130 R
.M. Lewis, G. Simm Livestock Production Science 67 2000 129 –141
records of animals in separate flocks and born in 100 females linked by the same reference sires. The
different years could not be fairly compared. intensity and criteria used for selecting reference
With the introduction of best linear unbiased sires and the extent of their use within the scheme
prediction BLUP methodologies to sheep breeding was not explicitly considered. Nimbkar and Wray
programmes, across-flock genetic evaluations are 1991 considered some of these selection strategies
possible. But for BLUP to disentangle genetic from but their combined effect on response in sire refer-
environmental sources of variation in performance, encing schemes has yet to be tested.
genetic relationships or links among animals are Our objective was to investigate the effects of
needed across flocks Garrick, 1991; Kennedy and specific operational characteristics of sire referencing
Trus, 1993. If the degree of linkage is adequate, the schemes in flocks of small size on rates of genetic
size of an individual flock is effectively enlarged, progress and inbreeding for a trait of moderate
since the genetic merit of animals in separate flocks heritability 0.25. This was undertaken using sto-
and born in different years can be directly and chastic simulation. We consider i the selection
accurately compared
Kinghorn and
Shepherd, intensity for reference sires, ii the criteria by which
1990. This allows more intense selection and there- reference sires are chosen and iii the extent to
fore quicker rates of genetic progress. which reference sires are used. For the latter, differ-
Sire referencing schemes are used to establish ent combinations of the number of reference sires
genetic links across flocks. In these co-operative used and the number of ewes mated to reference
breeding programmes a team of rams or reference sires were assessed.
sires is selected, typically from among member flocks. Each member of the scheme then uses some,
although not necessarily the same, rams from this
2. Methods
team to mate to a proportion of ewes within their own flock.
Stochastic simulation was used to investigate Several studies have focused on how reference
design alternatives for sire referencing schemes. The sires may be used to link dispersed flocks or herds to
population simulated was that of terminal sire sheep improve the accuracy of genetic evaluations of
in the United Kingdom, which have the common young male progeny tested in a sample of them
goal to improve lean meat production Guy and Foulley and Clerget-Darpoux, 1978; Hudson et al.,
Croston, 1994. This was done to model realistic 1980; Foulley et al., 1983; Miraei Ashtiani and
schemes in practice. The growth and carcass traits James, 1991, 1992. However, besides this role,
that dominate in such programmes are moderately reference sires themselves are a means to genetic
heritable. improvement. This is particularly important where
flock or herd sizes are small with reference sires 2.1. Genetic model
mated to a substantial proportion of the available females. In such situations, the optimal use of the
The lean growth trait to be improved the goal reference sires to accelerate genetic gain and the
trait was recorded in both sexes at around 150 days consequence of their use on inbreeding has not been
of age. An additive infinitesimal model was assumed. investigated.
For a base population of unrelated animals, the true Rates of genetic gain and inbreeding have been
breeding values were obtained from a normal dis- compared when flocks were independent or linked
tribution with mean zero and an initial additive
2
though reference sires both by prediction Morris et variance s
of 0.25. The number of base males
a
al., 1980 and by simulation Hanocq et al., 1996; and females created depended on the size of each
Roden, 1996. With sire referencing, genetic progress flock.
was increased by over 30 and inbreeding level was The trait litter size was also simulated to model
more than halved. However, these studies tested a reproductive rate within flocks. It was assumed that
single design for a sire referencing scheme with litter size was uncorrelated with the goal trait and
flocks of equal and comparatively large size at least was not under selection. True breeding values for
R .M. Lewis, G. Simm Livestock Production Science 67 2000 129 –141
131
litter size were drawn from a normal distribution each litter size category. Since extra variation was
2
with mean zero and s of 0.10. Permanent en-
introduced by flock and year, the variance of the
a 2
liability s was greater than one. This additional vironmental variation for the trait was assumed to be
2
flock and year variation s 2 1 had a non-linear negligible.
effect on the proportion of ewes placed in each litter For each trait, the true breeding values of off-
size category. Depending on the location of a spring
were simulated
as TBV 5 1 2TBV 1
i s
threshold, a change of equal size in threshold value TBV 1 m , where TBV , TBV and TBV are the true
d i
i s
d
defined a different proportion of the liability or area breeding value of the offspring i , its sire s and its
of the standard normal density. To account for this dam d , respectively. The Mendelian sampling term
non-linearity, an adjusted threshold value t was m was taken from a distribution with mean zero
i i
2
used to assign ewes to litter size categories. The and variance 1 2 1 2 F 1 F 2 s , where F
f g
s d
a s
value t was found using a search algorithm that
and F are the inbreeding coefficients of the sire and
i d
minimised the function dam, respectively. Inbreeding coefficients were ob-
tained using the algorithm of Meuwissen and Luo
` 2
2 t 2 t 1
i i
1992. The animal’s sex was assigned at random, ]]]]
]]]] p 2
E
qt exp dt
]]]
S D
i i
2 i
S D
2
with equal probability of being male or female. 2s 2 1
2ps 2 1
œ
2`
where qt is the value of the integral of the standard
i
normal distribution between t and infinity Amer, 2.2. Phenotypic model
i
personal communication. Litter size changes sys- tematically with age. Therefore the proportion of
For the goal trait, fixed environmental effects of ewes in each litter size category, and thus the
flock, year, dam age, lamb sex and rearing type adjusted threshold values, were different for 2, 3 and
single or multiple were generated. Flock and year 4-year and older ewes. The prescribed input aver-
effects were obtained by taking a random number age litter size values were equal to those obtained in
from a normal distribution with mean zero and the simulation and are shown in Table 1 by age
variance 0.20 for flock and 0.05 for record year. category. On average, ewes that lambed produced
Offspring of ewes 3 years old and older had an litters of 1.75 lambs.
advantage of 0.20 units in performance. Female offspring, and those reared in multiple litters, had
0.85 and 0.35 unit disadvantages, respectively, in 2.3. Reproductive and mortality parameters
performance for this trait. The size of these effects was based on evaluations of industry sire referencing
2.3.1. Conception rate schemes as described by Mercer et al. 1994. For
Rams and ewes were considered reproductively litter size, environmental influences of flock and year
mature at 6 and 15 months of age, respectively. were simulated as for the goal trait.
There was one mating season per year that lasted for The phenotypic value for each trait the goal trait
three oestrous cycles a total of 51 days and all ewes and litter size was generated as the sum of the fixed
were assumed to be cycling at the start of season. In environmental effects, the residual value obtained at
each of the oestrous cycles one and two, 65 of the random from a normal distribution with mean zero
2
and variance 1 2 s , and the true breeding value.
a
Table 1
Litter size was assumed to have a continuous,
Percentage of ewes within a lambing category
normal underlying distribution liability. Variation
Ewe age years Percentage of ewes within a lambing
on this liability scale was both genetic and en-
category
vironmental flock, year and residual values in
Single Twin
Triplet
origin. When a ewe’s liability exceeded specific
2 39.2
59.1 1.7
threshold values t on the liability scale, she gave
i
3 25.4
69.3 5.3
birth to one, two or three lambs. The thresholds
4 and older 31.6
61.8 6.6
corresponded with a set proportion of ewes p in
i
132 R
.M. Lewis, G. Simm Livestock Production Science 67 2000 129 –141
ewes conceived; at cycle three, of the remaining tion was then carried out for 15 years with genetic
open ewes, 20 conceived. Ewes mated to reference evaluation of all animals once each year. In all
sires RS had a 65 chance of conceiving to a scenarios a team of six rams were made available as
single artificial insemination AI. Ewes that failed to reference sires RS. RS were always chosen from
AI were then mated naturally to a ram from their among rams born in the member flocks of the
respective flock over the final two oestrous cycles of scheme.
the mating season. Overall, about 90 of the ewes Depending on the scenario tested, the specified
mated lambed. number of RS used within a flock was chosen at
random from the team of six rams. When more than 2.3.2. Mortality and culling
one RS were used in each flock, only one of these The rate of mortality from birth to recording of the
RS was re-used in the following year. On average goal trait was 13 for singletons, 13.6 for twins
the generation interval for reference sires and natural and 15.6 for triplets. Between recording and first
service sires was 2.7 years. mating 3 of the animals died. Thereafter, an annual
mortality rate of 2.5 was assumed. All rams chosen 2.5.1. Selection intensity
as sires were culled at 4 years of age. Ewes could The intensity at which rams were selected was
lamb a maximum of four times and thus were culled varied. This applied to both RS and rams chosen
at about 6 years of age. The average generation from within flocks for home use. Three selection
interval for ewes was 3.5 years. intensities were considered. Firstly, rams with the
highest ranking best for the selection criteria were 2.4. Establishing flocks
selected. For the RS team, the rams chosen were within the top 0.5 of the candidates available.
Fifteen flocks with sizes between 40 and 140 Rams were also selected at random from among the
breeding ewes average size of 70 with standard top sixth and top third once ranked on the selection
deviation 30 were evaluated. Before the start of sire criteria. No more than one ram from a full-sib family
referencing, each flock underwent 10 years of ran- in a birth year was chosen. Enough ewes were
dom selection. One half of these flocks used only chosen each year to maintain a constant flock size;
unrelated rams from outside the scheme. The remain- around 26 of the breeding ewes in a flock were
ing flocks used a mix of outside and homebred rams. replaced annually.
The mean genetic value for outside rams was equal to that of contemporary animals born within the
2.5.2. Selection criteria flocks under evaluation. That is, the true breeding
Rams and ewes were chosen on either of two value of outside rams was obtained from a normal
selection criteria: their estimated breeding value distribution with mean and additive variance ger-
EBV or their phenotype mass selection for the mane to the year of evaluation. Each flock used a
goal trait. BLUP EBVs were obtained by using an minimum of two rams, with each ram mated to about
individual animal model with genetic groups and 20 ewes. Rams were first used as sires at about 18
fitting flock, year, dam age, lamb sex and rearing months of age and at most for two mating seasons
type as fixed terms. Animals with unknown ancestors within a flock.
born during the years preceding sire referencing were Mate assignment was based on a strategy that
assigned to one genetic group. Thereafter, all outside avoided mating relatives. This involved two levels of
rams born in a year were assigned to a separate priority. Firstly, son–dam, daughter–sire and full-sib
genetic group. With mass selection, no pre-correction matings were avoided. Secondly, where possible,
for the influence of environmental factors was made. half-sib matings were avoided.
However, since selection decisions were made within year and sex, the noise introduced by these fixed
2.5. Sire referencing scheme effects was necessarily accounted for. For replace-
ment decisions made within-flock the selection of Following the period of random selection, the 15
homebred rams and ewes, flock effects were also flocks joined into a sire referencing scheme. Selec-
accounted for.
R .M. Lewis, G. Simm Livestock Production Science 67 2000 129 –141
133
2.5.3. Reference sire usage For animals born each year, the within-flock and
Scenarios in which 0, 1, 2 or 3 RS were used were between-flock additive variance and the accuracy of
considered. Since the specified number of RS needed selection the correlation between the true breeding
was chosen at random by each flock from the RS value for the goal trait and the selection criteria
team of six rams, individual flocks did not necessari- were computed. The selection differential was calcu-
ly use the same RS. When no RS were chosen, lated as the average superiority of the selected
selection was strictly within-flock. A RS was mated parents for the selection criteria weighted by progeny
to 5, 10, 15, 20 or 30 ewes. Since the smallest flock number. This statistic was computed separately for
had 40 ewes, only some combinations of the number RS across flocks, and for homebred sires and ewes
of RS used and the number of ewes mated were within flocks. All scenarios evaluated were repli-
considered Table 2. Within a flock, in total 10, 20 cated 100 times and the results averaged across the
or 30 ewes were mated to RS; this corresponds with replicates and, where appropriate, flocks.
|15, 30 and 45 of the ewe population being mated to RS. If excess ewes were available within a
flock above those mated to RS, these were mated to
3. Results