical methods for LD mapping
• Haplotype methods
j35
• LS / MM with fixed marker effects
y = Xb + u + e
b’ = [μ00 , μ01 , μ10 , μ11]
00100111001000010111
00111011001101101110
j36
j37
11101001001011101111
01011000001001101010
• IBD Mixed Model (Meuwissen & Goddard 2000)
y = ZgQ + u + e
gQ ~ N(0,VσQ2)
j4
01101000001001100010
00011001010001000111j3
• Max. Likelihood / Bayesian
• Perez-Enciso (2003)
• Boitard et al. (2006 BMC Genomics)
10011001001100110100
01111001001001011010
V = IBD matrix
Cov. from Prob(IBD at Q | marker haplos)
Calculated using genedrop (Maccluer et al. 1986)
j3
Slide 28
j35
Long and Langley 1999
Grapes et al. 2005
jdekkers, 8/7/2006
j36
MCMC to obtain joint distribution for multiple-marker haplotypes - avoids approximating composite likelihood by multiplication,
assuming independence
jdekkers, 8/7/2006
j37
Used 3-locus Wright-Fisher model to derive approximate expressions for expected haplotype frequencies - only uses flanking markers
- shown to do better than composite likelihood approach
jdekkers, 8/7/2006
j38
Allow for heterogeneity of allele effects at the QTL but derivation of IBD probabilities assume a single mutation - single mutation
occurred a pre-specified # generations ago in a population of certain Ne since then.
M+G however showed that mapping precision was little affected by deviations from these assumptions
jdekkers, 8/7/2006
j39
j40
Assume single mutation responsible for QTL - parameters can be fitted to account for allelic heterogeneity
jdekkers, 8/7/2006
require no assumptions about population history of origin of the mutation
jdekkers, 8/7/2006
1
Summary of QTL mapping strategies
r=.001
r=.01
0.9
0.8
0.7
0.6
r=.05
r=.1
r=.2
0.5
0.4
0.3
0.2
Linkage
analysis
0.1
0
0
5
10
15
Outbred population
Breed cross
20
Generation
25
F2 / BC
AIL
Assoc.analysis
LD mapping
Cand.
families pedigree genes
HS/FS
Ext.
LD used
Population wide
in cross
Recomb.
1 rnd
>1 rnd
1 rnd
>1 rnd
LD extent
Long
Smaller
Long
Smaller
Denser
Sparse
Marker map Sparse
Coverage
Map resol.
Genome wide
Poor
Better
Population wide
Within family
Poor
>>> 1 round
Small
Denser Few loci
Genome wide
Better
High
density
Local
Dense
Genome
High
LD-LA analysis
Summary QTL mapping
“ Breed cross QTL scan
Î QTL that differ between breeds
Î Wide QTL regions
Î LD-markers within cross - use for introgression,
composite development
“ Within-breed QTL scan
Î QTL that segregate within breeds
Î Wide QTL regions
Î LE-markers - difficult for MAS
“ Candidate gene approach
Î LD-markers for within-breed MAS
“ High-density genotyping LD mapping
Î LD-markers for within-breed MAS
So you
’ve found a QTL
you’ve
........
Now what ??
Use of Molecular Genetics in
Breeding Programs
Jack Dekkers
Department of Animal Science
Center for Integrated Animal Genomics
QTL detection Æ 3 types of selectable loci
Direct markers
LD
-markers
LD-markers
Functional mutations
- known genes
MQ
MQ
mq
MQ
mq
mq
In pop.-wide Linkage Equilibrium
with mutation
Linkage phase NOT consistent across families
Sire 2
Sire 1
Sire 3
M Q
M q
M Q
m q
q
In pop.-wide Linkage Disequilibrium
with mutation
Linkage phase
~consistent
across population
LE
-markers
LE-markers
Q
m Q
m Q
Sire 4
M q
m q
Outline
• Strategies and designs for QTL mapping
• QTL mapping in breed crosses
• Within-breed QTL mapping
• Half-sib designs
• Linkage disequilibrium mapping
• candidate genes
• high-density SNP panels
• Marker-assisted genetic evaluation
• Use of markers in selection
• MAS
• Marker-assisted composite line development
• Genomic selection
• MAS for commercial crossbred performance
Genes and LD-Markers
Linkage phase tends to be consistent
across families and generations
MQ
MQ
mq
MQ
mq
mq
Include marker genotype(s) as fixed effect in
animal model
y = marker genotype + a + e
ƒ
ƒ
Estimate effects in population under selection
Must be re-estimate on a regular basis
LE-markers
Linkage phase not consistent between
sires
Sire 1
Sire 2
Sire 3
Sire 4
M
M
M
M
Q
m q
q
m Q
Q
m Q
QTL effect must be estimated for each
individual/family
•
Based on family information
•
•
marker genotypes
phenotypes
q
m q
LE-markers
Marker-assisted BLUP
(Fernando and Grossman, 1989, GSE)
Sire
Dam
Ms Qsp
Md Qdp
Ms Qsm
Md Qdm
Progeny
Ms Qip
Md Qim
^ p + v^ m + u^
Total EBV = v
i
i
yi = μ + vip + vim + u + e
Paternal / Maternal PolyQTL allele effect
genic
Var(u
σ u2
Var(u)) = A
Aσ
Var(v
σv2
Var(v)) = G
Gσ
G = gametic relationship matrix
for QTL effects
Computed from
- marker genotypes
- marker
-QTL rec. rate
marker-QTL
Examples of
gene tests in
commercial
breeding
Trait
Congenital
defects
Appearance
Direct marker
BLAD (D)
Citrulinaemia (D,B)
DUMPS (D)
CVM (D)
Maple syrup urine (D,B)
Mannosidosis (D,B)
RYR (P)
CKIT (P)
MC1R/MSHR (P,B,D)
MGF (B)
κ-Casein (D)
β-lactoglobulin (D)
FMO3 (D)
RYR (P)
RN/PRKAG3 (P)
Milk quality
D = dairy cattle
B = beef cattle
C = poultry
P = pigs
S = sheep
Dekkers, 2004, JAS
“MAS has seen
limited
application”
Meat quality
LD marker
LE marker
RYR (P)
Polled (B)
RYR (P)
RN/PRKAG3 (P)
A-FABP/FABP4 (P)
H-FABP/FABP3 (P)
CAST (P, B)
>15 PICmarqTM (P)
THYR (B)
Leptin (B)
Feed intake
Disease
Reproduction
MC4R (P)
Prp (S)
F18 (P)
Booroola (S)
Inverdale(S)
Hanna (S)
Growth &
composition
Milk yield &
composition

MC4R (P)
IGF-2 (P)
Myostatin (B)
Callipyge (S)
DGAT (D)
GRH (D)
κ-Casein (D)
B blood group (C)
K88 (P)
Booroola (S)
ESR (P)
PRLR (P)
RBP4 (P)
CAST (P)
IGF-2 (P)
QTL (P)
QTL (B)
Carwell (S)
PRL (D)
QTL (D)
Outline
• Strategies and designs for QTL mapping
• QTL mapping in breed crosses
• Within-breed QTL mapping
• Half-sib designs
• Linkage disequilibrium mapping
• candidate genes
• high-density SNP panels
• Marker-assisted genetic evaluation
• Use of markers in selection
• MAS
• Marker-assisted composite line development
• Genomic selection
• MAS for commercial crossbred performance
Potential Gains from MAS
Meuwissen & Goddard, 1996
QTL with 1/3 of genetic variance haplotype
-marked
haplotype-marked
(%)
Extra response fr om MAS
h2=.27
64
70
62
55
60
50
38
38
40
39
37
31
30
30
25
21
20
15
9
10
5
0
1
1
2
Generation
4
2
2
3
3
5
4
Carcass trait
Sex
-limited trait
Sex-limited
Phenotyping after selection
Phenotyping before selection
Potential Gains from MAS
Effect of Heritability ((Meuwissen
Meuwissen & Goddard 1996)
S (%)
Ext ra respons e from MA
Single marked QTL with 1/3 of genetic variance
70
60
45
50
36
38
34
40
30
23
25
30
21
15
17
20
13
9
10
6
5
4
0
1
2
1
2
Generation
2
3
3
5
4
h2=.11
h2=.27 Phenotyping after
h2=.11
Phenotyping before
h2=.27
Potential Gains from MAS
Effect of QTL effect ((Meuwissen
Meuwissen & Goddard 1996)
70
60
47
40
50
33
40
29
25
23
30
19
20
13
12
12
10
10
7
5
5
4
0
1
4
1
2
3
Generation
2
3
4
5
va QTL
r ia
nc e
(%
)
(% )
Extra re sponse fr om MAS
Phenotyping after selection. Heritability=32%
46.7
26.7
13.3
6.7
Marker
-assisted composite line development
Marker-assisted
Line 1
Direct use
of QTL
detected
in cross
x
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
Line 2
F3 x F3
5
10
15
20
25
Generation
10-20 cM
marker
distance
sufficient
F1 x F1
F2 x F2
c=.001
c=.01
c=.05
c=.1
c=.2
0.9
QTL detection
Estimation of marker effects
y = marker genotype + BV + e
MAS
MAS
.
.
Pyiasatian et al. 2007
Theor
Theor.. Appl. Genetics
Composite of inbreds
QTL detection, MA-Evaluation,
and Selection
QTL
Model: yjk = generationk + ∑ β i * MSijk + u jk + e jk
i =1
90
∑ βˆ MS
i =1
i
ijk
# marker alleles
from line 1 in interval i
Significant intervals
QTL
Selection on
)
∑ β * MS
i =1
i
ijk
+ uˆ jk
Pyiasatian et al. 2007
Theor
Theor.. Appl. Genetics
Incorporating LD markers in MAS
Genome-scan
Detect QTL / Estimate effects
MAS on Significant
regions only
MAS on all
regions
Genomic
selection
(Meuwissen et al. ‘01)
Pyiasatian et al. 2007
Theor
Theor.. Appl. Genetics
Marker-Assisted Evaluation
and Selection
90
Model: yjk = generationk + ∑ β i * MSijk + e jk
i =1
90
∑ βˆ MS
i =1
i
# marker alleles
from line 1 in interval i
ijk
All intervals
90
Selection on
)
∑ β * MS
i =1
i
ijk
βi fixed or
random
Genomic
selection
(Meuwissen et al. ‘01)
Pyiasatian et al. 2007
Theor
Theor.. Appl. Genetics
50% -ve QTL Effects, 20 cM
200
Cumulatative Response
(% over BLUP)
R11
150
F11
Known QTL
100
Select on
significant
markers only
50
BLUP
0
BLUP
-50
F3
F4
F5
F6
F7
F8
Generation
F9
F10
F11
Pyiasatian et al. 2007
Theor
Theor.. Appl. Genetics
Genomic Selection in outbred population
Meuwissen et al. (2001)
Genotype
Many SNPs
Genotype
for SNPs
Genotype
for SNPs
MODEL
to predict
phenotype/BV
from SNP
genotypes
Predict
BV or
phenotype
A ccu racy
Phenotype
1.0
0.8
0.6
3
4
5
6
Generation
7
8
Genomic Selection in outbred population
Meuwissen et al. (2001 Genetics: 1819)
Mixed model for prediction of
haplotype effects
yi = μ + Σ(hijp + hijm) + ei
or of allele effects
yi = μ + Σβigij + ei
Ne = 100
Estimates from 2200 individuals
EBV accuracy
Marker distance
0.85
1 cM
0.81
2 cM
0.75
4 cM
0.9
0.8
0.7
0.6
500
1000
2200
# individuals
1
Accuracy
random (Bayesian)
Accuracy
1
0.9
0.8
0.7
0.6
1
2
3
4
5
Generation
6
7
8
Response from Genomic Selection - Simulation
Hong-hua Zhao, Jennifer Young, David Habier, Rohan Fernando, Jack Dekkers
Generation
0
.
.
.
.
1000
1001
1002
.
.
1012
20 chr of 150 cM
100,000 SNPs (freq. = ½ , LE)
Random mating, Ne =100
LD generated by drift and mutation
Allocate 100 loci with MAF>0.1 as QTL and 2000 as SNPs
Expand pop.size to 1000 – phenotype – h2=0.3
Estimate marker effects by Bayes-B
Mate random 20 males to random 60 females
Select 20/240 males
.
.
Select
60/240 females
Response
6
100 QTL GS-1
100 QTL GS-all
5
100 QTL BLUP-1
GS
-all
GS-all
100 QTL BLUP-all
4
BLUP
-all
BLUP-all
3
GS
-1
GS-1
2
1
BLUP
-1
BLUP-1
0
1001
1002
1003
1004
1005
1006
1007
1008
Generation
1009
1010
1011
1012
1013
Inbreeding
0.30
100 QTL GS-1
100 QTL GS-all
100 QTL BLUP-1
100 QTL BLUP-all
0.25
BLUP
-all
BLUP-all
0.20
GS
-all
GS-all
BLUP
-1
BLUP-1
0.15
GS
-1
GS-1
0.10
0.05
0.00
1001
1002
1003
1004
1005
1006
1007
1008
Generation
1009
1010
1011
1012
1013
Proportion of QTL fixed for favorable allele
0.20
GS
-all
GS-all
100 QTL GS-1
100 QTL GS-all
BLUP
-all
BLUP-all
100 QTL BLUP-1
0.15
100 QTL BLUP-all
0.10
GS
-1
GS-1
0.05
BLUP
-1
BLUP-1
0.00
1001
1002
1003
1004
1005
1006
1007 1008
Generation
1009
1010
1011
1012
1013
Proportion of QTL fixed for unfavorable allele
0.12
0.10
100 QTL GS-1
100 QTL GS-all
100 QTL BLUP-1
0.08
100 QTL BLUP-all
BLUP
-all
BLUP-all
0.06
GS
-all
GS-all
BLUP
-1
BLUP-1
0.04
GS
-1
GS-1
0.02
0.00
1001 1002
1003 1004 1005
1006 1007 1008
Generation
1009 1010 1011
1012 1013
Genomic Selection (GS)
••GS
GS provides unique opportunities to enhance
breeding programs by:
• increasing / maintaining response to selection
• reducing rates of inbreeding
• reducing loss of favorable alleles
• reducing generation intervals
• limiting need for pedigree
-based phenotypes
pedigree-based
Outline
• Strategies and designs for QTL mapping
• QTL mapping in breed crosses
• Within-breed QTL mapping
• Half-sib designs
• Linkage disequilibrium mapping
• candidate genes
• high-density SNP panels
• Marker-assisted genetic evaluation
• Use of markers in selection
• MAS
• Marker-assisted composite line development
• Genomic selection
• MAS for commercial crossbred performance
Current Pyramid Selection Programs
Limitations: - limited selection for performance in the field
- no selection for traits not recorded in nucleus
- disease traits
Pu
r eb
r ed
da
ta
NUCLEUS
herds
Sire
line
Multiplier
Dam
line
Multiplier
Production herds
High health
environment
rg < 1
Field
environment
Selection for Performance in Field
‘Traditional’ Breeding Solution:
Field data on relatives
Collect phenotypes on relatives in field
Æ Combined Crossbred
-Purebred Selection
Crossbred-Purebred
Î
Î
Purebred
data
Sire
line
ΔGfield
ΔF
Bijma & van Arendonk, ‘98
Requirements/limitations:
- Costly logistics - Pedigree-based
phenotyping in field
Multiplier
- Longer generation intervals
- Higher rates of inbreeding
- family data vs. own phenotype
Production herds
Selection for Performance in Field
Possible Molecular Genetic Solution:
Marker
Marker-assisted
EBV
Markerbased
EBV
Identify markers with effects on
performance in the field Æ MAS/GS
MAS
Pu
re
br
ph
ed
en
ot
yp
e
Advantages:
Sire
line
Genotype
Multiplier
SNP effect
estimates
Genotype
Phenotype
- No pedigree
-based phenotypes
pedigree-based
- Reduce generation intervals
- Select for low heritable traits
Current limitations:
- Few useful markers available
- Most detected in purebreds
- Effects not consistent across
breeds and environments
Production herds
Genomic Selection for Field Performance
Potential benefits
Marker
Marker-assisted
EBV
MAS/GS
red
sb e
os yp
Cr enot
ph
Pu
r
ph ebred
en o
t yp
e
Genomic
selection
EBV
Genotype
Genetic correlation
(purebred – commercial)
= 0.7
Sire
line
Multiplier
SNP effect
estimates
Genotype
Phenotype
(Dekkers 2007 JAS)
Production herds
Accuracy of selection
Selection for commercial performance
for
field ofperformance
Accuracy
selection
rpb,cb=0.7
Accuracy
Dekkers
JAS, 2007
by Crossbred Genomic Selection
PPb+
GSXb
+GS
((GS
GSXb)
0.8
GSXb
0.7
0.6
0.5
PPb+ PXbred
0.4
GSPb
PPurebred
0.3
0.2
0.1
0.2
0.3
0.4
0.5
0.6
Accuracy of GS-EBV
0.7
0.8
Impact on Inbreeding
ΔF (%)
Dekkers
JAS, 2007
Crossbred Genomic Selection
((GS
GSXb)
3.5
PPb+ PXbred
3
2.5
PPurebred
2
1.5
PPb+
GSXb
+GS
1
GSXb
0.5
0.2
0.3
0.4
0.5
0.6
Accuracy of GS-EBV
0.7
0.8
8 Partial LD
Estimation
q
6
6
4
of
marker
q
Q
q
2
effects in
Q
q
q
Q
Q
Q
GG GA AA Sire crosses Dam GG GA AA
10
Complete LD
GQ
Aq
line
line
Multiplier
Multiplier
GQ
Aq
Production herds
9
Marker effects
differ between
purebreds vs.
crossbreds
q
6.5
Q
Q Q
Q q
5.5
q
q Q
3
Reasons:
Q
q q
GG GA AG AA
Dominance
Epistasis
GxEnvironment
Marker-QTL LD
Genomic selection
in crossbred
populations
Methods
Founders
1 M chromosome
500 --6000
6000 SNPs
100 QTL
h2=0.3
Bayes
-B
Bayes-B
1000 generations
Ne = 500
Ali Toosi, Noelia Ibañez,
Rohan Fernando,
Jack Dekkers
Breed A
Breed B
50 generations
Ne = 100
50 generations
Ne = 100
Training data sets
N=1000
Breed A
Mix A+B
NRI Award
2007
-35205-17862
2007-35205-17862
Breed C
F1 AxB
C(AB)
F2 AxB
Breed B
8 generations
Ne = 100
Breed B
validation
Accuracy of GS based on
admixed training populations
0.9
#
markers
on 1 M
chrom
chrom..
500
2000
6000
0.8
0.6
0.5
0.4
0.3
0.2
0.1
(A
B)
(C
D)
(A
B)
C
AM
X
(A
B)
(A
B)
AB
A+
B
A
0
B
Accuracy
0.7
Training populations
Correlation of LD between breeds at 0.2 cM
64
Partial LD
MAS/GS
q
q
Q
q
for field
Q
q
q
Q
Q
Q
perforGG GA AA Sire mance Dam GG GA AA
Complete LD
line
line
Multiplier
Multiplier
Need to estimate breed-specific SNP effects?
Production herds
q
Q
Q Q
Q q
q
q Q
Q
q q
GG GA AG AA
Across-breed
Breed-specific
effects
1
0.9
Across-breed
Breed-specific
0.8
Accuracy
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Breed 0
separation 50
550
infin
50
550
infin
500 markers
2000 markers
1000 phenotypic records
50
550
infin
50
550
infin
500 markers
2000 markers
4000 phenotypic records
Summary / Conclusions
HD SNP genotyping offers unique opportunities
to enhance animal breeding programs
by removing limitations on when, where, and on whom
phenotypes are recorded
• Genomic selection of purebreds for field performance
of crossbreds
• With opportunities to
• reduce generation intervals
• reduce inbreeding
• Detect QTL for animal health
• Select for animal health
Simulation results look very promising
Empirical results are becoming available
Concluding remarks
Recent advances in genotyping technology and
reduced costs have increased opportunities for MAS
• Use of LD- instead of LE-markers
• Use all vs only significant markers (?)
• genomic selection
• MAS on commercial performance
• Increases response
• Reduces rate of inbreeding
• Requires continued emphasis on phenotypic recording
• Requires careful economic analysis
Implementation of MAS requires comprehensive
approach Business
R&D
objectives
Farms
Quality
DNA collection
Phenotyping
l
o
r
t
Con
Pedigree
n
D e c i s i oPhenotypic
Database
Support
Genotyping
Genotypic
Database
Analytical
tools
Selection
Acknowledgements
Rohan Fernando
Sue Lamont
Max Rothschild
Honghua Zhao
David Habier
Cristina Andreescu
Napapan Piyasatian
Jennifer Young
Ali Toosi
Noelia Iba
ñez
Ibañez
Abebe Hassen