| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Department of Animal Science and
and
Department of Statistics, Iowa State University, Ames 50011
2 Correspondence:
109 Kildee Hall, Iowa State University, Ames, IA 50011 (phone: 515-294-6728; fax: 515-294-5698; E-mail:
tjbaas{at}iastate.edu).
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Key Words: Genetic Parameters • Genetic Trend • Growth • Pigs
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Data were obtained from the National Swine Registry on Yorkshire, Duroc, Hampshire, and Landrace pigs born from 1985 to April of 2000. Details of data collection can be found in STAGES (NSR, 2000). The data included pedigree information for each pig, contemporary group (CG), sex of the pig, litter identification, birth date, date weighed, and measurements for weight, BF, and LEA at an approximate weight of 113.5 kg. Contemporary groups were defined by breeders as a group of pigs that were simultaneously raised in a common herd under similar environmental conditions. Data on boars, gilts, and barrows were included in the data set. Backfat and loin eye area were measured ultrasonically at the 10th rib. Days to 113.5 kg, backfat, and loin eye area were adjusted to 113.5 kg using recommendations in the Guidelines for Uniform Swine Improvement Programs (NSIF, 1997). Lean growth rate adjusted to 113.5 kg was predicted using the following fat-free lean prediction equation developed by the National Pork Producers Council (NPPC, 2000): LGR, kg/d = [0.3782 x sex (barrow and boar = 1; gilt = 2) - 2.9488 x (BF, cm) + 0.3817 x (LEA, cm2) + 0.291 x (adjusted live weight, kg) - 0.2424]/(days on test), where days on test is defined as the time between birth and weigh date. Single-sire contemporary group records were removed, as were records from sires not connected across contemporary groups and sires not mated to more than one dam. Numbers of records, contemporary groups, and litters represented by breed are shown in Table 1
, along with means and SD for LGR, DAYS, BF, and LEA. Distributions across years are shown in Table 2 for each breed.
|
|
Univariate Analyses.
Univariate analyses within breed were conducted using the REMLF90 program of Misztal (2000) to estimate single-trait variance components. The following animal model was fitted using REML: 
, where y represents the vector of observations; b is the vector of fixed effects of CG and sex; a is the vector of random additive genetic effects of animals, which is assumed to be distributed across N(0, A
), where A is the numerator relationship matrix among animals; c is the vector of common litter effects, assumed to be distributed across N(0, I
) and uncorrelated with random animal effects; m is the vector of random additive maternal genetic effects, assumed to be distributed across N(0, A
) and correlated with random animal effects (
am); and e is the vector of residual effects, which is assumed to be distributed across N(0, I
). The terms X, Z, S, and W are incidence matrices relating records to fixed, additive genetic, common litter, and maternal genetic effects, respectively. A simplified model was fitted also, in which maternal genetic effects were excluded. Standard errors of heritability estimates were estimated using the approximate method of Swiger et al. (1964).
Bivariate Analyses.
Bivariate REML analyses were conducted to estimate genetic and phenotypic correlations between traits, again using the programs of Misztal (2000). The model for the bivariate analyses was the same as for the univariate analyses, except maternal genetic effects were excluded. Standard errors of genetic correlation estimates (
A) from REMLF90 were not available; however, they can be approximated by the method of Falconer (1989), demonstrated by several studies (Lo et al., 1992; Chen et al., 2001):
 |
where 
2x and 2y are the heritability estimates of traits x and y, respectively, and 2x and 2y are the standard errors of heritability estimates of traits x and y, respectively. In the REML analyses, the convergence criterion was set to 10-8 for all analyses.
Maternal Effects. Breeding values for LGR and its components were estimated under two single-trait models both with and without maternal effects, using the programs of BLUPF90 of Misztal (2000). Spearman rank correlations between two sets of EBV were estimated to examine consequences of ignoring maternal effects.
Genetic Trends. Breeding values for LGR and its components were estimated under a multi-trait model, without maternal effects, using the programs of BLUPF90 of Misztal (2000). The average breeding values per year of birth of the pigs were regressed to year of birth of the pigs, and plotted to illustrate trends.
|
Results and Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Estimates of animal genetic, maternal genetic, common litter, and residual variances and covariances between animal genetic and maternal genetic effects are shown in Tables 3 to 6. Genetic variances in LGR, DAYS, and LEA were highest in Hampshires and lowest in Durocs. The Landrace had the highest and the Hampshires had the lowest genetic variance in BF. Total variation in LGR and DAYS was lower in Durocs than in the other three breeds. The Yorkshire and Landrace breeds had greater total variation in BF, while total variation in LEA was the lowest in Durocs and highest in Landrace. The Hampshire and Landrace breeds had the highest litter variance for LGR, while the Landrace breed had the highest litter variance for BF and LEA.
|
|
|
|
Results of analyses using single-trait models for LGR, DAYS, BF, and LEA are given in Tables 3 to 6
. Average estimates of heritabilities for LGR, DAYS, BF, and LEA from bivariate analyses are given in Table 7. Five estimates of heritability for each trait within a breed were obtained by single- and bivariate-trait analyses. Heritability estimates did not differ by more than 2% for any of these analyses; therefore, results of the single-trait analyses are discussed for simplicity. Estimates of heritability for LGR were 0.44, 0.44, 0.46, and 0.39 for the Yorkshire, Duroc, Hampshire, and Landrace breeds, respectively (Table 3). These estimates were in the range of previous estimates. Stern et al. (1993) reported an estimate of heritability of 0.37 for LGR between 25 to 90 kg in a selection experiment conducted in Swedish Yorkshire pigs free of the stress form of the halothane gene (NN). Mrode and Kennedy (1993) used Yorkshire, Landrace, and Duroc records from Canadian test stations and reported a heritability estimate of 0.39 for LGR between 29 to 90 kg. Cameron (1994) reported an estimate of heritability of 0.38 for LGR between 30 to 85 kg in a selection experiment in Large White pigs. Cameron and Curran (1994) also found an estimate of heritability of 0.25 in a selection experiment in Landrace pigs. Recently, Chen et al. (2001) reported a heritability estimate of 0.37 for LGR obtained from birth to 113.5 kg in a selection experiment in a synthetic line of Yorkshire-Meishan pigs. Estimates may depend on population structure, selection criterion, breed, sampling error, and appropriateness of the LGR equation for different breeds.
|
). These estimates are consistent with average literature estimates of 0.47 by Hutchens and Hintz (1981). Ducos et al. (1993) reported an estimate of 0.32 for days to 100 kg. Li and Kennedy (1994) also reported estimates of heritability for days to 100 kg ranging from 0.26 to 0.32 based on Yorkshire, Landrace, Duroc, and Hampshire data from the Canadian Swine Improvement Program.
Estimates of heritability for BF in this study were 0.48, 0.48, 0.49, and 0.48 for the Yorkshire, Duroc, Hampshire, and Landrace breeds, respectively (Table 5). These estimates were in the range of estimates from previous studies. Kennedy et al. (1985) reported heritabilities for ultrasonic measure of BF at 90 kg ranging from 0.40 to 0.44 for different breeds of Canadian performance tested pigs. Lo et al. (1992) found an estimate of 0.54 for ultrasonic measure of BF. Bryner et al. (1992) used Yorkshire records from U.S. Central test stations and reported a heritability estimate of 0.56 for ultrasonic measure of BF. Ferraz and Johnson (1993), using four animal models for herds of Landrace and Large White pigs, reported estimates of heritability that ranged from 0.39 to 0.50 for ultrasonic measure of BF. Mrode and Kennedy (1993) reported an average heritability of 0.59 for three breeds for ultrasonic measure of BF adjusted to 100 kg. Johnson et al. (1999) reported a heritability estimate of 0.36 for ultrasonic measure of BF in Large White boars.
Estimates of heritability for LEA were 0.33, 0.32, 0.35, and 0.31 for Yorkshires, Durocs, Hampshires, and Landrace, respectively (Table 6). These estimates were lower than previous estimates of 0.46 by Lo et al. (1992). Some previous studies did not properly account for common litter effects, which may be substantial, as found in this study (Tables 3 to 6); this could have contributed to an upward bias in previous estimates. Johnson et al. (1999) reported an estimate of heritability of 0.24 for LEA, with a common litter effect of 0.18.
The magnitude of the estimated heritability for LGR was in the range of its components for each breed. It was expected that estimates of heritability for LGR would be a compromise between heritability estimates of the three traits, since LGR is a combination of its components. Standard errors of heritability estimates from REMLF90 were not available; however, they can be approximated by the method of Swiger et al. (1964), as demonstrated by several studies (Lo et al., 1992; See, 1994). Standard errors of heritability estimates ranged from 0.01 to 0.02.
Common Litter Effects
Common litter effects were significant sources of variation for all traits and breeds, and the proportion of phenotypic variation accounted for by common litter effects ranged from 0.07 to 0.16 (Tables 3 to 6). Estimates of common litter effects were not greatly affected by including random maternal genetic effects in the model for any trait or breed. Also, estimates of common litter effects were similar for all traits and all breeds except a relatively large estimate of 0.16 for DAYS in Durocs, and a small estimate of 0.07 for LEA in Hampshires. Ferraz and Johnson (1993) reported that approximately 7% of the variation in ADG and 5% of the variation in BF was due to common litter effects in Landrace and Large White pigs. Li and Kennedy (1994) reported average common litter effects, expressed as a proportion of the total variance, of 0.26 for days to 100 kg and 0.10 for BF in Yorkshire, Landrace, Duroc, and Hampshire pigs. Crump et al. (1997) reported average estimates over various analyses of 0.05, 0.11, 0.06, and 0.06 for BF, average daily food intake, average daily gain, and food conversion ratio, respectively. Johnson et al. (1999) reported common litter effects of 0.13 and 0.18 for BF and LEA in Large White boars.
The magnitude of common litter effects reported in this and in previous studies indicates that common litter effects must be included in the model for estimation of variance components and breeding values. Several studies (Crump et al., 1997) reported that estimates of heritabilities with common litter effects in the model were approximately 10% less than those obtained with only random animal effects in the model. The use of models without common litter effects for genetic evaluation would, therefore, lead to biases in estimation of breeding values and overprediction of potential genetic gain.
Additive Maternal Genetic Effects
Additive maternal genetic effects were not large sources of variation in this study. Fractions of variance accounted for by maternal effects ranged from 0.01 to 0.05 for all traits across breeds (Tables 3 to 6). Correlations between maternal and direct genetic effects were negative for all traits and for all four breeds. Estimated correlations ranged from -0.31 to -0.61 for LGR, from -0.28 to -0.58 for DAYS, from -0.35 to -0.60 for BF, and from -0.28 to -0.41 for LEA. Crump et al. (1997) reported an average correlation of -0.18 for ultrasonic backfat depth. Ferraz and Johnson (1993) reported estimated correlations of -0.26 for BF and -0.34 for ADG. The estimates in this study were of greater magnitude than estimates reported in previous studies.
The correlations between two sets of EBV for direct genetic effects of LGR and its components, under the models with and without maternal genetic effects, over 15 yr ranged from 0.92 to 0.96. The correlations within year ranged from 0.88 to 0.97. Therefore, maternal effects could be ignored in the model to estimate breeding values in practice. This result agrees with the findings of Ferraz and Johnson (1993) and Crump et al. (1997), which reported that, for all practical purposes, the maternal effect and the correlation between maternal effects and direct effects could be ignored for performance traits in pigs. Robison (1972), however, indicated that maternal effects account for a significant proportion of the variance for 140-d weight and carcass BF in Yorkshire pigs. Bryner et al. (1992) also reported that maternal effects were significant for both BF and ADG, accounting for 11 and 23% of the variance, respectively. These relatively large estimates of maternal effects could be due to confounding between common litter effects and maternal effects, since common litter effects were not included in those studies (Robison, 1972; Bryner et al., 1992).
Genetic Correlations
Genetic and phenotypic correlations among LGR and its components are given in Table 7. Lean growth rate was estimated to have high negative genetic correlations with DAYS of -0.84, -0.86, -0.80, and -0.83 for the Yorkshire, Duroc, Hampshire, and Landrace breeds, respectively, and these estimates did not significantly differ from one another (P > 0.05). Lean growth rate was estimated to have moderately favorable genetic correlations with BF of -0.32, -0.40, -0.35, and -0.41 for Yorkshires, Durocs, Hampshires, and Landrace, respectively. Lean growth rate was estimated to have moderately favorable genetic correlations with LEA of 0.44, 0.43, 0.50, and 0.38 for Yorkshires, Durocs, Hampshires, and Landraces, respectively. Moderate genetic correlation estimates between LGR and LEA (Table 7) indicate that selection for increased LGR would result in increased LEA. Genetic correlation estimates of DAYS with BF and LEA did not differ from zero (Table 7). Kennedy et al. (1985) also reported genetic correlations between days to 100 kg and BF of -0.11 for Durocs and -0.17 for Yorkshires.
Moderately unfavorable genetic correlation estimates between BF and LEA, ranging from -0.35 to -0.45, were found for all breeds (Table 7). This finding is confirmed by Johnson et al. (1999). The estimated genetic correlations between LGR and its components suggests that selection in pigs based on LGR can be accomplished without adverse effects on DAYS, BF, and LEA.
There are small differences in genetic parameter estimates between breeds. Use of estimates pooled across breeds might be appropriate.
Genetic Trends
The average breeding values for animals in 1990 were adjusted to zero to illustrate genetic changes in later generations. All estimated genetic trends for DAYS, BF, LEA, and LGR were favorable (Figures 1 to 4, and Table 8). The Duroc breed showed the largest genetic changes of 3.28 g/d per year and -0.54 d/yr for LGR and DAYS, while Yorkshires had the largest genetic changes of -0.45 mm/yr and 0.41 cm2/yr for BF and LEA, respectively. Average genetic change in DAYS over four breeds was -0.40 d/yr, or -0.23% of the mean. This estimate was lower than the estimate of -0.63, or -0.36% of the mean, for four Canadian breeds reported by Kennedy et al. (1996). The average genetic change in BF over the four breeds was -0.40 mm/yr, or -2.3% of the mean. The average estimate of genetic change expressed as the percentage of the mean was higher than the estimate of 2.0% reported by Kennedy et al. (1996) in Canada and of 0.5% reported by Tibau i Font et al. (1994) in Spain, but lower than the estimate of 3.9% reported by Mantysaari et al. (1994) in Finland. Kennedy et al. (1996) suggests that 1.7 to 2.9% improvement per year for BF and 1.5% per year for days to 110 kg would be realistic for industry breeding programs, although Smith (1984) suggests that 3 to 5% and 2.7% for BF and growth rate is possible.
|
|
|
|
|
Lean Growth Rate
If LGR could be predicted on individual animals, as can other traits, the breeding objective would be a single-trait biological selection criterion (BLUP EBV). Therefore, use of LGR could simplify selection procedures for breeders when selecting for leanness and growth simultaneously.
Estimated breeding values for LGR in this study were estimated from a phenotypic trait that was calculated based on the lean prediction equation (NPPC, 2000). Therefore, lean growth rate used in this study was the predicted LGR. Bennett (1992) explained that the relationships between predicted LGR and other traits could be different from the relationships of actual LGR to the same traits. The accuracy of predicted LGR would rely on maximizing the correlation between actual LGR and a prediction equation consisting of its component traits. The equation used in this study might predict biased LGR due to sex, breed, and weight. Several researchers have demonstrated biases in estimations of carcass lean of swine associated with genotype, sex, and treatments (Gu et al., 1992; Hicks et al., 1998). Further research is needed to develop a sex-breed-weight-specific prediction equation, which can predict LGR more accurately.
An alternative method is to estimate EBVs for the components of LGR, DAYS, BF, and LEA, and then combine these EBVs to obtain the EBV for LGR. Goddard (1998), however, suggests that if there are complex relationships between the traits in the phenotypic profit function, calculating profit on individual animals directly may be more robust than using a complex bioeconomic model. Similarly, since the relationships between the traits of DAYS, BF, and LEA and LGR are complex, it might be better to use the trait of LGR than to use an index that consists of its components as a biological selection criterion. The results of a simulation study (our unpublished observations) show that direct selection of EBV for LGR with a multi-trait model yielded higher LGR when compared to a linear index of EBVs for DAYS, BF, and LEA.
Meuwissen and Goddard (1997) found that when profit was recorded as a trait in a multi-trait BLUP, the EBV for profit provided a robust and accurate selection criterion. The study further suggests that even if profit cannot be calculated for all individual animals, it may still be possible and useful to define traits (e.g., milk per day of herd life in dairy cattle) that are major components of profit. Since LGR expressed as lean gain per day is the major component of profit in the swine industry, and can be predicted and included as a trait in a multi-trait BLUP, the EBV for LGR could be a useful biological selection criterion when economic values are uncertain.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
Received for publication July 9, 2001. Accepted for publication February 27, 2002.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  T. D. Leeds, M. R. Mousel, D. R. Notter, H. N. Zerby, C. A. Moffet, and G. S. Lewis B-mode, real-time ultrasound for estimating carcass measures in live sheep: Accuracy of ultrasound measures and their relationships with carcass yield and value J Anim Sci, November 1, 2008; 86(11): 3203 - 3214. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. R. Schwab, T. J. Baas, K. J. Stalder, and J. W. Mabry Deposition rates and accretion patterns of intramuscular fat, loin muscle area, and backfat of Duroc pigs sired by boars from two time periods J Anim Sci, June 1, 2007; 85(6): 1540 - 1546. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. R. Schwab, T. J. Baas, K. J. Stalder, and J. W. Mabry Effect of long-term selection for increased leanness on meat and eating quality traits in Duroc swine J Anim Sci, June 1, 2006; 84(6): 1577 - 1583. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. B. Edwards, R. J. Tempelman, and R. O. Bates Evaluation of Duroc- vs. Pietrain-sired pigs for growth and composition J Anim Sci, February 1, 2006; 84(2): 266 - 275. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Chen, T. J. Baas, J. W. Mabry, and K. J. Koehler Genetic correlations between lean growth and litter traits in U.S. Yorkshire, Duroc, Hampshire, and Landrace pigs J Anim Sci, July 1, 2003; 81(7): 1700 - 1705. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Chen, T. J. Baas, J. C. M. Dekkers, K. J. Koehler, and J. W. Mabry Evaluation of strategies for selection for lean growth rate in pigs J Anim Sci, May 1, 2003; 81(5): 1150 - 1157. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
