Matt Culbertson
Cotswold USA
University of Georgia
INTRODUCTION
Genetic evaluation for livestock species has improved dramatically
in recent years with the widespread implementation of Best Linear
Unbiased Prediction (BLUP) of genetic merit for many traits of
economic importance. Due to the corresponding developments in
computing technology it has become relatively common to performance
test individual animals and subsequently, combine all available
information, i.e. from the animal, ancestors, collateral relatives,
and progeny, into a single value to represent genetic merit for
a given trait. However, although these methods are a marked improvement
over previous phenotypic selection schemes, the current methods
are only concerned with the additive component of genetic variation.
Therefore, if non-additive genetic effects are an important part
of variation, the accuracy of evaluations, even for "high-accuracy"
animals, is reduced.
Accounting for dominance genetic effects may contribute significantly
to accurate evaluation of performance records. However, dominance
effects have been ignored because evaluation methods were computationally
unfeasible for larger populations and the importance of dominance
effects have not been well established. Recent developments have
made computations with dominance effects much more manageable.
Hoeschele and VanRaden (1991) discovered rules for the inversion
of the dominance relationship matrix. Further, refinements in
computing have enabled evaluation with dominance in the model
to require less than twice the memory and computing time as a
model which includes only additive effects (Misztal, 1996). Finally,
use of Method R (Reverter et al., 1994) has enabled estimation
of dominance variance from large data sets.
The importance of dominance effects may prove larger in litter
bearing species that have added economic importance, and corresponding
selection programs, based on prolificacy. Falconer (1989) suggests
that fitness traits, such as measures of reproductive performance,
where the population mean is below the optimal value for fitness,
may have large dominance effects. Traditionally, the genetic
cause of crossbreeding and inbreeding has been thought to be dominance
(Falconer, 1989; Hill, 1982). Various studies have shown the
phenotypic results of crossbreeding and inbreeding, or heterosis
and inbreeding depression, to be greatest in those traits with
lower additive genetic variation, or typically reproductive traits
(Warwick and Legates, 1979). Finally, Hoeschele (1991) showed
that significant dominance effects appeared to exist in dairy
cattle populations analyzed for measures of fertility.
Assuming that dominance variation does exist, accuracy of genetic
evaluation for additive genetic merit will be improved by the
inclusion of these effects into the prediction model (Uimari,
1990). In addition, the genetic merit of parental combinations
(specific combining ability) can be estimated to enhance the productivity
of purebred matings for fertility (DeStefano and Hoeschele, 1992).
This, in turn, should improve the profitability of sections of
a genetic pyramid which must utilize pure-line matings.
As mentioned previously, inbreeding also poses a problem in the
genetic evaluation of animals for traits which are affected by
inbreeding depression. For traits and populations affected by
inbreeding, the genetic evaluation methods currently used are
inaccurate. However, if inbreeding is included in the model,
the genetic evaluation of inbred animals will be inflated if inbreeding
is not included in mating decisions (Uimari, 1990). Theoretically,
if mating pairs (dominance) and inbreeding are both included in
the model the resulting information should be accurate for purebreds
and crossbreds (Misztal et al., 1995). Therefore, in smaller
nucleus populations attempting to make rapid genetic improvement
and minimize inbreeding, the inclusion of dominance and inbreeding
in the estimation model may offer significant improvement over
current methods for evaluation of reproductive traits.
ESTIMATION PROCEDURES FOR DOMINANCE
EFFECTS
179,485 reproductive records and 239,354 growth records from purebred
American Yorkshire swine were obtained from the National Swine
Registry. Data sets were initially edited to insure connectedness
and eliminate biological extremes and were adjusted to a constant
basis (NSIF, 1987; Culbertson, 1997). NBA and LWT were analyzed
separately with a model which included the fixed effects of contemporary
group and regression on inbreeding percentage and the random effects
of additive genetic, parental dominance, animal permanent environment,
and mate within contemporary group. DAYS and BF were analyzed
separately with a model which included the fixed effects of contemporary
group, sex, and regression on inbreeding and the random effects
of additive genetic, dominance genetic, litter of birth, and
maternal permanent environment. Estimates were obtained by Method
R (Reverter et al., 1994) following the procedures of Misztal
(1997). Dominance variance was estimated as four times the parental
dominance estimate. Each trait was analyzed with 6 samples of
the data selected by a random number generator. The convergence
criterion was ri = 1 ± .0001, where ri
is the regression from the random effect i. Sampling standard
deviation was defined as the standard deviation of the 6 estimates
from the subsamples.
After estimating the relative variances, additive breeding values
were predicted for all animals using single-trait procedures and
three separate prediction models differing in the inclusion of
inbreeding and dominance genetic effects. The first model contained
only additive genetic effects, the second model contained additive
genetic and regression on inbreeding, and the third model contained
additive genetic, regression on inbreeding, and dominance genetic.
Additive breeding values were compared between alternative models
and trends and differences in specific groups of animals were
analyzed.
RESULTS FOR ESTIMATION STUDY OF
DOMINANCE IN SWINE
Estimates of inbreeding depression and variances for NBA and LWT
are presented in Table 1. All variances are expressed as a percentage
of phenotypic variation. Estimates of inbreeding depression were
found to be sizeable for both traits and similar to those found
by Bereskin et al. (1968). For example, a sow with an own inbreeding
level of F=.125 would have her record adjusted by approximately
.3 of a pig for NBA. Dominance variance was found to be larger
for LWT, 78% of additive variance, relative to NBA, 25% of additive
variance.
Table 1. Means (standard deviations) of estimates of inbreeding depression and additive, dominance, permanent environment (PE), and mate within contemporary group (mate) variances for NBA and LWT | |||||
|
| |||||
1 Per 10% of inbreeding
Estimates of inbreeding depression and variances for DAYS and BF are presented in Table 2. Inbreeding effects were found to be significant for DAYS and negligible for BF. Dominance variance for DAYS was found to be approximately 10% of phenotypic variation and a third of additive genetic variance.
Table 2. Means (standard deviations) of estimates of inbreeding depression and additive, dominance, litter, and maternal permanent environmental variances (PE) for days to 104.5 kg (DAYS) and backfat at 104.5 kg (BF) | |||||
|
| |||||
1 Per 10% of inbreeding
Additive breeding values were obtained after fitting of the three
alternative prediction models. For all 4 traits, the correlation
between additive breeding values after fitting of the three models
were > .99. This would suggest, on average, that accounting
for inbreeding and dominance effects has little effect on prediction
of additive genetic merit. However, by identifying groups of
animals which are the most affected it may be possible to determine
types of populations where dominance models are most beneficial.
Table 3 presents changes in sires' additive breeding for LWT
due to inclusion of regression of inbreeding with classes based
upon number and average inbreeding of daughters. Table 4 shows
changes in additive breeding value for inclusion of dominance
genetic effects with classes based upon the number of full-sibs,
i.e. the amount of available dominance information. Changes
for individual animals were noticeable for all traits except BF.
Inbreeding caused greatest change in prediction of additive merit
for inbred animals and those with inbred progeny. Changes in
prediction due to dominance genetic effects were generally smaller
than those due to inbreeding effects and were found to increase
as the amount of dominance information available for a given family
increases.
Table 3. Change in sire additive breeding value for LWT due to inclusion of inbreeding in the genetic prediction model with classes based on average inbreeding of daughters (Number of sires in parentheses) | ||||||||
|
|
||||||||
Table 4. Change in additive breeding value for LWT due to inclusion of dominance genetic effects in the genetic prediction model (number of animals in parenthesis) | ||||||||
In addition to improving prediction of additive genetic merit,
utilization of a more complete prediction model with inbreeding
and dominance genetic effects may allow for increased phenotypic
performance through a mate allocation program. A small simulation
study determined that utilizing conservative assumptions of 1
sire to 15 dams, no parental combinations previously tested, and
an average off-diagonal of the parental dominance relationship
matrix of .05, the expectation for increase would be approximately
5% of the parental dominance standard deviation. However, if
20% of the parental combinations are previously tested the increase
then becomes approximately 35% of the parental dominance standard
deviation. Utilizing these results and economic values of $17
per additional pig, $1.10 per kilogram of litter weight, $.17
per day to 104.5 kg and $.60 per millimeter of BF, increased revenue
may be up to $1.24 per litter for NBA, $0.69 per litter LWT, $0.11
per pig for DAYS, and $0.09 per pig for BF.
SUMMARY
The inclusion of dominance genetic and inbreeding effects in a
genetic prediction model may prove more beneficial for smaller,
closed populations of swine similar to those maintained by most
commercial breeding organizations. In addition, future research
should be conducted to look at mating programs and genetic system
designs which may be able to maximize the potential return of
a mating scheme which incorporates these effects.
REFERENCES
Bereskin, B., Shelby, C.E., Rowe, K.E., Urban, W.E., Jr., Blunn,
C.T., Chapman, A.B., Garwood, V.A., Hazel, L.N., Lasley, J.G.,
Magee, W.T., McCarty, J.W., and Whatley, J.A., Jr. (1968) J.
Anim. Sci. 27:339-350.
Culbertson, M.S., Mabry, J.W., Bertrand, J.K., and Nelson, A.H.
(1997) J. Anim. Sci. accepted.
DeStefano and Hoeschele. (1992) J. Dairy Sci. 75:1680-1690.
Falconer (1989) Introduction to quantitative genetics.
Hill (1982) Z. Tierz. Zuechtungsbiol. 99:161-168.
Henderson, C.R. (1985) J. Anim. Sci. 60:111-117.
Henderson, C.R. (1989) J. Anim. Sci. 72: 2592-2605.
Hoeschele (1991) J. Dairy Sci. 74:1743-1752.
Misztal, I., Fernando, R.L., Grossman, M., Lawlor, T.J., Lukaszewicz,
M. (1995) Animal Science Papers and Reports 13:251-265.
Misztal, I. (1997) J. Dairy Sci. 80:965-974.
NSIF (1987) "Guidelines for uniform swine improvement programs."
National Pork Producers Council, Des Moines.
Reverter, A., Golden, B.L., Bourdon, R.M., and Brinks, J.S. (1994)
J. Anim. Sci. 72:2247-2253.