| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
* ARS, USDA, Subtropical Agricultural Research Station, Brooksville, FL, 34601; and
and
Department of Animal Sciences, University of Florida, Gainesville, 32611
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
Key Words: Brahman • feedlot • height • random regression • body weight
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
One appropriate analysis technique involves the use of covariance functions with orthogonal polynomials of time as coefficients of these functions (Kirkpatrick et al., 1990
; Meyer and Hill, 1997). In this method, a set of random regression coefficients is fitted for additive genetic and permanent environmental effects for each animal. Covariances between these coefficients are equal to the coefficients of a covariance function. This covariance function is subsequently utilized to estimate variances (and other parameters) across time. These methods have been employed for analyses of BW traits in beef cattle (Meyer, 1998a; 2001a; Schenkel et al., 2002; Arango et al., 2004), traits measured by real-time ultrasound in beef cattle (Hassen et al., 2003, 2004), and in production traits of other species (Kominakis et al., 2001; Huisman et al., 2002).
The ratio of BW to hip height has been an indicator of BCS in cows (Klosterman et al., 1968) and calves (Brown et al., 1993, 1997) but may represent a unique body dimension. Greater BW at smaller hip heights may be desirable and could be important for Brahman cattle because the markets for domestic commercial females and international purebreds appear to have conflicting animal size requirements.
The objective of this research was to assess the genetic control of BW, hip height, and the ratio of BW to hip height in Brahman cattle through 170 d on feed using covariance function-random regression models.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
Experimental Design
In 1994, a progeny test of Brahman bulls for beef carcass and palatability attributes was begun at the Subtropical Agricultural Research Station (STARS). The project was originally designed to last 5 yr; matings were continued for 2 additional years at the request of the American Brahman Breeders Association to include some Brahman bulls in the Carcass Merit Project of the National Cattlemen’s Beef Association.
The experimental design was detailed previously (Riley et al., 2002). In brief, Brahman cows at STARS were exposed to 5 or 6 Brahman bulls in single-sire breeding herds each year. After the first year, a bull that sired calves in the previous year was again used. In the 1999 and 2000 breeding seasons, some cows were bred by AI to achieve target numbers of progeny for the Carcass Merit Project sires. The breeding season began on approximately March 20 of each year and lasted for 105 d. Calves were born from late December through late April or early May of each year. Shortly after birth, calves were weighed and tagged, and bull calves were castrated. They were weaned in September of each year at approximately 7 mo of age. Calves were maintained in a drylot for 2 to 3 wk postweaning.
Calves were sorted into STARS feedlot pens by sex and BW; the pen capacity was about 13 calves. All but the very smallest calves were fed. Calves began the feeding period on a diet that consisted of approximately (as-fed basis) 55% corn; 25% cottonseed hulls, ground hay, or both; 15% supplement (which contained melengestrol acetate for heifers, and monensin, vitamin A, and microminerals for all calves); and 5% molasses. The diet was gradually changed over 28 d to the final diet, which consisted of 72.5% corn; 15% cottonseed hulls, ground hay, or both; 7.5% supplement; and 5% molasses. Steers and heifers were implanted with Synovex-S or Synovex-H (Fort Dodge Animal Health, Fort Dodge, IA) at 0 and 112 d of feeding. After 140 d of feeding, external fat cover was estimated using real-time ultrasound in conjunction with monthly data collection. When the median backfat of the animals in a pen was 10 mm, the entire pen was slaughtered commercially in Central Florida.
The STARS Brahman herd was begun in 1949 with the purchase of heifers and a bull from a Texas breeder. The herd was increased by keeping replacement heifers and by periodic purchases of heifers or mature females. Brahman bulls have regularly been obtained from Florida breeders, and in several years bulls from Louisiana, Texas, and Georgia have been used. The Brahman bulls (n = 27) with calves in this project were mostly from Florida breeders, but a few were from Texas and Louisiana. They represented a broad sampling of Brahman bloodlines and included prominent bulls of the breed at that time.
Statistical Analyses
Over the 7 project years, steers (n = 342) and heifers (n = 382) had BW (kg) and hip height (cm) recorded every 28 d while on feed. The ratio of BW:hip height x 100 was calculated for each animal on each date. Some were fed longer, but all calves had records until 170 d on feed; subsequent records were removed from the final data set. This resulted in 6 records for almost all animals in the project. Calves that were injured, foundered, or sick (n = 17) at any time in the feeding period were removed from the project, and all of their records were removed from the data set.
The random regression model used can be represented by the following equation, which is based on that presented by Meyer (1998a):
 |
where yij is the jth record of the ith steer or heifer; FE represents the fixed effects; kA and kR represent the order of the fitted polynomial regressions for the additive genetic and permanent environmental components of variance; 
im and 
im represent the additive genetic and permanent environmental random regression coefficients for the ith steer or heifer, t* ij is the jth standardized day on feed of record of the ith steer or heifer (from –1 to 1 on the Legendre scale), and 
m (t*ij) is the mth polynomial evaluated for each standardized day on feed (t*ij). The number of records and structure of the data did not permit estimation of other effects, such as maternal genetic or permanent environmental variance.
The strategy for analyses was as follows: 1) Determination of the fixed effects portion of models; 2) Determination of the covariance structure for residual effects; and 3) Determination of the orders of the genetic and permanent environmental covariance polynomial regressions.
Preliminary analyses were conducted using the MIXED procedure of SAS (Littell et al., 1996; SAS Inst. Inc., Cary, NC) to build the fixed effects portion of the models. Sire and calf within sire were included in these preliminary analyses as random effects, and the fixed effects that were investigated included calf age at beginning of test in days as a covariate, sex of calf, feeding pen (n = 10; not all pens were used in all years), year of record (n = 7), and all interactions of sex, pen, and year. Contemporary groups (n = 63) were calves of the same sex, born in the same year, fed in the same pen, and slaughtered on the same day. The number of calves for individual sires within contemporary groups ranged from 1 to 7, and the number of sires represented in individual contemporary groups ranged from 4 to 7.
For each trait, after other fixed effects were determined, a fixed regression of dependent variable on orthogonal polynomials of days on feed was fitted to model the phenotypic population trajectories. The order of this regression for each trait was determined in preliminary mixed model analyses in SAS as the highest order of regression that was significant. In all subsequent steps, the finalized fixed effect portions of the models were held constant.
The residual error structure was investigated using models with a single set of random regression coefficients on Legendre polynomials of days on feed, representing an overall animal effect; that is, with no attempt to separate additive genetic from permanent environmental effects (Meyer, 1999). These residual error structures included from 1 to 7 discrete measurement error structures, in which a variance was estimated corresponding to intervals of days on feed. In each case, these intervals were distributed equally across days on feed; e.g., when 1 variance was estimated, it corresponded to the entire feeding period; when 2 variances were estimated, they corresponded to two 85-d periods, etc.; up to a maximum of 7 periods.
Likelihood ratio tests were used to compare the significance of additional variances estimated in nested models. Akaike’s information criterion values and Schwarz’s Bayesian information criterion values were also calculated and considered. This procedure was conducted separately for linear, quadratic, cubic, and quartic overall animal regressions. Results from these separate analyses were examined for consistencies across all 4 sets of analyses, in other words, to see if different measurement error structures would be preferred for different orders of overall polynomial regressions. The final covariance structure was determined considering these results, and the overall desire and need for parsimony. The program DXMRR (Meyer, 1998b,c) was used to complete this step of the analyses.
Two sets of random regression coefficients on Legendre polynomials of days on feed were used to model additive genetic and permanent environmental effects for each trait. Linear, quadratic, cubic, and quartic orders of each polynomial regression were investigated, and covariance functions for the additive genetic and permanent environmental effects were thereby estimated. Additive genetic and permanent environmental effects were assumed to be uncorrelated. The relationship matrix was constructed by incorporation of all available pedigree data of the STARS Brahman herd and 3 to 5 generations of pedigree information for the 27 sires. There were 3,896 animals in the mixed model equations; less than 20% (n = 724) had records. No parents themselves had records, and no sire had daughters that produced calves that had records. Log-likelihood values, estimates of covariance, and likelihood ratio tests were all considered in the selection of the most appropriate polynomial for the additive genetic and permanent environmental regressions. However, because most of the models compared were nonnested, Schwarz’s Bayesian information criterion values were used to make final model selections within each trait, in the same manner as described by Jaffrézic and Pletcher (2000) and Meyer (2001a). The program DXMRR (Meyer, 1998b,c) was used to complete this step.
Two-trait analyses were conducted for each trait for all pairs of days on feed using MTDFREML (Boldman et al., 1995) to estimate covariance components and genetic parameters for comparison with those from random regression analyses. Three-trait analyses were also conducted using MTDFREML to estimate covariances and genetic correlations for pairs of traits on the different days on feed. Data for these analyses consisted of 1 record per animal for a given time on feed, and these corresponded to 28-d intervals on which BW and hip height were recorded. These models included contemporary group, age in days at time of record as a linear covariate, and animal as a random effect.
For each analysis step, analyses were repeated with at least 3 sets of beginning values (for polynomial coefficients or variances) to attempt to avoid convergence to local maxima. A convergence criterion of 10–9 was used for all analyses. The estimated functions were used to generate estimates of covariances and the proportions of the total phenotypic variance for covariance components for selected days on feed.
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
. There were 724 calves with records in this project. Actual numbers of observations for each trait by average days on feed ranged from 720 to 724, for a total of 5,155 BW, 5,063 hip height, and 5,055 BW:hip height records.
|
Error Variance Structure
Although Akaike’s and Schwarz’s Bayesian information criterion values always favored the maximum number of measurement error categories (n = 7), likelihood ratio tests indicated that log likelihood values increased significantly only through the addition of the second measurement error variance. The most consistent result across the different phenotypic regression analyses for all traits was the addition of the nonsignificant third measurement error variance. Based upon these considerations, 2 measurement error categories were modeled in subsequent analyses of all traits for estimation of genetic and permanent environmental covariances: the first was used for records up to 85 d on feed and the second for records from 86 through 170 d on feed. Simple statistics for traits by measurement error category are shown in Table 2. In random regression analyses of bull test weights, Schenkel et al. (2002) utilized 6 measurement error categories that corresponded to the actual weighing periods. Hassen et al. (2003) investigated 6 error variances (analyses of intramuscular fat measured by ultrasound) but concluded that nothing more than a single measurement error term was warranted. Results from analyses of loin muscle area measured by ultrasound (Hassen et al., 2004) were conducted using 3 measurement error variances.
|
Schwarz’s Bayesian information criterion values favored models that included linear random regressions to model the additive genetic effects and cubic (hip height and BW:hip height ratio) or quartic (BW) polynomials to model the permanent environmental effects (Table 3). These results were consistent with other reports that a higher order polynomial was justified to model permanent environmental effects of BW traits in growing bulls (Schenkel et al., 2002) and in mature cow BW (Arango et al., 2004). Other studies have investigated and reported differing polynomials for various other effects (Albuquerque and Meyer, 2001; Meyer, 2001a), but those studies consisted of 5 to 10 times the number of animals and records. It was unexpected that the smaller data set of the current study would support estimation of differing polynomials. Linear regressions (for additive genetic and permanent environmental covariance functions) provided the best fit for random regression analyses of percentage intramuscular fat (measured by real time ultrasound, Hassen et al., 2003), but quadratic regressions provided best model fit for random regression analyses of longissimus muscle area (Hassen et al., 2004). Nobre et al. (2003), in analyses of BW from birth to approximately 22 mo of age in Nellore cattle, selected cubic regressions for all effects modeled, including additive direct and maternal, and direct and maternal permanent environmental effects.
|
Body Weight.
Coefficient estimates for covariance functions from random regression analyses for BW are shown in Table 4, and graphical depiction of the different estimates of variance is presented in Figure 1. Estimates of heritability for BW ranged from 0.31 early in the feeding period to 0.53 at the end of the feeding period (Figure 2). Schenkel et al. (2002) reported a range of estimates of heritability generated by random regression analyses from 0.32 to 0.4 for BW of bulls in growth through 140 d on feed. Estimates from the current study also appeared to be consistent with estimates from conventional (not random regression) analyses, as Newman et al. (2002) reported an estimate for 400-d BW from purebred data (including Brahman, among other breeds) in Australia of 0.45, and Koots et al. (1994) reported an average yearling BW estimate of heritability of 0.33 from compiled studies. The permanent environmental proportion of phenotypic variance was greatest at about d 28 (0.58) and lowest (0.44) at d 170 (Figure 2). Estimates of permanent environmental variance for bulls on a similar test period (Schenkel et al., 2002) also increased but were larger in magnitude (range from 490 to 930 kg2) and somewhat lower as a proportion of phenotypic variance (around 0.4) than the results from the current study. Estimates of repeatability for BW ranged from 0.8 to 0.97. Estimates of heritability from conventional (not random regression) analyses (Table 5) were less than 0.40 at 1 and 28 d on feed; were near 0.50 for 57, 85, and 113 d on feed; and were close to 0.58 for 140 and 168 d on feed.
|
|
|
|
). This increase was much smaller than the increases in the other modeled components of phenotypic variance and the total phenotypic variance (Figure 1). Estimation of 2 measurement errors may have been all that the data of the current study could support; however, there may be a more appropriate way to model a gradual increase in measurement error across days on feed, such as those alternatives proposed by Meyer (2000, 2001b).
Hip Height.
Estimated function coefficients from hip height random regression models and measurement error variances for hip height are shown in Table 4. Resultant estimates of variance for hip height for representative days on feed are shown in Figure 3. The estimates of heritability for hip height increased from 0.37 to 0.53 by about 85 d on feed and remained somewhat constant through 170 d on feed (Figure 4). Estimates of the proportion of phenotypic variance due to permanent environmental effects were low; they decreased from about 0.15 at the beginning of the feeding period to 0.09 through d 113, then increased to 0.13 at d 170 (Figure 4). Those estimates for hip height from conventional analyses in which BW from different months were analyzed as separate traits (Table 6) had a similar range (from 0.44 to 0.51) for all days on feed except d 85 (0.69 ± 0.11) and 140 (0.63 ± 0.11). Sampling error associated with the small number of records in this study seems a likely cause of these 2 high estimates. These high estimates were less pronounced in the trajectory from random regression analyses (Figure 4). This may represent a (beneficial) smoothing effect of the inclusion of repeated records in the random regression analyses on the 2 high estimates. High estimates (>0.70) of heritability for hip height for growing or mature cattle are not uncommon in the literature in Bos taurus (e.g., Gregory et al., 1995a,b; Choy et al., 2002), and in Brahman (Vargas et al., 1998, 2000); however, Kriese et al. (1991) reported a lower estimate of 0.27. Estimates of repeatability for hip height in the current study ranged from 0.56 to 0.66. The estimate of measurement error variance for the second half of the days on feed increased almost 40% (Table 4), which was approximately the same the increase observed for BW measurement error categories.
|
|
|
) were used in determination of covariance functions, which were in turn used to produce estimates of the various components of variance for 7 representative days on feed (Figure 5). Estimates of heritability ranged from 0.23 (d 1 on feed) to the greatest estimate of 0.60 at 170 d on feed (Figure 6). The proportion of phenotypic variance due to permanent environmental effects was greater (0.44 to 0.48) at earlier days on feed (28, 57, and 85 d on feed) than d 1 or the later days (Figure 6). Estimates of repeatability ranged from 0.61 to 0.9. Estimates of heritability from 2-trait analyses (Table 7) were similar to results from random regression analyses. They were lowest (0.21) at the first weigh date (d 1 on test) and steadily increased with days on feed to a high of 0.64 on d 170.
|
|
|
Within-Trait.
When in the sequence of repeated measures for a trait does it become a "different" trait? The question may be more related to the change (increase) in variance that occurs over time. Within all 3 traits, estimates of genetic correlations (calculated at different days on feed) generally were large and positive for any pair of rankings from 28 to 170 d on feed (Tables 5, 6, and 7); those estimates from random regression analyses were in almost all cases quite similar to those from 2-trait analyses. Results imply that genetic analyses of hip height (Table 6) from any single day within the time on feed would separate individuals in the same manner; multiple measurements or more sophisticated genetic analyses (random regression), may not be as valuable. However, there was a lower positive correspondence of breeding values for BW:hip height ratio from times early in the feeding period with those from later in the feeding period (Table 7). Estimates of genetic correlation for BW across the feeding period (Table 5) were intermediate to those of the other 2 traits yet were large and positive.
Across-Traits.
There were strong positive estimates of genetic correlation for BW:hip height ratio with BW and for BW with hip height for the different days on feed (Table 8). Those estimates for BW:hip height with hip height did not differ from 0. This lack of association between breeding values seems potentially important. It is relatively easy to envision that a range of BW:hip height ratio phenotypes could occur for a given hip height phenotype. Confirmation of the independence of breeding values for these 2 traits would imply ultimately that BW:hip height could be increased without affecting hip height.
|
Maternal effects may be important for BW as a carryover from weaning weight, at least early in the feeding period; they could also be important for the BW:hip height ratio. Others have reported the decreasing influence of maternal effects on BW from weaning up to a year of age (Albuquerque and Meyer, 2001; Arthur et al., 2001; MacNeil, 2003). The estimated variances in this study may be inflated due to maternal effects; Schenkel et al. (2002) also acknowledged this potential in their random regression analyses of postweaning bull BW. Estimates for hip height are likely less affected by maternal effects because their influence has been reported to be quite low (Vargas et al., 1999). Maternal permanent environmental effects may also be important for postweaning weights (Albuquerque and Meyer, 2001; Mercadante et al., 2003).
Body Weight:Hip Height Ratio Considerations
Results indicate substantial additive genetic control of Brahman BW:hip height ratio, and that control appears to increase with days on feed. Body weight:hip height ratio has only been considered as a substitution for subjective assessment of BCS in cows (Klosterman et al., 1968; Nelson et al., 1985) or in calves (Brown et al., 1993, 1997). Klosterman et al. (1968) reported a strong positive correlation of this ratio and condition score in Hereford and Charolais cows. Even though there appeared to be great breed differences for hip height (11 to 12 cm) in their data, there did not appear to be corresponding breed differences in BW:hip height ratio. The BCS were not recorded in the current study because it was expected that scores would be uniformly high for animals on feed, with minor exceptions for sick or injured animals.
It may be beneficial to try to relate Brahman research results to mature breeding animals, because one of the major contributions of the Brahman breed in the United States is through the commercial cow herd. The results of this study should be compared with similar evaluation of BW:hip height in adult animals. The current support among some commercial cow-calf producers for smaller-framed, yet heavier beef cattle may be based on belief in the higher reproductive (and perhaps overall) efficiency of smaller-framed animals. Superior pregnancy rates of smaller-framed Brahman cows have been reported (Vargas et al., 1998). There are many definitions for efficiency of beef production; these are often unique to a particular beef industry sector. The genetic control of such efficiency traits and their components is, in many cases, poorly understood.
Can BW:hip height be considered a useful measure of body dimensions in Brahman cattle? Selection for a ratio of 2 traits may be difficult and have limited effectiveness (Gunsett, 1987; MacNeil, 2005); an index created from breeding values of the 2 traits may be preferred. The large estimates of heritability at later days on feed seem promising for selective change. Additional work is needed to verify that BW:hip height measures more than BCS for cattle, especially in animals on pasture, and that it represents a trait that is distinct from BW alone.
|
IMPLICATIONS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
|
Footnotes |
|---|
2 Appreciation is extended to M. L. Rooks, E. J. Bowers, V. E. Rooks, E. L. Adams, and the STARS staff for technical assistance and animal care.
4 Current address: ARS, USDA, Pacific West Area, 800 Buchanan Street, Albany, CA 94710-1105.
3 Corresponding author: dgriley{at}ifas.ufl.edu
Received for publication November 8, 2005. Accepted for publication August 23, 2006.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONIMPLICATIONSLITERATURE CITED |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. J. Garrett, G. Rincon, J. F. Medrano, M. A. Elzo, G. A. Silver, and M. G. Thomas Promoter region of the bovine growth hormone receptor gene: Single nucleotide polymorphism discovery in cattle and association with performance in Brangus bulls J Anim Sci, December 1, 2008; 86(12): 3315 - 3323. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
X. Qiu, J. D. Arthington, D. G. Riley, C. C. Chase Jr., W. A. Phillips, S. W. Coleman, and T. A. Olson Genetic effects on acute phase protein response to the stresses of weaning and transportation in beef calves J Anim Sci, October 1, 2007; 85(10): 2367 - 2374. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
