ArticlePDF Available

Abstract and Figures

Aim of study: To assess selection methods via introgression to improve litter size in native and synthetic sheep breeds. Area of study: Sanandaj, Kurdistan, Iran. Material and methods: Selection approaches were performed using classical, genomic, gene-assisted classical (GasClassical) and gene-assisted genomic (GasGenomic) selection. Litter size trait with heritability of 0.1 including two chromosomes was simulated. On chromosome 1, a single QTL as the major gene was created to explain 40% of the total additive genetic variance. After simulation of a historical population, the animals from the last historical population were split into two populations. For the next 7 generations, animals were selected for favorable or unfavorable alleles to create distinct breeds of A or B, respectively. Then from the last generation, both males and females from breed B were selected to create a native population. On the other hand, males from breed A and females from breed B were mated to simulate a synthetic population. Finally, intra-population selections were carried out based on high breeding values during the last five generations. Main results: The genetic gain in the synthetic breed was higher than that of the native breed under all selection methods. The frequencies of favorable alleles after five generations in the classical, genomic, GasClassical and GasGenoimc selection approaches in the synthetic breed were 0.623, 0.730, 0.850 and 0.848, respectively. Research highlights: Combining gene-assisted selection with classical or genomic selection has the potential to improve genetic gain and increase the frequencies of favorable allele for litter size in sheep.
Content may be subject to copyright.
RESEARCH ARTICLE OPEN ACCESS
Spanish Journal of Agricultural Research
18 (1), e0403, 11 pages (2020)
eISSN: 2171-9292
https://doi.org/10.5424/sjar/2020181-15459
Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA)
Comparison of different selection methods for improving litter size
in sheep using computer simulation
Meysam Latifi (Latifi, M)1, Amir Rashidi (Rashidi, A)1, Rostam Abdollahi-Arpanahi (Abdollahi-Arpanahi, R)2
and Mohammad Razmkabir (Razmkabir, M)1
1 University of Kurdistan, Faculty of Agriculture, Dept. of Animal Science, Sanandaj, Iran 2 University of Tehran, College of Aburaihan,
Dept. of Animal Science, Tehran, Pakdasht, Iran
Abstract
Aim of study: To assess selection methods via introgression to improve litter size in native and synthetic sheep breeds.
Area of study: Sanandaj, Kurdistan, Iran.
Material and methods: Selection approaches were performed using classical, genomic, gene-assisted classical (GasClassical)
and gene-assisted genomic (GasGenomic) selection. Litter size trait with heritability of 0.1 including two chromosomes was simu-
lated. On chromosome 1, a single QTL as the major gene was created to explain 40% of the total additive genetic variance. After
simulation of a historical population, the animals from the last historical population were split into two populations. For the next 7
generations, animals were selected for favorable or unfavorable alleles to create distinct breeds of A or B, respectively. Then from
the last generation, both males and females from breed B were selected to create a native population. On the other hand, males from
breed A and females from breed B were mated to simulate a synthetic population. Finally, intra-population selections were carried
out based on high breeding values during the last five generations.
Main results: The genetic gain in the synthetic breed was higher than that of the native breed under all selection methods. The
frequencies of favorable alleles after five generations in the classical, genomic, GasClassical and GasGenoimc selection approach-
es in the synthetic breed were 0.623, 0.730, 0.850 and 0.848, respectively.
Research highlights: Combining gene-assisted selection with classical or genomic selection has the potential to improve ge-
netic gain and increase the frequencies of favorable allele for litter size in sheep.
Additional key words: major gene; synthetic breed; genomic evaluation
Abbreviations used: CBLUP (conventional or best linear unbiased prediction); GasClassical (gene-assisted classical selection);
GasGenomic (gene-assisted genomic selection); GPBV (prediction of genomic breeding values); LD (linkage disequilibrium); LS
(litter size); MCMC (Monte Carlo Markov Chain); PBV (prediction of breeding value); QTL (quantitative trait loci); SNP (single
nucleotide polymorphisms); TBV (true breeding value)
Authors’ contributions: Designed the study: ML and AR. Analyzed the data: ML. Wrote the paper: ML, AR. and RAA. Supervised
the work: AR. All authors read the paper and approved the final manuscript.
Citation: Latifi, M; Rashidi, A; Abdollahi-Arpanahi, R; Razmkabir, M (2020). Comparison of different selection methods for
improving litter size in sheep using computer simulation. Spanish Journal of Agricultural Research, Volume 18, Issue 1, e0403.
https://doi.org/10.5424/sjar/2020181-15459
Supplementary material (Fig. S1) accompanies the paper on SJAR’s website.
Received: 13 Jul 2019. Accepted: 02 Apr 2020.
Copyright © 2020 INIA. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0
International (CC-by 4.0) License.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Correspondence should be addressed to Amir Rashidi: arashidi@uok.ac.ir
Introduction
Litter size (LS) is defined as the number of lambs
born per ewe lambing. In sheep, a large variation in
litter size has been observed among and within
breeds. A trait such as LS can be genetically affected
by many genes with small effects as well as major
genes with large effects (Drouilhet et al., 2009). The
major genes frequency is one of the main factors in
improving breeding program efficiency for LS in
sheep (Elsen et al., 1994). Since the heritability of
LS is low (Hanford et al., 2005; Mokhtari et al.,
2010), the selection based on the ewe’s own perfor-
mance might induce a slow genetic gain. In addition,
DNA testing of major genes and learning about in-
heritance patterns shown that major genes could lead
to improvement of ovulation rates and LS in sheep
(Davis, 2005).
Meysam Latifi, Amir Rashidi, Rostam Abdollahi-Arpanahi and Mohammad Razmkabir
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
2
domly mated for 1000 generations with an effective
population size of 1000 animals (500 males and 500
females). At the second step, two random samples
consisting of 330 animals as founders (30 males and
300 females for each breed) were drawn out of the last
generation of historical population to simulate two dif-
ferent breeds, hereafter called A and B breeds. For the
next 7 generations, using a random mating design,
animals from breed A and B were selected based on
favorable and unfavorable alleles of the major gene,
respectively. Hence, two breeds were generated and
fixed for favorable and unfavorable alleles. After fixa-
tion of favorable and unfavorable alleles, the breeding
values of animals for LS were predicted using a thresh-
old model (Model 2 below). At the third step, 30 males
and 300 females from the last generation of Breed B
were selected based on high breeding values and
treated as founders to generate the native breed. Fur-
thermore, to generate the synthetic breed, 30 males and
300 females from the last generation of breed A and B,
respectively, were selected based on high breeding
values and treated as founders. The founder animals
expanded over five generations and were selected and
mated based on high breeding values and minimizing
inbreeding. In order to have a constant population size
across generations, each dam produced 5 offsprings
with an equal probability of each sex. Replacement
rates for males and females were 40% and 20%, re-
spectively. In generations G1 to G5, the native and the
synthetic breeds were kept at a constant size of 1500
breeding candidates per each breed and 30 males and
300 females were selected in each generation as parents
of the next generation.
Genome structure
The LS trait in sheep was simulated with heritabil-
ity of 0.1. The genome consisted of 2 chromosomes,
each 100 cM in length. For each chromosome, 10000
markers and 100 QTLs were simulated (20000 SNPs
and 200 QTLs in total). Due to the limitation in com-
putational requirements, only two chromosomes were
considered. One of the QTLs at position 25.7 cM on
chromosome 1 was considered as the major gene which
explained 40% of the additive genetic variance for LS
trait. This fraction of genetic variance for the major
gene was similar to that reported in the Lacaune sheep
by Bodin et al. (2014). Other QTLs were randomly
distributed over the chromosomes. The rest of the ad-
ditive genetic variance (60%) was allocated to the re-
maining 199 QTLs. The QTLs effects were sampled
from a gamma distribution with shape of parameter ϒ
= 0.4. Biallelic SNP markers were randomly distrib-
When a major gene with desirable characteristics is
detected in a particular breed, introduction and intro-
gression of this gene into other breeds might be desir-
able. The cross between two or more breeds and sub-
sequent mating among crossbred animals is considered
a synthetic breed. Almost 418 synthetic breeds have
been developed from the combination of two and more
breeds (Rasali et al., 2006). One of the main purposes
of developing synthetic breeds has been to increase
prolificacy and reproduction efficiency (Hulet et al.,
1984; Fahmy, 1990). Hence, crossbreeding between
high and low prolific breeds for increasing favorable
allele frequencies in a synthetic breed appear to be use-
ful for improving LS. Previous studies have shown that
breeding programs for introduction and introgression
of favorable alleles such as the Booroola gene (FecB)
into other sheep breeds is highly successful (Gootwine
et al., 2008; Mishra et al., 2009).
Conventional or best linear unbiased prediction
(CBLUP) selection is based on pedigree and pheno-
typic information (Henderson, 1975). Genomic selec-
tion, described by Meuwissen et al. (2001) is a method
for improving quantitative traits in plant and animal
breeding. This technique relies on the segmentation of
the genome in thousands of intervals bracketed by
contiguous markers and effects of all markers on the
whole genome are estimated (Gaspa et al., 2015).
Several major genes (casual mutations) have been
identified for LS in sheep (Galloway et al., 2000; Souza
et al., 2001; Hanrahan et al., 2004; Martinez-Royo et
al., 2008; Drouilhet et al., 2009). The genomic selec-
tion method (Duchemin et al., 2012) has opened its
way into sheep breeding programs. Combining ge-
nomic or classical selection with gene-assisted selection
could be used for creating a synthetic sheep breed.
Therefore, the objectives of this study were: (1) to
compare classical, genomic, gene-assisted classical
(GasClassical) and gene-assisted genomic (GasGe-
nomic) selection methods in introducing a favorable
gene for the creation of a synthetic breed to improve
LS; and (2) to track genetic gain between a native and
a synthetic sheep breed via classical and genomic selec-
tion over five generations in a sheep population by
simulation.
Material and methods
Simulation
Simulations were performed (Fig. 1) based on the
forward-in-time process using the QMSim software
(Sargolzaei & Schenkel, 2009). The first step involved
the historical population being generated and ran-
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
3
Introduction of a major gene to improve litter size in sheep population
trait such as LS. All parameters used for the simulation
are summarized in Table 1.
Selection
Two selection strategies were used from G0 to G5.
The first was creation of the synthetic breed to intro-
duce a favorable allele by classical, genomic, GasClas-
sical and GasGenomic selection and the second was
the selection of superior animals in the native breed
based on classical and genomic selection.
Prediction of breeding values
―Classical selection. The Bayesian method was
used for prediction of breeding values (PBVs) based
on pedigree and phenotypic information. For Bayesian
method, the MCMCglmm R package developed by
uted along the two chromosomes. The mutation rate
for both markers and QTL was assumed 2.5×10-4 per
locus per generation (Esfandyari et al., 2015). The true
breeding value (TBV) for an individual was calculated
by multiplying the genotype codes by the QTL ef-
fects. Finally, the phenotypic value of each individual
i (yi), was created by adding a normally distributed
residual
e
i
~
N
0,
σ
e
2
( )
to the sum over QTLs of ge-
netic values as shown below:
y
i
=
k
=1
m
x
ik
α
k
+
e
i
(1)
Above, xik (i=1, …, n; k=1,…, m) is an element of
the incidence marker matrix for additive genetic effects
k) and ei is a random residual [ e ~
N
0,
Iσ
e
2
( )
], where
σ
e
2is the residual variance. To convert a continuous
trait into a threshold trait in each generation, 20% of
the high-level phenotypes were considered to be 2, and
the rest were considered as 1 to simulate a categorical
19
1
Figure 1. Structure of the simulation schemes. 2

  (500 males and 500 females)
Generation = 1000
After fixation of favorable
and unfavorable alleles in
Breed A and B, respectively,
the breeding values of the
animals were predicted
Breed B (30 males and 300 females)
7 generations
Selection based on unfavorable allele
Breed A (30 males and 300 females)
7 generations
Selection based on favorable allele
Native Breed
Selection based on high breeding
values for 5 generations
Synthetic Breed
Selection based on high breeding
values for 5 generations
Figure 1. Structure of the simulation schemes.
Meysam Latifi, Amir Rashidi, Rostam Abdollahi-Arpanahi and Mohammad Razmkabir
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
4
Hadfield (2010) was applied to analyze classical selec-
tion using BLUP methodology. The following animal
model was used:
l = + Za + e (2)
where l is the vector of underlying latent variable for trait
(one threshold and two categories of response), μ is the
overall mean, 1 is a vector of ones, Z is the incidence
matrix relating phenotypic records to the animals, a is a
vector of random additive genetic effects with distribution
a~ N(0, A
σ
a
2
)
and e is a vector of random residuals with
e~ N(0, I
σ
e
2)
, where
σ
a
2 and
σ
e
2are additive genetic and
residual variances, respectively. A and I are the additive
genetic relationship and the identity matrices, respec-
tively. A total of 300,000 rounds of Gibbs sampler were
considered and the first 30,000 rounds were discarded as
a burn-in period. The thinning interval was set to 100
cycles. In each generation, pedigree and phenotypic re-
cords of new generation were added to the recent pedi-
gree and data file. Therefore, PBVs for every round of
selection were re-calculated and the predicted breeding
values were used for selection.
Genomic selection. The Bayes B method was used
to estimate marker effects with Monte Carlo Markov
Chain (MCMC). In Bayes B, it is assumed that each
SNP has either an effect of zero or non-zero with prob-
abilities of π and 1-π, respectively, and for those with
non-zero effect, it is assumed that each SNP has a dif-
ferent variance. The BGLR (Bayesian Generalized
Linear Regression) package of R software (Perez &
Campos, 2014) was used for prediction of genomic
breeding values (GPBVs). The Gibbs sampler was run
for 30,000 rounds with a 3,000 burn-in period. The
thinning interval was set to 10 cycles. The following
model was used to estimate the genomic breeding values:
l =
µ
+
j=1
n
xijmj+e
(3)
where l is the vector of underlying latent variable for
trait (one threshold and two categories of response), μ
is the overall mean, 1 is a vector of ones, xij is an inci-
dence matrix for marker j and individual i, mj is a
random effect for marker j and e is the random resid-
ual error with distribution of e~ N(0, I
σ
e
2). In each
generation, the effects of markers were estimated using
female phenotypes. GPBVs for male and females were
calculated as the sum of all marker effects according
to their marker genotypes:
GPBVi=
j=1
n
xijmj
!
(4)
where xij is the genotype of animal i at locus j, and mj
!
is the estimated substitution effect of marker j.
Table 1. Parameters of the simulation process.
Historical and current populations
No. of generations (effective population size) 1000 (1000)
Breed A B Native Synthetic
No. of founder sires (dams) 30 (300) 30 (300) 30 (300) 30 (300)
Criteria for selection/culling Favorable allele/
age
Unfavorable
allele/age
High breeding
value/age
High breeding
value/age
No. of generations for each breed 7 7 5 5
Mating system Random Random Minimizes
inbreeding
Minimizes
inbreeding
No. of offspring per dam 5
Replacement ratio for males (females) 0.4 (0.2)
Sex probability for offspring 0.5
Genome
No. of chromosomes 2
Total length of chromosome (M) 2
Position of markers and QTLs Random
No. of QTL 200
No. of markers 20000
Marker and QTL mutation rate 2.5 × 10–4
Heritability 0.1
Variance of QTL as the major gene (%) 40
Position of major gene (cM) 25.7
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
5
Introduction of a major gene to improve litter size in sheep population
where D = freq(AB) – (freq(A) × freq(B)); freq(AB)
is frequency of observed haplotype; and freq(A) and
freq(B) are frequencies of alleles A and B, respec-
tively.
Evaluation of scenarios
The genetic gain, frequency of favorable allele and
mean of inbreeding coefficient were monitored across
generations for all selection schemes. The genetic gain
per generation was computed as the average TBV over
time. In addition, the accuracy of prediction for both
classical and genomic selection methods was obtained
as the Pearson correlation between the TBVs and (G)
PBVs in each generation. A total of 10 replicates were
produced for each scenario and selection method. The
10 replicates were simulated separately, but the initial
generation (G0) was identical for all scenarios within
each replicate.
Results
Native breed
The results obtained using simulations performed
for the native breed are summarized in Table 2. The
results obtained under classical and genomic selection
methods showed that the trend of true breeding values
increased across generations. The mean of true breed-
ing values for both methods were -0.22 at G0 and in-
creased to 0.264 and 0.621 at G5 for classical and
genomic selection, respectively. Furthermore, the mean
of true breeding values obtained under genomic selec-
Combining gene-assisted selection with
classical and genomic selection
In the present study, selection and culling of animals
in each generation was based on (G)PBVs and age, re-
spectively. For combining gene-assisted selection with
classical and genomic selection (GasClassical and
GasGenomic), we developed an alternative approach that
allows selection of animals with favorable allele and the
highest (G)PBVs as the parents of the next generation.
According to this approach: 1) the PBV or GPBV for
each animal was predicted using model 2 and 3, respec-
tively; 2) homozygosity or heterozygosity of each animal
was determined based on the favorable allele, and 3)
based on the equation below, the weighted predicted
breeding value of each animal was obtained:
G
PBVW=G
PBVp+NFA (5)
where G
( )
PBVW
=G
( )
PBVp+NFA
is the weighted PBV or GPBV,
G
( )
PBVW=
G
( )
PBVp
+NFA
is the predicted PBV or GPBV for each ani-
mal based on model 2 and 3, respectively, and NFA is
number of favorable alleles in each animal. If an animal
had 2, 1 or zero copies of the favorable allele, NFA was
2, 1 and 0, respectively (Fig. S1 [suppl.]).
Linkage disequilibrium (LD)
The LD in the native and the synthetic breeds from
G0 to G5 was assessed by r2 among all pairs of markers
using QMSim software (Sargolzaei & Schenkel, 2009):
r2=D2
fre(A) ×freq(a) ×freq(B) ×freq(b) (6)
Table 2. Mean of true breeding values, accuracy of prediction (r) and mean of inbreeding coefcient (F) in each generation
based on classical and genomic selection in the native breed.
Selection method Gen MeanTBV r(PBV,TBV) F
Classical 0 -0.220 ± 0.155 0.253 ± 0.056 0.000 ± 0.000
1 -0.146 ± 0.154 0.309 ± 0.064 0.000 ± 0.000
2 -0.064 ± 0.168 0.374 ± 0.066 0.000 ± 0.000
3 0.042 ± 0.187 0.370 ± 0.102 0.000 ± 0.000
4 0.146 ± 0.206 0.375 ± 0.055 0.004 ± 0.004
5 0.264 ± 0.226 0.358 ± 0.075 0.016 ± 0.007
Genomic 0 -0.220 ± 0.155 0.669 ± 0.077 0.000 ± 0.000
1 -0.145 ± 0.155 0.696 ± 0.067 0.000 ± 0.000
2 0.017 ± 0.170 0.722 ± 0.061 0.000 ± 0.000
3 0.209 ± 0.197 0.763 ± 0.047 0.000 ± 0.000
4 0.416 ± 0.215 0.779 ± 0.047 0.000 ± 0.000
5 0.621 ± 0.228 0.738 ± 0.046 0.002 ± 0.001
Meysam Latifi, Amir Rashidi, Rostam Abdollahi-Arpanahi and Mohammad Razmkabir
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
6
tion was 74% higher than that of the classical selection.
Thus, the accuracy of prediction using genomic selec-
tion was higher compared to the classical selection.
The highest accuracies for classical and genomic selec-
tion was obtained in G4 and afterwards, decreased in
G5 (Table 2). At G5, the mean of inbreeding coeffi-
cients for the native breed under classical and genom-
ic selection were 0.016 and 0.002, respectively.
Synthetic breed
The results obtained using simulations performed
for the synthetic breed are summarized in Table 3. As
expected, regardless of selection methods, mean of true
breeding values increased across generations. Genetic
gain obtained under GasGenomic (1.347) was higher
than that of obtained using the GasClassical selection
(1.007), and the value of 1.319 obtained for Genomic
selection was higher than that of obtained for the clas-
sical selection method (0.884). Combining gene assist-
ed-selection with classical and genomic selection led
to higher genetic gain. Consequently, genetic gain
obtained by GasGenomic and GasClassical selection
methods was greater than those of genomic and clas-
sical selection (2% and 16% higher genetic gain from
G0 to G5, respectively). As shown in Table 3, genom-
ic and GasGenomic selection resulted in better accura-
cies of PBV prediction in comparison with classical
and GasClassical selection methods. The highest ac-
curacy for classical, genomic, GasClassical and
GasGenomic selection in the synthetic breed were
obtained at G4 which decreased slightly at G5 (Table
3). For the synthetic breed, a similar pattern to that of
the native breed was observed for the mean of inbreed-
ing coefficient. At G5, mean of inbreeding coefficients
were 0.011, 0.002, 0.008 and 0.002 for classical,
genomic, GasClassical and GasGenomic genomic se-
lection, respectively. The mean of inbreeding coeffi-
cient in classical and GasClassical selection varied
more compared to genomic and GasGenomic selection.
The frequency of the favorable allele in each gen-
eration under classical, genomic, GasClassical and
GasGenomic selection in the synthetic breed for 10
replicates is shown in Fig. 2. The frequency of the fa-
vorable allele started with an initial value of 0.091 at
G0 and reached to 0.623, 0.730, 0.850 and 0.848 at G5
for classical, genomic, GasClassical and GasGenomic
selection methods, respectively (Table 3). The fre-
quency of favorable allele under classical and genomic
Table 3. Mean of true breeding values, accuracy of prediction (r) and mean of inbreeding coefcient (F) in each generation
based on classical, genomic, GasClassical and GasGenomic selection methods in the synthetic breed.
Selection
method Gen MeanTBV r(PBV,TBV) Frequency1F
Classical 0 -0.142 ± 0.151 0.263 ± 0.034 0.091 ± 0.000 0.000 ± 0.000
1 0.282 ± 0.154 0.298 ± 0.066 0.500 ± 0.000 0.000 ± 0.000
2 0.332 ± 0.149 0.361± 0.095 0.451 ± 0.005 0.000 ± 0.000
3 0.384 ± 0.163 0.506 ± 0.057 0.429 ± 0.024 0.000 ± 0.000
4 0.530 ± 0.168 0.514 ± 0.059 0.491 ± 0.051 0.002 ± 0.003
5 0.742 ± 0.176 0.462 ± 0.086 0.623 ± 0.072 0.011 ± 0.004
Genomic 0 -0.142 ± 0.151 0.581 ± 0.089 0.091 ± 0.000 0.000 ± 0.000
1 0.281 ± 0.153 0.646 ± 0.056 0.500 ± 0.000 0.000 ± 0.000
2 0.399 ± 0.144 0.720 ± 0.052 0.449 ± 0.003 0.000 ± 0.000
3 0.579 ± 0.177 0.799 ± 0.043 0.442 ± 0.003 0.000 ± 0.000
4 0.852 ± 0.198 0.834 ± 0.045 0.537 ± 0.004 0.001 ± 0.001
5 1.177 ± 0.207 0.820 ± 0.048 0.730 ± 0.003 0.002 ± 0.003
GasClassical 0 -0.142 ± 0.151 0.251 ± 0.044 0.091 ± 0.000 0.000 ± 0.000
1 0.279 ± 0.152 0.303 ± 0.084 0.500 ± 0.000 0.000 ± 0.000
2 0.304 ± 0.168 0.386 ± 0.053 0.449 ± 0.005 0.000 ± 0.000
3 0.460 ± 0.184 0.510 ± 0.060 0.550 ± 0.033 0.000 ± 0.000
4 0.666 ± 0.220 0.516 ± 0.049 0.700 ± 0.053 0.002 ± 0.003
5 0.883 ± 0.251 0.482 ± 0.065 0.850 ± 0.043 0.008 ± 0.005
GasGenomic 0 -0.142 ± 0.151 0.635 ± 0.089 0.091 ± 0.000 0.000 ± 0.000
1 0.281 ± 0.151 0.672 ± 0.068 0.500 ± 0.000 0.000 ± 0.000
2 0.407 ± 0.145 0.741 ± 0.052 0.449 ± 0.004 0.000 ± 0.000
3 0.637 ± 0.150 0.784 ± 0.050 0.538 ± 0.027 0.000 ± 0.000
4 0.915 ± 0.159 0.805 ± 0.041 0.698 ± 0.003 0.001 ± 0.000
5 1.205 ± 0.184 0.793 ± 0.052 0.848 ± 0.005 0.002 ± 0.003
1Frequencies of favorable allele
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
7
Introduction of a major gene to improve litter size in sheep population
selection. Moreover, genetic gain in the synthetic breed
under GasGenomic (60%) and Genomic (57%) selec-
tion methods was higher than that obtained for the
native breed using the genomic selection method.
Linkage disequilibrium in the native and the
synthetic breeds
The average LD decay obtained using genomic se-
lection method for all possible pairs of markers in the
native and the synthetic breeds are shown in Fig. 4. As
illustrated, the maximum average of r2 for the native
and the synthetic breeds (genomic and GasGenomic
selection) at distance 0.0 to 50 kb at G0 were 0.19 and
0.18, respectively. However, the minimum average of
r2 at a distance of 400 to 500 kb for the native and the
selection (Figs. 2a and 2c) showed high fluctuations
whereas when applying GasClassical and GasGenomic
selections, no fluctuation was observed over 10 repli-
cates (Figs. 2b and 2d).
Comparison between the native
and the synthetic breeds
Mean of true breeding values for the native and the
synthetic breeds obtained using different methods are
shown in Fig. 3. As illustrated, genetic gains for LS in
the synthetic breed under classical or genomic selec-
tions were higher than selections in the native breed.
Genetic gain in the synthetic breed with GasClassical
(117%) and classical (82%) selection methods was
greater than that in the native breed under classical
Figure 2. Pattern of allele frequencies of the favorable allele for the synthetic breed under a) Classical b) Gene-assisted classical
(GasClassical), c) Genomic and d) Gene-assisted genomic (GasGenomic) selection across 10 replicates (the black line is the mean
of 10 replicates).
20
1
2
3
Figure 2. Pattern of allele frequencies of the favorable allele for the synthetic breed under a) 4
Classical b) Gene-assisted classical (GasClassical), c) Genomic and d) Gene-assisted 5
genomic (GasGenomic) selection over 10 replicates (the black line is the mean of 10 6
replicates). 7
8
0
0,2
0,4
0,6
0,8
1
012345
Frequency
Generation
Classical
a
0
0,2
0,4
0,6
0,8
1
012345
Frequency
Generation
GasClassical
b
0
0,2
0,4
0,6
0,8
1
012345
Frequency
Generation
Genomic
c
0
0,2
0,4
0,6
0,8
1
012345
Frequency
Generation
GasGenomic
d
Meysam Latifi, Amir Rashidi, Rostam Abdollahi-Arpanahi and Mohammad Razmkabir
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
8
synthetic breeds (genomic and GasGenomic selection)
at G0 were 0.11 and 0.10, respectively. For the native
and the synthetic breeds (genomic and GasGenomic
selection) at G5, the average maximum values of r2 for
distance of 0.0 to 50 kb were 0.30 and 0.26, respec-
tively. In addition, the average minimum values of r2
for the native and the synthetic breeds (genomic and
GasGenomic selection) from 400 to 500 kb at G5 were
0.23 and 0.18, respectively. For the native and the
synthetic breeds, by increasing marker pair distance,
the values of r2 decreased at G0 and G5 (Fig. 4). The
native breed showed higher levels of LD in comparison
with the synthetic breed at G0 and G5.
Discussion
The results of this study showed that employing
genomic and GasGenomic selection methods rather
than Classical and GasClassical selection lead to in-
crease in genetic gain in the native and the synthetic
breeds. Meuwissen et al. (2001) showed that using
genomic selection method resulted in an increased ac-
curacy of selection for traits with low heritability and
female sex-limited such as LS. A reason for the high
accuracy of prediction in genomic selection could be
due to use of the marker information to capture the
Mendelian sampling terms (Daetwyler et al., 2007).
The results reported by other researchers indicated that
the accuracy of prediction depends on the extent of
linkage disequilibrium, the density of markers, statisti-
cal methods, heritability of trait and selection design
(Wang et al., 2013; Gowane et al., 2018; Karimi et al.,
2019). Using simulated data, higher accuracies were
reported for genomic selection compared to classical
selection in dairy cattle (Gaspa et al., 2015), swine
(Putz et al., 2018) and American mink (Karimi et al.,
2019). In the current study, the accuracies in all selec-
tion methods increased from G0 to G4. Afterward, a
decrease was observed in G5, because individuals in
this generation did not have any recorded offspring.
When the combination of gene-assisted selection with
classical and genomic selection was carried out for
improvement of a trait by introducing a major gene,
genetic gain was higher than when classical and ge-
nomic selection were used alone. These findings indi-
cate that if the major gene is controlling a large propor-
tion of genetic variance in a population, the combination
of classical and genomic selection with gene-assisted
selection leads to an increase in genetic gain. Pedersen
et al. (2009) indicated that using genotype information
increased within-family selection and lead to an increase
in genetic gain as the accuracy of breeding values in-
creased. Generally, this result was obtained using clas-
sical and genomic selection methods in the synthetic
breed which resulted in higher genetic gain compared
to the native breed. High genetic gain for the synthetic
breed led to an increase of heterozygosity at the QTL
level and accuracy of selection method (Ødegard et al.,
2009b). Our results indicated that two breeds (A and B)
Figure 3. Mean of breeding values for the native and the synthetic breeds with Genomic and Clas-
sic selection methods across 10 replicates.
21
1
2
3
4
Figure 3. Mean of breeding values for the native and the synthetic breeds with Genomic and 5
Classic selection methods over 10 replicates. 6
7
-0,4
-0,2
0
0,2
0,4
0,6
0,8
1
1,2
1,4
012345
Mean of true breeding value
Generation
Classical-Native Breed
Classical-Synthetic Breed
GasClassical-Native Breed
Genomic-Native Breed
Genomic-Synthetic Breed
GasGenomic-Synthteic Breed
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
9
Introduction of a major gene to improve litter size in sheep population
GasClassical selection method (31%). These results are
in agreement with Ødegard et al. (2009a), who stated
that combining genomic selection with gene-assisted
selection for the target QTL acted as an additional ac-
tion against decline of the target QTL and gave surpass
genetic gain.
The study presented here about the extent of the LD
can be used to interpret the difference observed among
the breeds. Observed LD is a function of the recombi-
nation rate between loci within a breed and the selec-
tion performed for specific quantitative or qualitative
traits of interest (Prieur et al., 2017). Each of the breeds
showed the decrease in r2 with increased distance be-
tween markers. In the present study, the levels of LD
for the native and the synthetic breeds at a distance
from 0.0 to 50 kb at G0 and G5 was in agreement with
the results reported by Kijas et al. (2014) at a distance
of 10 kb. For the native and synthetic breeds under
selection, the level of LD increased over the genera-
tions. The high level of LD showed a high level of
selection intensity over generations for these breeds.
The extent of LD could be increased by several factors,
including inbreeding, population structure, and bottle-
necks (Pritchard & Przeworski, 2001). In this study,
the extent of LD was found to be remarkably different
between the native and the synthetic breeds (Fig. 4).
This could be due to the creation of the synthetic breed
from two cross-breeds (Breed A and B). Toosi et al.
(2010) mentioned that individual animals are less re-
lated to each other in a crossbreed population; there-
fore, LD haplotypes in crossbred populations are nar-
rower than in a purebred. Results of the current study
showed that the LD decay was different between
breeds, and similar results in sheep were reported by
Al-Mamun et al. (2015).
were separated based on the frequencies of the major
gene after 7 generations. If the two breeds were consid-
erably divergent, the relative advantage of crossing
could be higher.
In both the native and the synthetic breeds, the rate
of increase of inbreeding coefficient was lower in ge-
nomic selection compared to classical selection. Dae-
twyler et al. (2007) stated that the low inbreeding rate
in genomic selection compared to the classical selection
method is due to an increase in prediction accuracy of
the Mendelian sampling. In our simulation, the mating
system was based on optimized mating design. Thus,
pairs of mates were chosen based on minimizing ge-
netic relationship to create the next generation (Sargol-
zaei & Schenkel, 2009). Hence, the rate of inbreeding
coefficient in both genomic and classical selection was
lower than the results reported in previous research
(Ødegard et al., 2009b; Gaspa et al., 2015).
The frequency of favorable allele after five genera-
tions in classical, genomic, GasClassical and GasGe-
nomic selection methods in the synthetic breed were
0.62, 0.73, 0.85 and 0.85, respectively (Table 3). Re-
sults obtained in our study showed that the frequency
of favorable allele in genomic selection increased by
more than 20% compared to the classical selection.
These findings indicate that genomic selection is more
effective in increasing the frequency of favorable allele
than the classical selection without any knowledge of
the major gene’s position. Under GasClassical and
GasGenomic selection methods, the frequency of the
favorable allele was never declined in comparison with
classical and genomic selections (Fig. 2). The fre-
quency of the favorable allele was the same in the
GasClassical and GasGenomic, but the rate of genetic
gain in the GasGenomic was higher than that of the
Figure 4. The average LD decay for the native and the synthetic breeds (genomic and GasGenomic selection) at generation zero
(G0) and ve (G5) across 10 replicates.
22
1
2
Figure 4. The average LD decay for the native and the synthetic breeds (genomic and 3
GasGenomic selection) at generation zero (G0) and five (G5) across 10 replicates. 4
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
Average LD (r2)
Distance Range (kb)
G0
Genomic-Native Breed Genomic-Synthetic Breed GasGenomic-Synthetic Breed
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
Average LD (r2)
Distance Range (kb)
G5
Genomic-Native Breed Genomic-Synthetic Breed GasGenomic-Synthetic Breed
Meysam Latifi, Amir Rashidi, Rostam Abdollahi-Arpanahi and Mohammad Razmkabir
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
10
Fahmy MH, 1990. Development of DLS breed of sheep:
genetic and phenotypic parameters of date of lambing and
litter size. Can J Anim Sci 70: 771-778. https://doi.
org/10.4141/cjas90-096
Galloway SM, McNatty KP, Cambridge LM, Laitinen MP,
Juengel JL, Jokiranta TS, McLaren RJ, Luiro K, Dodds
KG, Montgomery G, 2000. Mutations in an oocyte-derived
growth factor gene (BMP15) cause increased ovulation
rate and infertility in a dosage-sensitive manner. Nat Genet
25: 279-283. https://doi.org/10.1038/77033
Gaspa G, Veerkamp RF, Calu MPL, Windig JJ, 2015. Assess-
ment of genomic selection for introgression of polledness
into Holstein Friesian cattle by simulation. Livest Sci 179:
86-95. https://doi.org/10.1016/j.livsci.2015.05.020
Gootwine E, Reicher S, Rozov A, 2008. Prolificacy and lamb
survival at birth in Awassi and Assaf sheep carrying the
FecB (Booroola) mutation. Anim Reprod Sci 108: 402-
411. https://doi.org/10.1016/j.anireprosci.2007.09.009
Gowane GR, Lee SH, Clark S, Moghaddar N, Al-Mamun
HA, van der Werf JHJ. 2018. Effect of selection on bias
and accuracy in genomic prediction of breeding values.
bioRxiv 298042. https://doi.org/10.1101/298042
Hadfield JD, 2010. MCMC methods for multi-response
generalized linear mixed models: the MCMCglmm R
package. J Stat Softw 33: 1-22.
Hanford KJ, Van Vleck LD, Snowder GD, 2005. Estimates
of genetic parameters and genetic change for reproduction,
weight, and wool characteristics of Rambouillet sheep.
Small Rumin Res 57: 175-186. https://doi.org/10.1016/j.
smallrumres.2004.07.003
Hanrahan JP, Gregan SM, Mulsant P, Mullen M, Davis GH,
Powell R, Galloway SM, 2004. Mutations in the genes
for oocyte-derived growth factors GDF9 and BMP15 are
associated with both increased ovulation rate and steril-
ity in Cambridge and Belclare sheep (Ovis aries). Biol
Reprod 70: 900-909. https://doi.org/10.1095/biol-
reprod.103.023093
Henderson CR, 1975. Best linear unbiased estimation and
prediction under a selection model. Biometrics 31: 423-
447. https://doi.org/10.2307/2529430
Hulet CV, Ercanbrack SK, Knight AD, 1984. Development
of the Polypay breed of sheep. J Anim Sci 58: 15-24.
https://doi.org/10.2527/jas1984.58115x
Karimi K, Sargolzaei M, Plastow GS, Wang Z, Miar Y, 2019.
Opportunities for genomic selection in American mink:
A simulation study. PLOS ONE 14: e0213873. https://doi.
org/10.1371/journal.pone.0213873
Kijas JW, Porto-Neto L, Dominik S, Reverter A, Bunch R,
McCulloch R, Hayes BJ, Brauning R, McEwan J, 2014.
Linkage disequilibrium over short physical distances
measured in sheep using a high-density SNP chip. Anim
Genet 45: 754-757. https://doi.org/10.1111/age.12197
Martinez-Royo A, Jurado JJ, Smulders JP, Marti JI, Alabart
JL, RocheA, Fantova E, Bodin L, Mulsant P, Serrano M,
Folch J, Calvo JH, 2008. A deletion in the bone morpho-
genetic protein 15 genecauses sterility and increased
prolificacy in Rasa Aragonesa sheep. Anim Genet 39:
294-297. https://doi.org/10.1111/j.1365-2052.2008.
01707.x
The results of this study confirmed that reproduction
and sex-limited traits such as LS in sheep can be im-
proved by using genomic selection methods. The results
also demonstrated that inclusion of information about
known causal mutations into genomic and classical
selection methods can lead to increase in frequency of
favorable allele and subsequently resulted in higher
genetic gain in the synthetic breed.
Acknowledgments
The authors are sincerely grateful to the Dr. M Hos-
sein Yazdi (Head of Technical Analysis Division,
Akvaforsk Genetics Center AS, N-6600 Sunndalsøra,
Norway) and Dr. Mehdi Bohlouli and Carsten Scheper
(Institute of Animal Breeding and Genetics, Justus-
Liebig-University Gießen, 35390 Gießen, Germany)
for the valuable discussions.
References
Al-Mamun HA, Clark SA, Kwan P, Gondro C, 2015. Ge-
nome-wide linkage disequilibrium and genetic diversity
in five populations of Australian domestic sheep. Genet
Sel Evol 47: 1-14. https://doi.org/10.1186/s12711-015-
0169-6
Bodin L, Martin PM, Raoul J, 2014. Effects of the FecL
major gene on mean and variance of litter size in the
Lacaune meat sheep population. Proc 10th World Conf
on genetics applied to livestock production (WCGALP),
Vancouver (Canada), Aug 17-22.
Daetwyler HD, Villanueva B, Bijma P, Woolliams JA, 2007.
Inbreeding in genome-wide selection. J Anim Breed Genet
124: 369-376. https://doi.org/10.1111/j.1439-0388.2007.
00693.x
Davis GH, 2005. Major genes affecting ovulation rate in
sheep. Genet Sel Evol 37: 11-23. https://doi.
org/10.1186/1297-9686-37-S1-S11
Drouilhet L, Lecerf F, Bodin L, Fabre S, Mulsant P, 2009.
Fine mapping of the FecL locus influencing prolificacy
in Lacaune sheep. Anim Genet 40: 804-812. https://doi.
org/10.1111/j.1365-2052.2009.01919.x
Duchemin S, Colombani C, Legarra A, Baloche G, Larroque
H, Astruc JM, Barillet F, Robert-Granié C, Manfredi E,
2012. Genomic selection in the French Lacaune dairy
sheep breed. J Dairy Sci 95: 2723-2733. https://doi.
org/10.3168/jds.2011-4980
Elsen JM, Bodin L, Francois D, Poivey JP, Teyssier J, 1994.
Genetic improvement of litter size in sheep. Proc 5th
World Conf on genetics applied to livestock production
(WCGALP). Guelph, Ontario, (Canada), pp: 237-244.
Esfandyari H, Sorensen AC, Bijma P, 2015. Maximizing
crossbred performance through purebred genomic selec-
tion. Genet Sel Evol 47: 16. https://doi.org/10.1186/
s12711-015-0099-3
Spanish Journal of Agricultural Research March 2020 • Volume 18 • Issue 1 • e0403
11
Introduction of a major gene to improve litter size in sheep population
disequilibrium and efective population size in New Zea-
land sheep using three diferent methods to create genetic
maps. BMC Genet 18: 68. https://doi.org/10.1186/s12863-
017-0534-2
Pritchard JK, Przeworski M, 2001. Linkage disequilibrium
in humans: models and data. Am J Hum Genet 69:1-14.
https://doi.org/10.1086/321275
Putz AM, Tiezzi F, Maltecca C, Gray KA, Knauer MT, 2018.
A comparison of accuracy validation methods for genom-
ic and pedigree-based predictions of swine litter size traits
using Large White and simulated data. J Anim Breed
Genet 135: 5-13. https://doi.org/10.1111/jbg.12302
Rasali DP, Hrestha JNB, Crow GH, 2006. Development of
composite sheep breeds in the world: A review. Can J
Anim Sci 86: 1-24.
Sargolzaei M, Schenkel FS, 2009. QMSim: a largescale
genome simulator for livestock. Bioinformatics 25: 680-
681. https://doi.org/10.1093/bioinformatics/btp045
Souza C, MacDougall C, Campbell B, McNeilly A, Baird D,
2001. The Booroola (FecB) phenotype is associated with
a mutation in the bone morphogenetic receptor type 1 B
(BMPR1B) gene. J Endocrinol 169: R1-R6. https://doi.
org/10.1677/joe.0.169r001
Toosi A, Fernando RL, Dekkers JCM, 2010. Genomic selec-
tion in admixed and crossbred populations. J Anim Sci
88: 32-46. https://doi.org/10.2527/jas.2009-1975
Wang CL, Ding XD, Wang JY, Liu JF, Fu WX, Zhang Z, Yin
ZJ, Zhang, 2013. Bayesian methods for estimating GEBVs
of threshold traits. Heredity 110: 213-219. https://doi.
org/10.1038/hdy.2012.65
Meuwissen THE, Hayes B, Goddard ME, 2001. Prediction
of total genetic value using genome-wide dense marker
maps. Genetics 157: 1819-1829.
Mishra AK, Arora AL, Kumar S, Prince LLL, 2009. Studies
on effect of Booroola (FecB) genotype on lifetime ewe’s
productivity efficiency, litter size and number of weaned
lambs in Garole × Malpura sheep. Anim Reprod Sci 113:
293-298. https://doi.org/10.1016/j.anireprosci.2008.06.002
Mokhtari MS, Rashidi A, Mohamadi Y, 2010. Estimation of
genetic parameters for post-weaning traits of Kermani
sheep. J Anim Sci 80: 22-27. https://doi.org/10.1016/j.
smallrumres.2008.08.002
Ødegård J, Sonesson Ak, Yazdi MH, Meuwissen THE, 2009a.
Introgression of a major QTL from an inferior into a su-
perior population using genomic selection. Genet Sel Evol
41: 1-10. https://doi.org/10.1186/1297-9686-41-38
Ødegard J, Yazdi MH, Sonesson AK, Meuwissen THE,
2009b. Incorporating desirable genetic characteristics from
an inferior into a superior population using genomic selec-
tion. Genetics 181: 737-745. https://doi.org/10.1534/genet-
ics.108.098160
Pedersen LD, Sørensen AC, Berg P, 2009. Marker-assisted
selection can reduce true as well as pedigree-estimated
inbreeding. J Dairy Sci 92: 2214-2223. https://doi.
org/10.3168/jds.2008-1616
Perez P, de los Campos G, 2014. Genome-wide regression and
prediction with the BGLR statistical package. Genetics 198:
483-495. https://doi.org/10.1534/genetics.114.164442
Prieur V, Clarke SM, Brito LF, McEwan JC, Lee MA, Braun-
ing R, Dodds KG, Auvray B, 2017. Estimation of linkage
... (Dekkers, 2007;Daetwyler et al., 2007;Daetwyler et al., 2008 (Brzustowicz et al., 1993;Göring &Terwilliger, 2000;Marquard et al., 2009 (Pompanon et al., 2005) . (Wang et al., 2013;Bazzi et al., 2017;Foroutanifar, 2017;Naderi, 2018;Latifi et al., 2020 ...
... Recently, the identification of GDF9 gene has been of interest for some researchers [9,14]. In particular, the effects or presence of the mutations of this gene on fertility have been focused on small ruminants [9,[15][16][17][18]. ...
Article
Full-text available
Accuracy of prediction. Litter size (LS) is one of the important economic traits in sheep. This trait is influenced by some genes with a large effect. The purpose of this simulation study was to evaluate the fixation selection designs for favorable major gene allele under scenarios of overlapping and discrete generations. In this regard, a trait with heritability of 0.1, and a genome with 26 chromosomes and a major gene was simulated in sheep population. Animals breeding value was predicted using Bayesian threshold model. Selection of animals was based on estimated breeding value (EBV), phenotype (PHEN) and random (RND). After ten generations, genetic gain in selection based on EBV, PHEN and RND under scenario of discrete generation were 8, 23 and 26 percent higher than those in scenario of overlapping generation, respectively. The accuracy of prediction in scenario of discrete generation was higher than scenario of overlapping generation. Means of inbreeding coefficient under scenario of overlapping generation and selection for EBV, PHEN and RND were 0.317, 0.029 and 0.027, respectively, and for scenario of discrete generation were 0.327, 0.058 and 0.056, respectively. In generation ten, the favorable allele of the major gene, in scenario of discrete generation selection based on EBV, PHEN and RND was 1, 46 and 38 percent higher than those in the scenario of overlapping generation, respectively. The results indicated that the scenario of discrete generation selection based on EBV leads to more fixation the favorable allele of the major gene and of genetic gain. Cite this article: Latifi, M., & Naderi, Y. (2024). Evaluation of frequency of the favorable allele of major gene under overlapping and discrete generation: A study based on simulation.
Preprint
Full-text available
12 Reference populations for genomic selection (GS) usually involve highly selected individuals, 13 which may result in biased prediction of estimated genomic breeding values (GEBV). In the present 14 study, bias and accuracy of GEBV were explored for various genetic models and prediction methods 15 when using selected individuals for a reference. Data were simulated for an animal breeding program to 16 compare Best Linear Unbiased Prediction of breeding values using pedigree based relationships 17 (PBLUP), genomic relationships for genotyped animals only (GBLUP) and a Single Step approach 18 (SSGBLUP), where information on genotyped individuals was used to infer a matrix H with relationships 19 among all available genotyped and non-genotyped individuals that were linked through pedigree. In 20 SSGBLUP, various weights (α=0.95, 0.80, 0.50) for the genomic relationship matrix (G) relative to the 21 numerator relationship matrix (A) were applied to construct H and in another version (SSGBLUP_F), 22 inbreeding was accounted for while computing A-1. With GBLUP, accuracy of GEBV prediction 23 increased linearly with an increase in the number of animals selected in reference. For the scenario with 24 no-selection and random mating (RR) prediction was unbiased. For GBLUP, lower accuracy and bias 25. CC-BY-NC-ND 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint. http://dx.doi.org/10.1101/298042 doi: bioRxiv preprint first posted online Apr. 9, 2018; observed in the scenarios with selection and random mating (SR) or selection and positive assortative 1 mating (SA), in which prediction bias increased when a smaller and highly selected proportion 2 genotyped. Bias disappeared when all individuals were genotyped. SSGBLUP_F showed higher accuracy 3 compared to GBLUP and bias of prediction was negligible even with selective genotyping. However, 4 PBLUP and SSGBLUP showed bias in SA owing to not fully accounting for allele frequency changes 5 because of selection of quantitative trait loci (QTL) with larger effects and also due to high inbreeding 6 rate. In genetic models with fewer QTL but each with larger effect, predictions were less accurate and 7 more biased for selection scenarios. Results suggest that prediction accuracy and bias is affected by the 8 genetic architecture of the trait. Selective genotyping lead to significant bias in GEBV prediction. 9 SSGBLUP with appropriate scaling of A and G matrices can provide accurate and less biased prediction 10 but scaling requires careful consideration in populations under selection and with high levels of 11 inbreeding. 12
Article
Full-text available
Genomic selection can be considered as an effective tool for developing breeding programs in American mink. However, the genetic gains for economically important traits can be influenced by the accuracy of genomic predictions. The objective of this study was to investigate the prediction accuracies of traditional best linear unbiased prediction (BLUP), multi-step genomic BLUP (GBLUP) and single-step GBLUP (ssGBLUP) methods in American mink using simulated data with different levels of heritability, marker density, training set (TS) sizes and selection designs based on either phenotypic performance or estimated breeding values (EBVs). Under EBV selection design, the accuracy of BLUP predictions was increased by 38% and 44% for h² = 0.10, 27% and 29% for h² = 0.20, and 5.8% and 6% for h² = 0.50 using GBLUP and ssGBLUP methods, respectively. Under phenotypic selection design, the accuracies of prediction by ssGBLUP method were 11.8% and 15.4% higher than those obtained by GBLUP for heritability of 0.10 and 0.20, respectively. However, the efficiency of ssGBLUP and GBLUP was not influenced by selection design at higher level of heritability (h² = 0.50). Furthermore, higher selection intensity increased the bias of predictions in both pedigree-based and genomic evaluations. Regardless of selection design, TS sizes for GBLUP and ssGBLUP methods should be at least 3000 to achieve more accuracy than using BLUP for heritability of 0.50 and marker density of 10k and 50k. Overall, more accurate predictions were obtained using ssGBLUP method particularly for lowly heritable traits and low density of markers. Our results indicated that TS sizes should be optimized in accordance with heritability level, marker density, selection design and prediction method for genomic selection in American mink. The results provided an initial framework for designing genomic selection in mink breeding programs.
Article
Full-text available
The objective of this study was to compare and determine the optimal validation method when comparing accuracy from single-step GBLUP (ssGBLUP) to traditional pedigree-based BLUP. Field data included six litter size traits. Simulated data included ten replicates designed to mimic the field data in order to determine the method that was closest to the true accuracy. Data were split into training and validation sets. The methods used were as follows: (i) theoretical accuracy derived from the prediction error variance (PEV) of the direct inverse (iLHS), (ii) approximated accuracies from the accf90(GS) program in the BLUPF90 family of programs (Approx), (iii) correlation between predictions and the single-step GEBVs from the full data set (GEBVFull), (iv) correlation between predictions and the corrected phenotypes of females from the full data set (Yc), (v) correlation from method iv divided by the square root of the heritability (Ych) and (vi) correlation between sire predictions and the average of their daughters' corrected phenotypes (Ycs). Accuracies from iLHS increased from 0.27 to 0.37 (37%) in the Large White. Approximation accuracies were very consistent and close in absolute value (0.41 to 0.43). Both iLHS and Approx were much less variable than the corrected phenotype methods (ranging from 0.04 to 0.27). On average, simulated data showed an increase in accuracy from 0.34 to 0.44 (29%) using ssGBLUP. Both iLHS and Ych approximated the increase well, 0.30 to 0.46 and 0.36 to 0.45, respectively. GEBVFull performed poorly in both data sets and is not recommended. Results suggest that for within-breed selection, theoretical accuracy using PEV was consistent and accurate. When direct inversion is infeasible to get the PEV, correlating predictions to the corrected phenotypes divided by the square root of heritability is adequate given a large enough validation data set.
Article
Full-text available
Background: Investments in genetic selection have played a major role in the New Zealand sheep industry competitiveness. Selection may erode genetic diversity, which is a crucial factor for the success of breeding programs. Better understanding of linkage disequilibrium (LD) and ancestral effective population size (N e) through quantifying this diversity and comparison between populations allows for more informed decisions with regards to selective breeding taking population genetic diversity into account. The estimation of N e can be determined via genetic markers and requires knowledge of genetic distances between these markers. Single nucleotide polymorphisms (SNP) data from a sample of 12,597 New Zealand crossbred and purebred sheep genotyped with the Illumina Ovine SNP50 BeadChip was used to perform a genome-wide scan of LD and N e. Three methods to estimate genetic distances were investigated: 1) M1: a ratio fixed across the whole genome of one Megabase per centiMorgan; 2) M2: the ratios of genetic distance (using M3, below) over physical distance fixed for each chromosome; and, 3) M3: a genetic map of inter-SNP distances estimated using CRIMAP software (v2.503). Results The estimates obtained with M2 and M3 showed much less variability between autosomes than those with M1, which tended to give lower N e results and higher LD decay. The results suggest that N e has decreased since the development of sheep breeds in Europe and this reduction in Ne has been accelerated in the last three decades. The N e estimated for five generations in the past ranged from 71 to 237 for Texel and Romney breeds, respectively. A low level of genetic kinship and inbreeding was estimated in those breeds suggesting avoidance of mating close relatives. Conclusions M3 was considered the most accurate method to create genetic maps for the estimation of LD and Ne. The findings of this study highlight the history of genetic selection in New Zealand crossbred and purebred sheep and these results will be very useful to understand genetic diversity of the population with respect to genetic selection. In addition, it will help geneticists to identify genomic regions which have been preferentially selected within a variety of breeds and populations.
Article
Full-text available
Abstract Research conducted since 1980 in relation to inheritance patterns and DNA testing of major genes for prolificacy has shown that major genes have the potential to significantly increase the reproductive performance of sheep flocks throughout the world. Mutations that increase ovulation rate have been discovered in the BMPR-1B, BMP15 and GDF9 genes, and others are known to exist from the expressed inheritance patterns although the mutations have not yet been located. In the case of BMP15, four different mutations have been discovered but each produces the same phenotype. The modes of inheritance of the different prolificacy genes include autosomal dominant genes with additive effects on ovulation rate (BMPR-1B; Lacaune), autosomal over-dominant genes with infertility in homozygous females (GDF9), X-linked over-dominant genes with infertility in homozygous females (BMP15), and X-linked maternally imprinted genes (FecX2). The size of the effect of one copy of a mutation on ovulation rate ranges from an extra 0.4 ovulations per oestrus for the FecX2 mutation to an extra 1.5 ovulations per oestrus for the BMPR-1B mutation. A commercial DNA testing service enables some of these mutations to be used in genetic improvement programmes based on marker assisted selection.
Article
Full-text available
In the 1850s, breeders established breed societies which maintained flock registries and actively participated in show rings, exhibitions and auctions promoting pure breeds that conformed to the breeders' vision of uniform and distinct morphological characteristics. At that time, the performance of the pure breeds over grade sheep was overwhelming. As a result influential government officials, specialists and scientists had to acknowledge their superiority. The supply of purebred seed-stock that was initially adequate became sparse in the years following the World War II because demand arising from the increasing human population exceeded the capacity for production. Thus, stimulated interest in higher productivity for commercial production resulted in the development of multi-breed synthetic populations for specific objectives e.g. reproduction, meat quality, meat and wool combinations, wool for textile, carpet wool, fur, and milk production. Today, as many as 418 sheep breeds that have been documented in nearly 75 countries are combinations of two or more distinct breeds, populations and landraces that have records on pedigree and morphological characteristics, and in some cases, production performance similar to those for long established breeds. The present day methods of forming composite breeds are based on a wealth of knowledge and creative skills passed on from breeder to breeder over the centuries, as well as the application of quantitative genetic principles that had demonstrated tremendous success in the development of hybrid corn. Research findings also suggested that in well-planned efforts to create composite breeds, the genetic background of the foundation breeds should be as broad as possible to achieve increased additive genetic variance and heterozygosity from a large proportion of segregating genes and lower probability for the expression of lethal recessive genes. There was always the possibility of lower performance in the composite breeds compared with specific crosses derived from two or more breeds, because of their inability to fully realise any benefit from individual and maternal heterosis (considered important in the mammalian species). Though unlikely, loss in performance of composite breeds could occur from recombination of non-allelic genes and the migration of undesirable genes. Nevertheless, there is opportunity for accelerating genetic response to selection as well as heterosis retention unless a rise in inbreeding occurs in the early generations causing loss in performance. The real operational advantage of a multi-breed composite population lies in raising a single breed with nearly the same productivity instead of two or more breeds for subsequent crossbreeding. Sheep in the domain of developing countries adapt well to severe climatic conditions and arduous terrains that are often not suitable for cultivation or for raising the larger livestock and are fully exploited for their ability to produce a variety of commodities without disrupting the cultural harmony. Likewise, more breeds were developed in sheep than in other livestock species with the United Kingdom, Russian Federation, Australia, Poland, China, France, the United States of America and New Zealand each having established 15 or more composite breeds that offer real advantages. In North America and worldwide, there is further prospect for complementing established breeds with exotic breeds to achieve potential biological ceiling in reproduction, lean muscle growth, feed efficiency and milk yield of sheep. In the future, application of advances in molecular methodologies could possibly lead to the development of composite breeds for specific objectives from a combination of widely different breeds, populations and landraces, probably, even from unrelated species.
Article
Full-text available
Knowledge of the genetic structure and overall diversity of livestock species is important to maximise the potential of genome-wide association studies and genomic prediction. Commonly used measures such as linkage disequilibrium (LD), effective population size (N e ), heterozygosity, fixation index (F ST ) and runs of homozygosity (ROH) are widely used and help to improve our knowledge about genetic diversity in animal populations. The development of high-density single nucleotide polymorphism (SNP) arrays and the subsequent genotyping of large numbers of animals have greatly increased the accuracy of these population-based estimates. In this study, we used the Illumina OvineSNP50 BeadChip array to estimate and compare LD (measured by r 2 and D′), N e , heterozygosity, F ST and ROH in five Australian sheep populations: three pure breeds, i.e., Merino (MER), Border Leicester (BL), Poll Dorset (PD) and two crossbred populations i.e. F1 crosses of Merino and Border Leicester (MxB) and MxB crossed to Poll Dorset (MxBxP). Compared to other livestock species, the sheep populations that were analysed in this study had low levels of LD and high levels of genetic diversity. The rate of LD decay was greater in Merino than in the other pure breeds. Over short distances (<10 kb), the levels of LD were higher in BL and PD than in MER. Similarly, BL and PD had comparatively smaller N e than MER. Observed heterozygosity in the pure breeds ranged from 0.3 in BL to 0.38 in MER. Genetic distances between breeds were modest compared to other livestock species (highest F ST = 0.063) but the genetic diversity within breeds was high. Based on ROH, two chromosomal regions showed evidence of strong recent selection. This study shows that there is a large range of genome diversity in Australian sheep breeds, especially in Merino sheep. The observed range of diversity will influence the design of genome-wide association studies and the results that can be obtained from them. This knowledge will also be useful to design reference populations for genomic prediction of breeding values in sheep.
Conference Paper
Full-text available
Abstract Text: Genomic selection (GS) can be used to select purebreds for crossbred performance (CP). As dominance is the likely genetic basis of heterosis, explicitly including dominance in the GS model may be beneficial for selection of purebreds for CP, when estimating allelic effects from pure line data. The objective of this study was to investigate the benefits of GS of purebreds for CP based on purebred information. Results demonstrate that when at least one of the purebreds involved in a two way crossbreeding system was selected for CP, crossbred offspring had better performance than when selection in both breeds was for purebred performance. Under the hypothesis that crossbreds differ from purebreds due to dominance, GS can be applied to select purebreds for CP without crossbred data, by using a dominance model. Keywords: genomic selection crossbreeding dominance
Article
Naturally hornless cattle are called polled. The possibility to introgress the allele responsible (P) for polledness in a cattle population that is more intensively selected for other traits is investigated in this paper. Gene introgression, generally carried out by several steps of backcrossing and selection takes a long time and may lead to unacceptable genetic loss in other traits or inbreeding. The main objective of the current study was to evaluate the use of genomic selection to speed up the introgression of a target allele in a conventional dairy cattle breeding scheme with overlapping generations. A cattle population and a breeding program were simulated and run over a 12 year selection period. Assuming that the polled population was inferior for overall genetic merit , two selection strategies were evaluated to introgress: i) selection on conventional BLUP-EBV (CBLUP); ii) selection on genomic EBV (GEBV) obtained with the genomic relationship matrix used in BLUP (GBLUP). Both selection strategies were applied with (PSEL) and without (NOPSEL) selection for the single polled locus (P). The overall level of genetic merit, the P allele frequency and the inbreeding level (F) in the new born animals were monitored each, as well as the average genetic gain per year of selection (ΔG). The overall genetic level of new born animals was higher for GBLUP compared to CBLUP, with an average ΔG/year of 8.34% (GBLUP) against 7.49% (CBLUP). The PSEL scenario reduced genetic gain, but P allele frequency increased from 0.130 to 0.415 (CBLUP) and from 0.128 to 0.440 (GBLUP) for PSEL, after 12 years of selection. No substantial changes in allele frequency were recorded for NOPSEL scenarios, both for CBLUP and GBLUP breeding schemes. The overall inbreeding rate for GBLUP were 0.28%/y (NOPSEL) and 0.30%/y (PSEL) and for CBLUP 0.52%/y (NOPSEL) and 0.44%/y (PSEL). In conclusion, application of GS to gene introgression helped to speed up the process of introgression of a gene while simultaneously increasing the genetic gain and reducing the inbreeding rate.
Conference Paper
Abstract Text: The major gene FecL controlling prolificacy and evidenced in the meat Lacaune population was routinely genotyped in the Ovi-Test selection nucleus on all serving sires since 2002 and all replacement ewes since 2010. Litter size at lambing of these genotyped females have been collected and analysed. The population was composed of 71% homozygous ++, 27% heterozygous L+ and 2% mutated LL females. Analyses were performed with a threshold animal mixed model which showed an interaction between thresholds and genotypes. Polygenic heritability on the underlying scale was 0.11 and the gene accounted for 42% of the total additive genetic variance. On average L+ ewes produced 0.4 more lambs than ++ ewes. A new parameterization of the underlying variables and thresholds showed that the gene had an effect on the variability of the trait additionally to the direct variability increase induced by the mean change. Keywords: Major gene Sheep Prolificacy