| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
Department of Animal Science, Iowa State University, Ames 50011
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Key Words: Breeding Value Estimation • Intramuscular Fat Percent • Pigs • Selection
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
). Until recently, genetic evaluation and selection for IMF has been limited to sib and progeny testing to evaluate a sire’s ability to produce progeny with higher degrees of marbling. Real-time ultrasound has been reported to be a reliable method to estimate IMF in live pigs (Newcom et al., 2002). In addition, Newcom et al. (2003) reported similar heritability estimates for carcass and ultrasound IMF, and genetic correlations of similar magnitude for both IMF measures and indicators of meat quality (color, pH, tenderness, juiciness, flavor, etc.).
To maximize genetic response for a trait of interest, superior breeding stock must be accurately identified and selected. Direct selection for a trait measured on sibs after slaughter has been shown to result in less genetic gain than indirect selection using a closely related trait measured on the live animal (Martin and Fredeen, 1967). Incorporating pedigree or multiple trait information has been shown to yield more accurate EBV for traits of economic importance (Henderson and Quaas, 1976; Mabry et al., 1987).
The objectives of this study were to 1) compare selection candidate rankings based on breeding values estimated from three different models; 2) investigate the effects of breeding value estimates on selection differentials and expected response when selecting the top 1, 5, and 10% of boars and 50% of gilts; and 3) to simulate a progeny testing scheme by comparing breeding values for boars estimated immediately after their test period with breeding values estimated for them after progeny test records were completed.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
.
All barrows within a litter meeting the minimal weight requirement (>97 kg) were slaughtered 5 d after scanning. If a barrow was not available, a randomly chosen gilt was slaughtered. At 24 h postmortem, a 3.2-mm slice of the LM from the 10th- to 11th-rib interface was analyzed for carcass IMF by the method of Bligh and Dyer (1959). In total, 367 pigs were scanned, of which 130 were subsequently slaughtered in the base generation.
For Generation 1 matings, littermate pairs of gilts were randomly selected within each base generation litter, with one gilt designated to the control line and the other to the select line. Gilts within each littermate pair across both lines were mated to the same boar (via natural mating or AI) to maintain genetic ties between the lines before selection was initiated. A total of 24 sires from 14 sire families were used to produce 50 control and 45 select line litters in Generation 1. At weaning, two boars in each litter were randomly selected to remain intact, and all other boars in the litter were castrated. When Generation 1 animals reached an average of 110 kg, pigs were scanned and slaughtered according to the protocol discussed previously. In total, 616 pigs were scanned and 150 pigs were subsequently slaughtered.
Breeding Value Estimation
Breeding values used in the selection experiment were estimated for ultrasound and carcass IMF by fitting a two-trait animal model and the full relationship matrix with AI-REML in MATVEC (Model 1; Wang et al., 2003). Variance components and genetic parameters were estimated for ultrasound and carcass IMF records with data from 280 slaughtered pigs using the following model: y = Xb + Za + Hd + e, where y = the vector of observations; b = the vector of fixed effects (scan contemporary group, slaughter contemporary group, and sex); a = the vector of random additive genetic effects, which includes the numerator relationship matrix among animals; d = the vector of common litter effects, which was assumed to be uncorrelated with the random animal effects; and e = the vector of residuals. The incidence matrices relating observations to fixed, random animal, and common litter effects are X, Z, and H, respectively.
In the select line, the 10 boars and 75 gilts with the highest EBV for carcass IMF were selected. To minimize inbreeding, no more than two boars per sire family were selected; selection of full-sib boars was not permitted, and no more than five gilts per litter were selected. In the control line, one boar from each of the 14 sire families and 60 gilts representing all 14 sire families were randomly selected. Animals within each line were mated to produce Generation 2, with matings designed to minimize inbreeding (<5.0%) and ensure several litters by each selected boar.
In Generation 2, 56 select and 36 control line litters were produced. A total of 560 pigs were scanned, and 103 pigs were slaughtered. In Generation 3, 54 select and 38 control litters were produced. A total of 610 pigs were scanned, and 145 pigs were slaughtered. The genetic evaluation procedure described above was performed to rank selection candidates. In the select line, no more than two full-sib pairs and three half-sib boars per sire were selected to control inbreeding. In the control line, one boar from each sire family and 65 gilts representing all viable litters were randomly selected.
Model Comparison
The models compared were Model 1, a two-trait animal model used in the selection experiment, which included ultrasound IMF from all pigs scanned and carcass IMF from pigs slaughtered, to estimate breeding values for both carcass (C1) and ultrasound IMF (U1); Model 2, a single-trait animal model that included ultrasound IMF values on all pigs scanned to estimate breeding values for ultrasound IMF (U2); and Model 3, a multiple-trait model including carcass IMF from slaughtered pigs and the first three principal components, which explained 96% of the variation, from a total of 10 image parameters averaged across four longitudinal ultrasound images to estimate breeding values for carcass IMF (C3). Model 3 was included to compare whether the ultrasound IMF value needs to be estimated or whether the phenotypic image parameter values could be included in the breeding value estimation. Breeding values from Models 2 and 3 were estimated using the same model effects as Model 1. Estimated breeding values were compared using Spearman rank correlation coefficients of SAS (SAS Inst., Inc, Cary, NC). Simulating a selection scenario where the top 1, 5, and 10% of boars and top 50% of females were selected based on EBV, selection differentials, expected responses, and loss in selection differentials were compared for the three models using select line boars and gilts from Generation 3 (n = 299). Models also were compared using EBV for all sires (n = 83) with progeny data across the four generations to compare results as if a progeny test had been completed.
|
Results and Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
. Mean values for carcass- and ultrasound-derived IMF were similar for the first three generations of selection with the exception of a slight decrease in mean ultrasound IMF in Generation 2. Reasons for this decrease are unclear, but it may be due to random genetic drift or normal biological variation. Pearson correlations and Spearman rank correlations within year were similar. This indicates that the ability to predict IMF and/or rank animals was similar for all three generations.
|
. Genetic, common litter, and residual variances and standard errors for U1 and U2 were similar, indicating similar variances for common model effects. Variance component differences existed between models used for estimation of carcass and ultrasound IMF. In particular, genetic variance for C1 and C3 were much larger than for U1 and U2. These differences, combined with variation in litter, residual, and total phenotypic variances, resulted in heritability estimates that were nearly or more than twofold greater for C1 and C3 compared with U1 and U2. These greater heritability estimates offer the potential for greater genetic improvement of IMF by using carcass measurement compared with ultrasonic evaluation.
|
when evaluating ultrasound IMF across multiple breeds. Heritability estimates for carcass IMF were numerically different, but had similar standard errors, having a higher heritability for carcass IMF when including principal components in the multiple-trait model compared with the model that included ultrasound IMF. However, both heritability estimates for carcass IMF were close to the mean of 0.50 from 19 estimates reported by Sellier (1998) and within the reported range of 0.26 to 0.86. The genetic correlation between U1 and C1 was 0.82, which indicates the two traits are likely controlled by many of the same genes. This is slightly larger than the value of 0.76 reported by Newcom et al. (2003).
Rank Correlations
Spearman rank correlation coefficients between breeding values estimated from three different models for all pigs scanned are shown in Table 3. The largest correlation was between U1 and U2 (0.97). This high correlation between breeding values could be due to the limited effect the inclusion of sib carcass IMF into the genetic evaluation has on ultrasound IMF breeding value estimation. These results indicate that animals ranked nearly the same when ultrasound IMF was evaluated with carcass IMF in a two-trait animal model or alone in a single-trait model. This similarity in boar rankings by including or excluding sib carcass data is similar to the effect of including or excluding reference sires from the genetic evaluation of swine test station data reported by Mabry et al. (1987). The rank correlation coefficient between C1 and U2 was 0.86. This indicates that animals are not ranked exactly the same, and differences in selection differential and response would be expected depending on the EBV used for selection. The rank correlation between C1 and C3 was 0.92, indicating the two EBV rank animals similarly. Rank correlations between U1 and U2 with C3 were smaller (0.81 and 0.68, respectively), indicating that changes in EBV rank are expected when comparing selection candidates for IMF across models.
|
). When evaluating EBV ranking from pigs in Generation 3, the rank correlation coefficient between U1 and U2 was 0.98, whereas the correlation coefficient between C1 and U2 was 0.93 (data not shown). This finding indicates pigs from the current generation rank more similarly for C1 and U2 than do all pigs from all four generations. This was most likely due to an improvement in ultrasound image quality, which should lead to more accurate ultrasound IMF values. The rank correlation coefficients reported should be interpreted carefully, as the magnitude results, at least in part, from the high genetic correlation between the traits and the structure of the data, with approximately 25% of the pigs having both measures of IMF.
Selection Differential and Estimated Response
Selection differentials, loss in selection differential, estimated response, and loss in estimated response across the three models for select line pigs from Generation 3 and for sires with progeny records are shown in Table 4. When selecting the top 10% of boars in the select line, the greatest decrease in carcass IMF response was found when ranking boars based on C3 (0.49 vs. 0.38%; 22.8%). Selection based on U2 was only slightly better at 18.5% (0.49 vs. 0.40%). Selection based on U1 resulted in a loss of 5.3% response (0.49 vs. 0.46%) in carcass IMF. When selecting the top 5% of boars, the three alternative EBV (U1, U2, and C3) differed little in lost selection differential compared with C1. For selection of the top 1% of boars, ranking boars based on C3 resulted in a 27.1% loss (0.59 vs. 0.43%) in selection differential. Because selecting the top 1% resulted in only a single boar being selected, the other three models all ranked the top boar the same, so no loss in selection differential was found for rankings based on either U1 or U2. If the population had been large enough that the top 1% had resulted in more boars being selected, Models 1 and 2 may not have selected identical sires. The greatest loss in selection differential when selecting gilts was also from rankings based on C3.
|
) was calculated as (EBVS + EBVD)/2, where EBVS is equal to the mean genetic superiority of the selected boars (mean EBV of selected group minus mean EBV of selection candidates), and EBVD is equal to the average genetic superiority of the selected gilts (mean EBV of selected group minus mean EBV of selection candidates; Falconer and Mackay, 1996). Results from the estimated response and loss in estimated response analysis closely mirror the results from the selection differential and loss in selection differential analysis. Selection based on IMF EBV from Model 3 resulted in the lowest estimated response in C1, regardless of the selection intensity of the boars. When selecting the top 10% of boars, the lowest loss in estimated response was noted when selection is based on the rankings for ultrasound IMF EBV from Model 1, and could be indicative of the added information provided by inclusion of sib carcass IMF data into the genetic evaluation for ultrasound IMF. When selecting the top 5 or 1% of boars, selection based on ultrasound IMF EBV from Models 1 and 2 was similar. The difference in loss of estimated response between selecting the top 10% and 1 or 5% could be due to the fact that the population was small enough that the top boar (1%) or top six boars (5%) may have been so outstanding that the model used for ranking selection candidates was inconsequential.
Results from the simulated progeny testing scheme were similar to the selection differentials and loss in selection differential from Generation 3 (Table 4). The loss in selection differential by selecting on ultrasound IMF EBV from Model 2 compared with selection based on carcass IMF EBV from Model 1 ranged from 1.7 to 11.0%, depending on whether selection was based on individual or progeny EBV and the percent selected. The differences in the results between the individual selection and progeny testing methods are attributable to the selection differentials being twice as large when selecting boars based on their own EBV after progeny are tested compared with selection based on an individual’s own EBV immediately after the test period. Corresponding values for selection based on the mean progeny EBV for tested boars was only 1.5 times greater than with individual selection without progeny testing. If one considers the lengthened generation interval (2 yr as opposed to 1 yr) created by progeny testing (Falconer and Mackay, 1996), the selection differentials are approximately equal for individual selection or for progeny testing and selecting sires based on their own EBV. Selection of sires after the progeny test based on the mean EBV of their progeny would produce smaller selection differentials than the other two methods of selection. In addition, if a progeny-testing scheme were implemented, fewer boars would likely be tested due to space limitations, thereby further decreasing the selection intensity. However, in a more practical selection program, where males and females are kept for more than one farrowing season, sires could be ranked by including individual records along with pedigree records and selected across generations. Although the breeding values for ultrasound IMF were half as large as for carcass IMF, a correlated response in carcass IMF would be expected due to the high genetic correlation.
Using the formula i x 
P x h2, where i is the intensity of selection, P is the phenotypic standard deviation, and h2 is the heritability (Falconer and Mackay, 1996), estimated genetic gains for ultrasound and carcass IMF from Generation 3 data are shown in Table 5. The intensity of selection for carcass IMF selection is the proportion the top 10 boars represented in Generation 3 and the top 50% of gilts. The intensity of selection for ultrasound IMF selection is based on an assumption that all males (boars and barrows) would be available for selection if no carcass data were collected; therefore, the proportion made up by the top 10 boars is smaller, thereby increasing intensity. The correlated responses in carcass IMF based on ultrasound IMF selection and ultrasound IMF based on carcass IMF selection are also shown. The results demonstrate that expected genetic progress is greater when selection is based on carcass IMF than when selection is based on ultrasound IMF. When selecting for ultrasound IMF, the correlated response in carcass IMF was estimated to be 0.52% compared with the 0.75% estimated by direct selection on carcass IMF, a loss in estimated genetic gain of 31%. This contrasts the results of Martin and Freeden (1967), who showed indirect selection for carcass percent lean cuts based on selection for decreased backfat produced genetic gains from 9 to 38% greater than when selecting on sib data collected after slaughter, depending on the heritability assumed for percent lean cuts. The contradiction between the result of Martin and Freeden (1967) and the current study can be attributed to the differences in the traits studied, where the heritability of the directly selected trait is higher than that of the indirectly selected trait. Additionally, the differences in selection intensity between the two methods of selection were small (Table 5), and the directly measured trait had substantially greater variation than the indirect. Stalder et al. (2003) found greater estimated genetic gain in pH when selecting sires (progeny test) based on hydrogen ion concentration compared with selecting sires based on pH, primarily due to the higher heritability for hydrogen ion concentration.
|
) also were used to estimate genetic gain. Genetic gain was estimated with improvement in carcass IMF as the breeding goal using the Selection Index Program (Wagenaar et al., 1995) with three different indices: 1) phenotypic ultrasound IMF records collected on an individual, parents, and sibs (full and half); 2) phenotypic ultrasound IMF records collected on an individual, parents, and sibs, and phenotypic carcass IMF records collected on full and half sibs; and 3) phenotypic carcass IMF records collected on full and half sibs. Selection intensities were taken from Table 5. The greatest genetic gain was estimated from the index that included only ultrasound IMF records (0.86%). Genetic gain estimated from ultrasound and carcass IMF records was similar (0.83%). The index with the highest accuracy included ultrasound and carcass IMF records, but with the decrease in selection intensity, genetic gain was lower than when including only ultrasound IMF records. The lowest estimated genetic gain was from the index that included only carcass IMF records from close relatives, indicating that the addition of ultrasound IMF records improves the potential genetic gain in a breeding program. This finding shows that when resources for boar production are unlimited, genetic gain could be maximized when all boars remain intact to increase the size of the selection candidate pool. Selections could then be based on an EBV calculated with a selection index from ultrasound records on the boars themselves and close relatives.
Of the top 12 boars in Generation 3 ranked on carcass IMF EBV from Model 1, nine of those boars were in the top 12 based on ultrasound IMF EBV from Model 1, and six were in the top 12 based on ultrasound IMF EBV from Model 2 and carcass IMF EBV from Model 3. This shows that more of the same boars would be selected based on rankings for ultrasound IMF EBV from Model 1, the same model currently used to select boars based on carcass IMF EBV.
When implementing a selection program, the goal is to select individuals that are superior for the trait of interest and keep them for breeding purposes. In this study, the genetic evaluation using a two-trait animal model including IMF from real-time ultrasound on all pigs and carcass IMF from those pigs slaughtered resulted in the greatest levels of genetic gain compared with a single-trait evaluation for IMF only from ultrasound. In addition, the results show that incorporation of sib carcass data for IMF improves the accuracy of identifying individuals to keep for breeding purposes. Rank correlations show that similar animals would be selected based on rankings for EBV from Models 1 or 2, but rankings based on Model 3 resulted in substantial loss in estimated response.
The preferred method to rank individuals in a breeding program is to estimate breeding values for several traits (growth, backfat, loin muscle area) with a multiple-trait, mixed-model analysis. A selection index is then used to weight each breeding value by its appropriate economic weight and rank animals according to that index value. Incorporating IMF estimated from real-time ultrasound into an index of this type would most likely be more beneficial than carcass IMF. A real-time ultrasound estimate of IMF could be collected at the same time backfat and loin muscle area are measured, decreasing the time needed to collect the data and eliminating the need for carcass data. One drawback to this selection index method is the lack of reliable estimates of economic values for IMF. Further research needs to be conducted to fully understand this problem.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
2 Current address: Genetic Improvement Services of North Carolina, Inc., P.O. Box 9, Newton Grove, NC 28366.
3 Correspondence: 109 Kildee Hall (phone: 515-294-6728; fax: 515-294-5698; e-mail: tjbaas{at}iastate.edu).
Received for publication June 14, 2004. Accepted for publication December 23, 2004.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. P. Zhao, J. L. Chen, M. Q. Zheng, J. Wen, and Y. Zhang Correlated Responses to Selection for Increased Intramuscular Fat in a Chinese Quality Chicken Line Poult. Sci., November 1, 2007; 86(11): 2309 - 2314. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
