Published December 11, 2014
Effect of Culling on Selection Response Using
Phenotvnic Selection or Best Linear Unbiased
Prediction of Breeding Values
in Small, Closed Herds of Swine'
I A
Daryl L. Kuhlers2 and Brian W. Kennedy
Department of Animal and Poultry Science, University of Guelph,
Guelph, Ontario N1G 2W1, Canada
ABSTRACT: Records from 7,200 separate closed
herds with either 12 or 25 sows that were mated to
either four or eight boars per year were simulated
by computer. Effects of selection method, herd
size, and contemporary group variability on average genetic change, genetic variance, and inbreeding over 10 yr of selection were analyzed for traits
with heritabilities of .1, 3, and .6. Selection of
replacement animals was on individual phenotype
or BLUP of breeding value using a reduced animal
model. For both of these selection methods, two
culling schemes were imposed: 1) based only on
involuntary culling because of losses due to
conception rate and age and 21 when a n available
replacement animal was projected to be superior
to an existing breeding animal in the herd in
addition to the involuntary culling. The contemporary group standard deviation was set at either
.I or .5 of a phenotypic standard deviation.
Selection with BLUP gave 72, 36, and 12% more
genetic improvement for heritabilities of .1, .3, and
.6, respectively, than selection on individual
phenotype after 10 yr. However, inbreeding increased 20 to 52% more rapidly and there was a
decrease in genetic variance. Culling based on
Scheme 2 increased genetic improvement over
Scheme 1 by about 75% with coincident increases
in inbreeding level and decreases in genetic
variance. The largest changes in inbreeding and
genetic variance were associated with culling on
BLUP. Culling when a superior animal was available with individual phenotype had little effect on
inbreeding and genetic variance. Use of four boars
rather than eight boars and 25 rather than 12 sows
per herd increased genetic response. Use of four
boars also increased inbreeding and deceased
genetic variance. Genetic variance was higher in
herds with 25 sows, but the size of the sow herd
had little effect on inbreeding. Contemporary
group variation influenced only the genetic response of individual phenotypic selection with
culling.
Key Words: Best Linear Unbiased Prediction, Selection, Culling, Pigs, Herd Sizes
J. Anim. Sci. 1992. 70:2338-2348
Introduction
Swine evaluation programs have been established in many countries to aid breeders in their
selection decisions. Generally, these evaluation
programs include a measure of growth, daily gain
'Financial support was provided by the Ontario Ministry of
Agric. and Food and Auburn Univ. The authors gratefully
acknowledge the simulation program written by Gail M.
Belonsky.
2 0 n leave from the Dept. of Anim. and Dairy Sci., Auburn
Univ., AL 36849-54 15.
Received June 19, 1991.
Accepted March 21, 1992.
or days to market weight, and a measure of
carcass composition (usually ultrasound backfat
thickness measurements) and may also include a
measure of sow productivity. The newer genetic
evaluation programs are based on BLUP techniques, in contrast to simple evaluation of the
phenotype of the candidate for selection. Theoretically, for large populations in which selection is
practiced across management units and time,
BLUP is most efficient.
However, innovative purebred breeders and
researchers with experimental swine herds may be
interested in traits not measured by available
evaluation programs and conduct selection for
these traits in small, closed herds. Accumulation
2338
CULLING WITH BLUP AND MASS SELECTION
of inbreeding can be of concern in small, closed
populations. There are two questions that need to
be answered for these herds: 11 What is the best
selection procedure? and 2) How many sires
should be used to obtain maximum genetic improvement while keeping inbreeding a t an acceptable level? (Quinton et al., 1992). Belonsky and
Kennedy (19881, by use of simulation, have shown
that culling on estimated breeding values along
with the use of BLUP of breeding values resulted
in more genetic improvement than selection on
phenotype with and without this type of culling in
100-sow herds. However, in the Ontario swine
testing program, the average numbers of sows per
herd with tested progeny per year are 10.6, 15.8,
22.3, and 30.1 for the Hampshire, Duroc, Landrace,
and Yorkshire breeds, respectively. Because many
breeder herds are considerably smaller than 100
sows, the objective of the present study was to
determine whether culling on estimated breeding
value and BLUP estimation of breeding values
were superior to phenotypic selection in closed
sow herds of 12 or 25 cows mated to four or eight
boars with two levels of variability in contemporary group effects.
Materials and Methods
Simulation. Ten years of selection in 7,200 s e p a
rate herds were simulated. Simulation of the data
followed the procedures of Belonsky and Kennedy
(19881, which was to represent “typical” Canadian
purebred breeder herds. Age at first breeding for
selected gilts was 235 k 25 d and for boars was
210 d. Gestation length was 114 d and litter size
was 9 for gilts and 10 for sows. Sows in the
breeding herd were culled after their fifth litter or
when they did not conceive after two consecutive
breeding cycles. Boars were culled after reaching 3
yr of age along with random loss of one additional
boar each year. These maximum age levels portray industry generation intervals (Kennedy et al.,
1986). Probabilities of survival from birth to
measurement age, a replacement boar’s being
fertile, and conception rate in a single mating were
.9, .9, and .85, respectively.
An infinitesimal additive genetic model (Bulmer,
1980) was assumed. Genotypic values (GV) for the
base population of animals were sampled from a
normal distribution with mean equal to zero and
variance equal to .1, .3, or .6 according to GV =
+ v*og, where v is a random normal deviate (ISML,
1981). Parities of the base population sows were
randomly assigned from a uniform distribution
(ISML, 1981) from 0 to 5, and comparable designations for ages of base population boars were from
0 to 3 yr.
2339
Sows and boars were mated at random, with the
exception that full-sib and parent-offspring matings were avoided. Simulation of progeny genotypic values were according to GV, = p + .5GV, +
.5GVd + ~,[.25(1- Fs)1.5~g
+ vd[.25(1- Fd)].%(Tg, where
F, and Fd are the inbreeding coefficients of the sire
and dam, respectively. The parts of the equations
involving v, and Vd represent Mendelian sampling
of gametes from the sire or dam, where v, and Vd
are uncorrelated standard normal random deviates.
Phenotypic values (PV)were simulated as PV =
t + GV + vo,, where t is a fixed contemporary
group effect for the 3-mo time period in which the
record was made and 0, is the environmental
standard deviation. For all simulations, the underlying within-contemporarygroup phenotypic variance was 1.0. Contemporary-group constants were
simulated from a normal distribution with mean of
zero and a standard deviation of .1 or .5.
Progeny were assigned a generation number
one greater than the average of their parents’
generation numbers. When fractions of a generation were encountered, because of overlapping
generation intervals, the generation number was
rounded up to the next whole generation.
Selection. Replacement animals were selected at
28-d intervals from among the animals available
during the previous 28-d period. Selection was
either at random (RAND), on individual phenotypic performance (IND), or on BLUP of breeding
value under a n involuntary culling scheme due to
conception rate and age. Truncation selection was
practiced for IND and BLUP selection.
A second culling scheme was imposed in addition to the involuntary culling. With this scheme,
existing breeding animals were culled from the
herd and replaced with available young breeding
stock that were “expected” to be superior for
individual phenotype, which was not deviated
from contemporary group average (IND-CULL) or
BLUP estimates of breeding value (BLUP-CULL).
Selection was across contemporary groups when
a n older animal already in the breeding herd was
culled in favor of a superior, younger replacement
animal. The BLUP evaluations were based on a
reduced animal model (Quaas and Pollak, 1980;
Hudson and Kennedy, 1985) and were run every 28
d.
Therefore, five selection methods (RAND, IND,
IND-CULL, BLUP, and BLUP-CULL),herds of four
different sizes (4 males and 12 females, 4 males
and 25 females, 8 males and 12 females, and 8
males and 25 females), three heritabilities c.1, 3,
and .6), and two contemporary group standard
deviations (.1 and .5) were simulated. Forty replicate herds were simulated in which the sow herd
size was 25, and 80 replicate herds were simulated
in which the sow herd size was 12. In addition to
2340
KUHLERS AND KENNEDY
the average genetic merit of the populations,
average actual genetic variance and average
inbreeding coefficients for each herd were calculated based on animals that were performancetested in the 10th yr.
Statistical Analyses. Weighted, generalized least
squares analyses (SAS, 1989) were conducted on
the average genetic values, genetic variances, and
inbreeding coefficients in the loth yr for each of
the three heritabilities using the following model:
+ Mi + Sj + Gk + MSij + MGik + SGjk + eijkl,
Yijkl =
where M, S, G, and e represent the effects of
method of selection, size of herd, contemporary
group standard deviation, and residual, respectively. Weighted analyses were used because the
number of replicates simulated differed between
the sow herd sizes. Weights used in the analyses
of each of the traits were (standard error)t2. Sums
of squares associated with individual df orthogonal contrasts were partitioned in the following way: 1) selected lines vs random selection
(IND, IND-CULL, BLUP, and BLUP-CULL minus
RAND), 2) BLUP selection vs individual selection
(BLUP AND BLUP-CULL minus IND and INDCULL), 31 culling vs no culling (IND-CULL and
BLUP-CULLminus IND and BLUP), and 4) selected
lines by culling interaction (IND-CULL and BLUP
minus BLUP-CULL and IND). For S, the sums of
squares were partitioned as follows: 1) four vs
eight sires, 2) 12 vs 25 dams, and 3) number of sires
by number of dams interactions. The G effect only
contained one df. In addition, interactions of the
main effects were split into single df comparisons
for tests of hypotheses.
Results and Discussion
Tables 1 through 3 give the average genetic
merit, remaining genetic variance, and inbreeding
coefficients of the populations a t the end of 10 yr
for each selection method, herd size, and contemporary group standard deviation. The individual df
tests of hypotheses are in Tables 4 through 6.
Selection Method. As expected, selection was
effective. Average genetic improvement after 10 yr
was .456, 1.198, and 2.250 phenotypic standard
deviations (PSD) for heritabilities of . l , .3, and .6,
respectively (Table 4). These responses were about
one-half of those observed by Belonsky and Kennedy (1988) using the same herd age limits and
survival and fertility rates as in the present study.
They were lower partly because of the lower
selection intensities resulting from smaller herd
sizes in the present study. Associated with these
changes were decreases in genetic variance (Table
5) and a n increase of > .04 in the inbreeding
coefficient (Table 6).
Selection by BLUP resulted in .238, .361, and .263
PSD additional selection response compared to
selecting on individual phenotype (Contrast 21, for
heritabilities of . l , .3, and .6, respectively. This
corresponds to a 72, 36, and 12% advantage of
BLUP over individual selection. However, inbreeding was increased by .03 to .07 (a 20 to 52%
increase) and genetic variances were decreased.
Similarly, culling of breeding animals when a n
expected superior prospective replacement was
available gave a n additional response of .251, .638,
and 1.239 PSD for heritabilities of .l,.3, and .6,
respectively (Contrast 3). This is equivalent to 77,
73, and 77% additional improvement from culling
relative to not culling on the availability of
expected superior replacement animals.
This additional improvement is due in part to
the shorter generation intervals of the herds using
culling. Randomly selected herds averaged, across
all herd sizes and contemporary group standard
deviations, 2.0 yr per generation, similar to those
found with individual selection and BLUP selection without culling. In herds using culling with
individual selection, generation intervals averaged
1.8 to 1.9 yr, depending on the heritability. In herds
that used BLUP with culling, generation intervals
averaged 1.7 to 1.8 yr. In addition to the shorter
generation interval contributing to additional genetic gain as a result of culling, further genetic
gain is realized with culling because superior
animals are used in the herd when they are
estimated to be superior. The BLUP and individual
selection methods without culling take the best
available replacement only when a replacement
animal is needed and animals superior to the
existing breeding herd can be passed over if a
replacement is not needed at that time. With
culling, however, superior animals are not
bypassed when they are estimated to be superior.
The advantages from selection on BLUP and
culling in these 12- and 25-SOW herds were in the
same direction as those found by Belonsky and
Kennedy (1988)for 100-sowherds but were of lesser
magnitude. Culling, however, resulted in a n additional 5% loss in genetic variance, as a proportion
of the base genetic variance, compared to not
culling. Also, there was an increase in average
inbreeding of .048 to .058. Culling based on either
individual phenotype or BLUP estimates gives
substantial improvements regardless of the heritability but will also result in increased inbreeding
and loss in genetic variance. Some form of
reasonable culling based on estimated breeding
value should be a recommended practice rather
than culling on the basis of some prearranged age
and level of reproductive performance criteria that
are not related to the breeding value estimates of
the animals. This will turn over generations more
CULLING WITH BLUP AND MASS SELECTION
rapidly and incorporate in the breeding herd
superior replacements whenever they are available.
There was a n interaction between BLUP and
individual phenotypic selection by culling and no
2341
culling (Contrast 4). Increases in genetic merit and
inbreeding coefficient and the decreases in genetic
variance between BLUP with culling and BLUP
without culling were larger than those between
culling on individual phenotype and not culling on
Table 1. Least squares means for genetic merit after 10 years of selection
Heritabilitv
Item
4-12X
4-12X
4-25X
4-25X
8-12x
8-12x
8-25 X
8-25 X
.3
.1
Selection method [MI
BLUP
BLUP-CULL
IND
IND-CULL
RAND
Herd size [SI
4 Boars, 12 sows
4 Boars, 25 sows
8 Boars, 12 sows
8 Boars, 25 sows
Contemporary group (GI
.1 SD
.5 SD
M x S means
BLUP x 4-12
BLUP x 4-25
BLUP x 8-12
BLUP x 8-25
BLUP-CULL x 4-12
BLUP-CULL x 4-25
BLUP-CULL x 8-12
BLUP-CULL x 8-25
IND x 4.12
IND x 4-25
IND x 8-12
IND x 8-25
IND-CULL x 4-12
IND-CULL x 4-25
IND-CULL x 6-12
IND-CULL x 8-25
RAND x 4-12
RAND x 4-25
RAND x 8-12
RAND x 8-25
M x G means
BLUP x .1
BLUP x .5
BLUP-CULL x .I
BLUP-CULL x .5
IND x .1
IND x .5
IND-CULL x .1
IND-CULL x .5
RAND x .1
RAND x .5
S x G means
.1
.5
.1
.5
.1
.5
.1
.5
.6
,390 f
.751 f
.263 f
.402 f
-.004 f
,007
.007
,007
,007
,006
,939 f
1.803 f
,804 f
1.215 f
-.008 f
.010
,010
,010
,010
.010
1.640 f
3.100 f
1.598 f
2.617 f
-.011 f
,016
,017
,015
,015
,017
.326 f
,454 f
,271 f
,391 f
,005
,008
,004
,006
,874 f
1.196 f
,736 f
,996 f
.008
.011
.007
.008
1.640
2.262
1.399
1.855
,012
,019
,012
.013
,365 f ,004
,356 f ,004
.965 f ,006
,937 f ,006
f
f
f
f
1.820 f ,010
1.758 f ,010
,327 f
,497 f
274 f
,462 f
,696 f
,929 f
,583 f
,798 f
,221 f
,338 f
,189 f
,302 f
,382 f
,498 f
,324 f
,406 f
,002 f
,008 f
-.018 f
-.012 f
,012
,017
,009
.017
.011
.017
.011
.012
.013
,016
.010
,014
.013
.017
.009
.013
.009
.018
.010
.011
,803 f
1.192 f
,747 f
1.012 f
1.747 f
2.273 f
1.347 f
1.847 f
366 f
1.041 f
.654 f
.856 f
1.149 f
1.470 f
.958 f
1.285 f
,003 f
.015 f
-.028 f
-.021 f
,019
,026
,014
,020
.020
,025
,014
.018
,017
,023
,017
,021
,020
,023
,016
.018
,015
,029
,015
,017
1.356 f
2.223 f
1.272 f
1.711 f
2.956 f
3.820 f
2.443 f
3.182 f
1.354 f
1.968 f
1.283 f
1.785 f
2.528 f
3.277 f
2.035 f
2.628 f
,005 f
,021 f
-.040 f
-.029 f
,393 f
.387 f
,754 f
.749 f
.264 f
,261 f
.416 f
,389 f
-.001 f
-.008 f
.010
.010
,009
,009
,009
,009
,009
,009
,008
,008
.937 f
,940 f
1.801 f
1.806 f
,798 f
311 f
1.297 f
1.133 f
-.009 f
-.007 f
,014
,014
,013
,013
,014
,014
,014
,013
,013
,013
1.639
1.642
3.098
3.102
1.605
1.591
2.770
2.464
-.013
-.009
f
f
f
f
f
f
f
.024
,024
,021
,022
,023
,020
,022
.022
.322 f
,329 f
,475 f
,433 f
,270 f
,271 f
,394 f
,389 f
,007
,008
,011
,011
,006
,006
,008
,009
,878 f
,869 f
1.202 f
1.194 f
.755 f
,716 f
1.025 f
,967 f
.012
,011
.016
,016
.010
,010
.012
,012
1.672 f
1.607 f
2.282 f
2.241 f
1.437 f
1.360 f
1.887 f
1.824 f
.018
,017
,028
.027
,017
,017
,018
,018
,026
,046
,029
,027
,029
,046
.030
,029
,029
,040
,023
.026
,026
,037
,025
.030
,025
.049
,025
,030
f ,022
f ,022
k
2342
KUHLERS AND KENNEDY
individual phenotype. Indeed, most all of the
changes in genetic variance and inbreeding level
connected with culling occurred with culling on
BLUP estimates rather than with culling on
individual phenotypic values. Because there were
small changes in genetic variance and inbreeding
coefficients due to culling on individual phenotype, selecting on individual phenotype should
always be accompanied by culling to maximize
genetic response, for population structures similar
Table 2. Least squares means f o r genetic variance after 10 years of selection
Heritability
Item
Selection method [MI
BLUP
BLUP-CULL
IND
IND-CULL
RAND
Herd size (SI
4 Boars, 12 sows
4 Boars, 25 sows
8 Boars, 12 sows
8 Boars, 25 sows
Contemporary group (GI
.1 SD
.5 SD
M x S means
BLUP x 4-12
BLUP x 4-25
BLUP x 8-12
BLUP x 8-25
BLUP-CULL x 4-12
BLUP-CULL x 4-25
BLUP-CULL x 8-12
BLUP-CULL X 8-25
IND x 4-12
IND x 4-25
IND x 8-12
IND x 8-25
IND-CULL X 4-12
IND-CULL X 4-25
IND-CULL x 8-12
IND-CULL X 8-25
RAND x 4-12
RAND x 4-25
RAND x 8-12
RAND x 8-25
M x G means
BLUP x .1
BLUP x .5
BLUP-CULL x . I
BLUP-CULL x .5
IND x .1
IND x .5
IND-CULL x .I
IND-CULL x .5
RAND x .1
RAND x .5
S x G means
4-12X
4-12X
4-25x
4-25 X
8-12x
8-12x
8-25 X
8-25 x
.3
.1
.1
.5
.1
.5
.1
.5
.1
.5
.6
.0705 f
.OS99 f
,0736 f
,0740 f
,0762 f
,0004
.0003
,0004
.0005
.0005
,2110 f
,1829 f.
.2190 f
.2135 f
.2287 f
.0016
.0658 f
,0682 f
,0740 f
.0754 f
,0003
.0004
,0004
,0005
.1980 f
,2045 f
,2193 f
,2223 f
,0013
.0016
.0012
.0012
,0017
,0016
,0018
,0015
.4212 f
,3615 f
.4155 f
,4079 2
,4575 f
,0017
.0012
,0016
.0014
.3894 f
.4305 f
,4256 f
,4324 f
,0013
,0017
,0012
,0019
,0016
.0710 f .0003
,0708 f .0003
.2109 f ,0010
,2111 f ,0010
.4100 f .0010
,4155 f ,0010
,0665 f
.0653 f
.0732 f
.0769 f
.0557 f
,0541 f
,0676 f
,0625 f
.0708 f
.0721 f
.0739 f
,0774 f
,0671 f
,0715 f
,0782 f
,0794 f
.0689 f
.0778 f
.0772 f
,0810 f
,0007
,0008
,0008
,0009
,0004
.0006
.0007
.0006
.0009
,0009
,0007
,0010
.0007
.0009
,0010
,0015
,0008
,0012
,0007
,0010
,2093 f
,2072 f
,2201 f
,2074 f
,1699 f
.1657 f
.2036 f
,1924 f
.2042 f
,2116 f
,2200 f
,2402 f
,1997 f
.2045 f
.2212
,2285 f
.2068 f
,2335 f
,2316 f
,2340 f
.0035
.0038
.0028
,0029
,0022
.0026
,0023
.0025
.0030
.0036
,0028
,0039
,0028
,0035
,0030
.0034
.0030
.0046
,0028
,0037
,4055 f
.4161 f
,4279 f
,4355 f
,3490 f
,3418 f
,3796 f
,3758 f
.3877 f
.4080 f
,4242 f
,4422 -+
.3910 f
,3848 f
,4332 f
,4225 f
,4137 f
,4670 f
,4632 f
,4860 f
,0033
,0040
,0026
,0038
,0024
.0029
,0020
,0027
.0024
,0037
,0025
,0042
,0026
,0028
,0028
,0030
,0032
,0049
,0030
,0039
,0705 f
,0705 f
,0600 f
,0600 f
.0741 f
,0730 f
,0738 f
,0743 f
.0763 f
,0761 f
.0006
,0006
,0004
.0004
,0006
,0006
,0007
.0007
.0006
.0006
.2111 f
,2110 f
,1827 f
,1831 f
,2181 f
.2199 f
.2145 f
,2124 f
,2284 f
,2291 f
,0023
,0023
.0017
.0017
,0024
,0023
,0024
,0021
,0025
.0025
,4210 f
,4215 f
.3613 f
,3618 f
,4131 f
,4180 f
.3975 f
,4182 f
,4571 _+
.4579 f
,0024
,0024
,0017
,0017
.0021
,0023
,0018
,0021
.0026
,0026
.0657 f
,0659 f
,0888 f
,0677 f
,0737 f
,0743 f
,0757 f
,0752 f
,0004
,0004
,0005
.0006
,0005
,0005
.0006
,0006
,1994 f
,1966 f
,2040 f
,2050 i
.2203 f
,2183 f
,2201 f
,2245 f
,0019
,0018
,0023
,0022
.3884 f
,3904 f
,3987 f
.4084 f
,4234 f
.4278 f
,4295 f
,4353 f
,0017
.0018
,0022
,0023
,0016
,0016
,0021
,0023
*
,0018
,0017
,0020
,0021
CULLING WITH BLUP AND MASS SELECTION
to those simulated in the present study.
Herd Size. Using four boars per year instead of
eight boars per year (Contrast 5) increased genetic
response .059, .170, and .324 PSD. Herds of 25 sows
rather than 12 sows (Contrast 6) gave .125, ,292,
and ,539 PSD of additional response for heritabili-
2343
ties of . l , .3, and .6, respectively. This agrees with
the conclusions of DeRoo (1988b1, who indicated
that the number of sows in the herd (in a
comparison of 25, 50, 100, and 150 sows) was more
important than was the number of boars used
(comparing 5, 10, 15, and 20 boars) in 25-yr
Table 3. Least squares means for inbreeding coefficient after 10 years of selection
Heritability
Item
.1
Selection method (M)
BLUP
BLUP-CULL
IND
IND-CULL
RAND
Herd size (SI
4 Boars, 12 sows
4 Boars, 25 sows
8 Boars, 12 sows
8 Boars, 25 sows
Contemporary group (GI
.1 SD
.5 SD
M x S means
BLUP x 4-12
BLUP x 4-25
BLUP x 8-12
BLUP x 8-25
BLUP-CULL x 4-12
BLUP-CULL x 4-25
BLUP-CULL x 8-12
BLUP-CULL x 8-25
IND x 4-12
IND x 4-25
IND x 8-12
IND x 8-25
IND-CULL x 4-12
IND-CULL x 4-25
IND-CULL x 8-12
IND-CULL x 8-25
RAND x 4-12
RAND x 4-25
RAND x 8-12
RAND x 8-25
M x G means
BLUP x .1
BLUP x .5
BLUP-CULL x .I
BLUP-CULL x .5
IND x .1
IND x .5
IND-CULL x .I
IND-CULL x .5
RAND x ,1
RAND x .5
S x G means
4-12 X
4-12 x
4-25 X
4-25 x
8-12 x
8-12x
8-25x
8-25X
.1
.5
.1
.5
.1
.5
.1
.5
.3
.6
,1573 f
,2813 f
.1334 f
.1414 f
.1319 f
,0022
,0033
,0015
,0017
,0016
,1536 f
,2463 f
.1342 f
.1567 f
.1319 f
,0013
,0019
.0011
,0014
,0011
,1571 f
,2230 f
,1441
.1737 f
,1319 f
*
.0008
.0011
,0006
,0009
,0007
f
f
f
f
,0017
,0027
,0012
,0017
.2093 f
.2119 f
,1189 f
,1180 f
,0012
,0016
.0008
,0012
.2107
.2109
.1255
.1196
f
f
f
f
.0007
,0010
,0005
,0007
.2095
.2147
.1179
.1181
,1653 f ,0013
,1649 f ,0013
.1648 f ,0009
,1643 f ,0009
,1663 f ,0005
,1656 f ,0005
,1922 f
,2094 f
.1139 f
,1136 f
.3157 f
,3503 k
,1757 f
,2036 f
,1685 f
.1880 f
,1034 f
,0939 f
,1978 f
.1786 f
.0985 f
.0928 f
.1734 f
.le94 f
.0982 f
.0868 f
.0034
.0087
.0031
.0035
.0049
.0092
,0040
,0067
,0032
.0038
.0022
,0026
,0042
,0043
,0024
,0022
,0035
.0045
,0017
,0023
,1904 f
,1953 f
,1140 f
,1147 f
,2984 f
.3289 f
,1869 f
,1910 f
,1708 f
.le94 f
,1047 f
,0917 f
.2135 f
.1988 f
.llO6 f
.lo58 f
,1734 f
.1694 f
.0982 f
,0868
*
.0025
.0035
,0017
.0025
.0034
.0048
,0024
,0042
,0022
.0029
.0016
,0017
.0032
,0033
,0019
,0023
,0024
,0031
,0012
,0016
,1881 f
,2163 f
,1166 f
,1075 f
,2739 f
,2716 f
,1689 f
.1796 f
,1841 f
,1801 f
,1088 f
,1034 f
,2343 f
,2174 f
,1222 f
.1208 f
.1734 f
.1694 f
.0982 f
.0868 f
,0014
,0023
,0012
.0015
.0019
.0028
,0016
,0021
.0014
.0014
,0011
.0011
,0018
,0022
.0013
,0016
.0015
,0019
,0007
,0010
,1578 f
,1568 f
,2618 f
,2608 f
,1319 f
,1350 f
,1427 f
,1400 f
,1321 f
,1317 f
,0029
,0029
,0042
,0042
,0020
,0021
,0024
,0022
,0021
.0021
,1533 f
,1539 f
,2460 f
,2467 f
,1343 f
,1340 f
,1585 f
.1549 f
,1317 f
,1322 f
,0018
.0018
,0025
.0025
.0014
,0015
,0019
,0018
.0014
.0014
.1569 f
,1574 f
.2228 f
,2232 f
,1453 f
,1429 f
.1746 f
,1727 f
,1318 f
,1321 f
,0011
,0011
,0014
,0014
,0009
,0009
,0012
,0013
.0009
,0009
.2077 f
,2113 f
,2173 f
,2121 f
,1180 f
,1178 f
,1181 f
,1182 f
.0024
,0025
.0035
.0035
,0017
,0017
,0022
,0022
,2103 f
,2083 f
,2111 f
,2128 f
,1198 f
,1179 f
,1178 f
.1183 f
.0017
.0017
.0022
.0023
.0011
,0011
,0016
,0015
,2119 f
,2096 f
,2099 f
,2120 f
,1228 f
,1222 f
,1205 f
,1187 f
,0010
.0010
.0013
.0013
,0007
.0008
.0009
.0009
2344
KUHLERS AND KENNEDY
response to index selection on individual performance. The additional gains expected due to increased selection intensity by keeping fewer boars
was partially offset by losses in genetic variance
due to increased inbreeding. In the present study,
using fewer boars in the herd also resulted in an
additional 5 to 7% reduction in genetic variance,
as a proportion of the original genetic variance,
and a .09 increase in inbreeding level for all three
heritabilities. Using 25- rather than 12-sow herds
had a significant effect on genetic improvement
but little effect on inbreeding level. Belonsky and
Kennedy (1988) used four boars with 100 sows and
found greater selection responses than in the
present study, which used 12 or 25 sows, but levels
of inbreeding differed little from those in the
present study. The 25-SOW herd size gave greater
selection response than 12-sow herds largely be-
cause of greater selection intensities, mostly
through the males, because more animals were
available for selection in a contemporary group.
There was little effect on inbreeding level and only
a small increase in genetic variance.
Interactions between the number of boars used
in the herd and the number of sows in the herd
were evident (Contrast 71 at heritabilities of .3and
.6, but not at .l. The difference in genetic merit
between sow herd sizes of 25 and 12 was greater
when only four boars per year were used compared to the difference between the sow herd sizes
when eight boars per year were used. The reduction in genetic variance (for traits with h2 = .6)
in herds with 25 sows compared to herds with
12 sows mated to four boars was greater than
the difference when eight boars were used in the
herd.
Table 4. Effect of single degree of freedom contrasts
for average genetic merit after 10 years of selection
Heritability
Description of contrast
Selection method (MI
(1) Selected lines - random
(2) BLUP - individual lines
(3) Culling - no culling lines
(4) (2) x (3)
Herd size (SI
(5) 4 boars - 8 boars
(6)25 SOWS - 12 SOWS
(7) ( 5 ) x (6)
.3
.6
.456***
.238***
.251* * *
.Ill***
1.198**
.361***
.638***
.227***
2.250***
.263** *
1.239***
.220** *
.059***
.125***
.004NSa
,170** *
.292***
.032**
.324***
.539***
.083** *
.OlONS
.028**
.062***
.086***
.177***
.OlONS
.173***
.329** *
.050*
.044*
.1
Contemporary group (GI
(8) .1 - .5
M x S interaction
(9) (1) x (51
(10) (1) x (61
(111 (1) x (71
(12) (2) x (5)
(13) (2) x (6)
(14) (2) X (7)
(15) (3) X ( 5 )
(16) (3) x (6)
(171 (3) x (7)
(18)(4) x ( 5 )
(19) (4) x (61
(201 (41 x (71
M x G interaction
(211 (1) x (81
(22) (2) x (81
(231 (3) x (8)
(24) (4) x (81
S x G Interaction
(25) (5) x (8)
(261 (6)x (8)
(27) (7) X (8)
&NS = not significant.
+P < .lo.
*P
< .05.
**P < .01.
***P <
,001.
.025**
.074**
.002NS
.014t
.047** *
-.005NS
.030** *
.007NS
.008NS
.009NS
.015*
.001NS
.06 1* *
.057***
-.OO2NS
.096***
.056***
-.035**
.OS1***
.038**
.O1ONS
.om+
.001NS
-.005NS
.006NS
-.006NS
-.040**
.044***
-.045***
.OO8NS
.014*
.01l t
-.020*
.005NS
-.OOSNS
.056**
.036*
.180***
.065**
-.032+
-.042*
.OO9NS
.043*
.042*
-.082* * *
,073* * *
-.073***
-.OO8NS
-.OOONS
-.002NS
CULLING WITH BLUP AND MASS SELECTION
Contemporary Group Variation. The amount of
variation between contemporary groups did not
have a significant effect on genetic improvement,
variance, and inbreeding for traits with h2 = .1.
However, increasing contemporary group variation from .1 to .5 decreased genetic response after
10 yr by .028 and .062 PSD for heritabilities of .3
and .6, respectively. Most of this difference was
due to the selection method of individual phenotypic selection with culling (IND-CULL), which
gave a difference between the two contemporary
group variation levels of .164 and .306 PSD for
heritabilities of .3 and .6, respectively. This corresponds to a loss of 11 to 13% for contemporary
group standard deviation of .5 compared to .l.
Differences between the two levels of variation for
each of the other selection methods were small
and not significant. Genetic variance increased as
2345
the contemporary group standard deviation increased from .I to .5 only with heritability equal to
.e. Inbreeding coefficients were not influenced
significantly by the change in contemporary group
standard deviation. Responses from selection on
BLUP estimation of breeding values with or
without culling were equivalent for both contemporary group standard deviations, which shows
that BLUP accounts effectively for contemporary
group variation (Henderson, 1973).
Interactions. Method of selection, herd size, and
contemporary group standard deviation interacted for a number of contrasts. Two of the largest
interactions for genetic gain, genetic variance, and
inbreeding level involved selection lines. Contrast
13 involved the interaction of BLUP vs individual
selection x 25 cows vs 12 sows. Average genetic
values after 10 yr of selection indicated that BLUP
Table 5, Effect of single degree of freedom contrasts
for average genetic variance after 10 years of selection
Heritability
DescriDtion of contrast
Selection method (MI
(1) Selected lines - random
(2) BLUP - individual lines
(3) Culling - no culling lines
(4)(21 x (31
1
.3
.6
-.0067***
-.0221* **
-.0193***
-.0168* * *
-.0113***
-.0559***
-.0203** *
-.0337***
-.0260***
-.0077***
.oo1Q*
.0005NSa
-.O 196* * *
-.0325 *
.0105***
.0037*
.0002NS
-.OOO I N S
-.0013*
-.0028** *
-.0013*
-.OO 14**
-.oo 18***
-.0003NS
-.0015**
-.0015NS
-.0086***
-.0050***
-.0055***
Herd size (SI
(5) 4 boars - 8 boars
(6) 25 SOWS - 12 SOWS
(7) (5) x (8)
.0048**
.0018NS
Contemporary group (GI
(8) .1
-
.5
-.0055**
M x S interaction
(a)
(11 x (5)
(10) (1) x (6)
(11) (1) x (7)
(12) (21 x (5)
(13) (2) x (6)
(14) (2) x (7)
(15) (3)x (5)
(10) (31 x (6)
(17) (3) x (7)
(181 (4) x (5)
(191 (4) x (8)
(20) (4) x (7)
M x G interaction
(211 (11 x (8)
(22) (2) x (8)
(23) (3) x (8)
(24) (4) x (8)
S x G interaction
(25) (5) x (8)
(28) (6) x (8)
(27) (7) x (8)
~_________
*NS = not significant.
tP < .lo.
*P < .05.
**P < .01.
***P < ,001.
.0011NS
-.0089**
-.O 173***
-.0072* *
.0055**
-.OO18NS
-.0010*
-.0037t
.0023NS
-.0087* **
,004 1
-.0063*
-.OO2ONS
.0017*
.0011*
-.0012*
.0004NS
-.0060* *
.001QNS
-.0017NS
-.OOOONS
-.0001NS
-.0004NS
.0004NS
.0002NS
.0006NS
.0000NS
.OOOBNS
.0003NS
-.OOO I N S
.000QNS
-.OO11NS
.0011NS
-.002Bt
.0007NS
-.OOOQNS
-.0040*
-.0106***
-.0005NS
-.0017NS
.0033t
-.OO11NS
-.OO28NS
.0062**
-.003Q*
.0040*
-.0004NS
-.0023NS
-.OO l 6 N S
2346
KUHLERS AND KENNEDY
year. This greater genetic improvement was accompanied by a reduction in the genetic variance
and a n increase in the level of inbreeding.
Interactions between selection method and contemporary group standard deviation were associated almost exclusively with culling with individual phenotypic selection on traits with
heritabilities of .3 and .6. Genetic improvement
was less effective at these heritabilities with
individual phenotypic selection with culling when
the contemporary group standard deviation was
increased. None of the other selection methods
was affected by the increase in contemporary
group standard deviation. At heritability of .l, the
inaccuracy of individual phenotypic selection was
not significantly affected by a n increase in the
contemporary group standard deviation. Interactions between herd size and contemporary group
methodology was able to take advantage of the
larger amount of information on relatives for
estimating breeding values generated in herds of
25 sows compared with herds with 12 sows. With
individual phenotypic selection, which does not
take into consideration family information, additional animals produced in the herd do not
contribute to the accuracy of the selection and,
therefore, the response to selection. A similar
result was found by Rohe et al. (19901, who showed
that the animal model gave relatively higher
accuracies of selection than did index selection in
herds of 180 sows compared with herds with 120
sows.
Contrast 15 involved the interaction of culling
vs no culling x 4 vs 8 boars. The effect of culling
compared with not culling is greater when four
rather than eight boars are used in the herd each
Table 6. Effect of single degree of freedom contrasts
for inbreeding coefficient after 10 years of selection
Heritability
Description of contrast
.1
.3
.8
.04 14***
.0719***
.0560***
,0481***
.0408***
.0545***
.0578**
.0351***
.0425***
.0312***
.0477***
.o 182** *
.0941***
.0027NSa
.0025NS
.0922***
.0009NS
.0017NS
.0004NS
.0004NS
.0008NS
.0095** *
.0083**
.0054**
-.OO12NS
.0111***
.0120***
.OO 14NS
.0203***
.0068***
.0040** *
-.0013NS
.0023*
.0071***
.0046***
.0094***
-.0022*
-.0087* * *
-.0048***
.0000NS
-.0044***
~~~~~~~~~~~
Selection method (M)
(1) Selected lines - random
(2) BLUP - individual lines
(31 Culling - no culling lines
(41 (2) x (3)
Herd size (SI
(5) 4 boars
(8) 25 SOWS
(7) (51 x ( 8 )
- 8 boars
- 12 SOWS
.0898** *
-.OO 14'
.0016t
Contemporary group (GI
(8) .1 - .5
M x S interaction
(91 (1) x (5)
(10) (1) x (8)
(11) (1) x (7)
(121 (21 x (5)
(13) (21 x (61
(14) (21 x (7)
(151 (3) x (51
(16) (31 x (81
(17) (31 X (71
(18) (4) x (51
(19) (4) x (131
(201 (4) x I71
M x G interaction
(211 (11 x (8)
(22) (21 x (8)
(23) (3) X (8)
(241 (4) x (81
.0088*
-.0007NS
.O173**
.0145***
.0038NS
.0195***
.0038NS
-.0044t
.0086**
.0078*
.0017NS
.0052**
-.0027t
.0078* * *
.0070***
.0032*
.0008NS
.0000NS
.0008NS
.OO14NS
-.0015NS
-.0013NS
.0008NS
-.OOO8NS
.0008NS
-.0001NS
.0001NS
.0004NS
.0022NS
.0023NS
-.0003NS
-.0015NS
-.0003NS
-.0008NS
-.0008NS
-.0014t
-.OO 13NS
S x G interaction
(25) (5) x (8)
(281 (81 x (8)
(271 (71 x (8)
&NS = not significant.
+P < .lo.
*P < .os.
**P < .01.
***P <
,001.
CULLING WITH BLUP AND MASS SELECTION
standard deviation were, in general, not important
sources of variation.
Discussion. This work shows that selection on
BLUP and culling on estimated breeding values
when a n expected superior replacement animal is
available were more effective than individual
phenotypic selection and not culling on estimated
breeding values in small (12 or 25 sows), continuously farrowing, closed herds. Similar results were
shown by Mabry and See (19901 for 60-sow herds,
by Belonsky and Kennedy (19881for 100-sowherds,
and by Rohe et al. (1990) for 120- and
180-sowherds. At the same time, inbreeding levels
increased to -27 to .35 after 10 yr using four boars
per year when there was voluntary culling with
BLUP selection. Using eight boars per year and
culling with BLUP selection increased inbreeding
levels to .17 to .20. Inbreeding levels were about
twice as large as with randomly selected lines of
equal sizes. Comparable inbreeding coefficients
for individual phenotypic selection were .18 to .23
and .09 to .12 for four and eight boars, respectively.
If traits such as litter size and survival rate,
although not selected for directly, are negatively
affected by the increases in inbreeding level, then
the number of progeny produced and the selection
intensities may be reduced for those selection
methods that increase inbreeding to a greater
extent, leading to reduced response. This has been
shown to be the case by DeRoo (1988a1 when
correction for inbreeding depression was not
made. This outcome, therefore, may lead to less
superiority of BLUP and culling on estimated
breeding value over individual phenotypic selection and not culling than the present results would
suggest. Similarly, the conflict between increased
response associated with using fewer boars in
these small herds and the dramatic increase in
inbreeding needs to be considered by breeders
with small numbers of sows in a continuous
farrowing system. If the breeder plans to keep the
herd closed for more than 10 yr, alternative
strategies may be needed. Avoidance of inbreeding, or using more boars per year and deferring
some short-term genetic merit to obtain longerterm gain are possible options. Another option is
to include the fitness traits that are affected by
inbreeding and the effects of inbreeding depression in the selection criterion. DeRoo (1988~1
showed that avoidance of mating of relatives
resulted in lower responses than when mating of
relatives was not avoided over 25 yr of index
selection in simulated herds. Which of these is
superior in the long term is presently not known.
An alternative to maintaining the herd closed is to
open the herd to unrelated animals, which brings
the inbreeding level back to zero but at a cost of
losing some of the genetic response acquired if
2347
information on the traits of interest is not available on the outside animals.
Additional work needs to be conducted on the
effectiveness of BLUP superiority in small, closed
herds beyond the 10-yr period of time. Demfle
(1990) indicated, in simulated populations of five
full-sib families of eight males and eight females,
that a selection index that optimally weighted
family and individual performance, although superior to individual phenotypic selection for 5 to 10
generations, was not best a t the end of 20 and 40
generations. DeRoo (1988bl concluded that as the
time period included in the evaluation of a
breeding scheme increased, the optimal number of
boars used per year also needed to increase.
Decreasing the maximum ages of boars and
sows in the herd may also influence the differences among the selection methods and herd
sizes. Bichard et al. (1986)has indicated that boars
should be kept for no more than 6 mo and sows for
no more than two litters to optimize selection
response per unit of time. The present study
attempted to depict the existing purebred industry
and not necessarily optimize the results, except for
BLUP-CULL,which does maximize the response to
selection, a t least over the short term. The results
of Belonsky and Kennedy (1988) and those of this
study showed that the effect of culling on estimated breeding values was to shorten the generation interval compared to the interval that would
have occurred from not culling on estimated
breeding value. Therefore, a breeding policy reducing the length of breeding life for boars and
sows as given in the present study probably would
reduce the magnitude, but not the ranking, of the
selection and culling methods studied.
Batch farrowing (i.e., farrowing all 12 or 25 sows
in one 28-d period twice per year) would give
greater numbers of progeny from which to select
and might also alter the size of the differences
between the selection methods and herd sizes.
One of the strengths of BLUP evaluation is that it
effectively ties together small groups of related
animals. With batch farrowing, fewer groups
would need to be tied together, thereby reducing
some of the advantage of BLUP. However, examination of Ontario Swine Improvement Program
Provincial data shows that more than one-half the
Duroc and Hampshire purebred breeders tested
three or fewer litters per contemporary group,
probably to meet the needs of their clientele.
Finally, interactions of selection method with
and without culling and with different herd sizes
were evident in this work. This suggests that
rankings of and magnitude of differences between
selection methods may be influenced by both herd
sizes and structure. Each herd size and structure
should be evaluated on a n individual basis to
KUHLERS AND KENNEDY
2348
ascertain the relative effectiveness of different
selection procedures given a policy decision on
length of life of breeding animals.
Implications
Computer simulations of closed sow herds
showed that 25-sow herds, best linear unbiased
prediction of breeding values, and culling when an
expected superior replacement animal was available gave greater selection response than
12-sow herds, phenotypic selection, and not culling
when a superior replacement was available. Because response is greater in the larger sow herds,
exchange of breeding stock among small breeders
and selection of breeding stock on best linear
unbiased prediction of breeding values allows
small breeders to have effective population sizes
and a powerful evaluation system similar to those
of larger purebred breeders and breeding companies.
Literature Cited
Belonsky, G. M., and B. W. Kennedy. 1988. Selection on individual phenotype and best linear unbiased predictor of breeding value in a closed swine herd. J. Anim. Sci. 66:1124.
Bichard, M., P. J. David, and M. Bovey. 1986. Selection between
and within lines and crossbreeding strategies for worldwide production of hybrids. Proc. 3rd World Congr. Genet.
Appl. Livest. Prod. 10:130.
Bulmer, M. G. 1980. The Mathematical Theory of Quantitative
Genetics. Clarendon Press, Oxford, U.K.
Demfle, L. 1990. Conservation, creation and utilization of genetic variance. J. Dairy Sci. 73:2593.
DeRoo, G. 1988a. Studies on breeding schemes in a closed pig
population. Doctoral Thesis. Dept. of Anim. Breeding, Agricultural University, Wageningen, The Netherlands.
DeRoo, G. 1988b. Studies on breeding schemes in a closed pig
population. 1. Population size and selection intensities.
Livest. Prod. Sci. 19:417.
DeRoo, G. 1988~.Studies on breeding schemes in a closed pig
population. 2. Mating policy. Livest. Prod. Sci. 19:443.
Henderson, C. R. 1973. Sire evaluation and genetic trends. In:
Proc. of Anim. Breed. Genet. Symp. in Honor of Dr. J. L.
Lush. p 10. ASAS and ADSA, Champaign, IL.
Hudson, G.F.S., and B. W. Kennedy. 1985. Genetic evaluation of
swine for growth rate and backfat thickness. J. Anim. Sci.
61:83.
IMSL. 1981. International Mathematical and Statistical
Libraries, Inc. Edition 8 , Houston, TX.
Kennedy, B. W., G.F.S. Hudson, and L. R. Schaeffer. 1988. Evaluation of genetic change in performance tested pigs in Canada. Proc. 3rd World Congr. Genet. Appl. Livest. Prod. 10:
149.
Mabry, J. W., and M. T. See. Selection with the animal model
versus selection within contemporary groups for swine. J.
Dairy Sci 732857.
Quaas, R. L., and E. J. Pollak. 1980. Mixed model methodology
for farm and ranch beef cattle testing programs. J. Anim.
sci. 51:1277.
Quinton, M., C. Smith, and M. E. Goddard. 1992. Comparison of
selection methods a t the same level of inbreeding. J. Anim.
Sci. 70:1060.
Rohe, R., J. Drieter, and E. Kalm. 1990. Efficiency of selection in
closed nucleus herds of pigs using a n animal model. Proc.
4th World Congr. Genet. Appl. Livest. Prod. 15:469.
SAS. 1989. SAS/STAT@ User's Guide (Release 6.03).SAS Inst.
Inc., Cary, NC.