Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Genetic prediction for first-service conception rate in Angus heifers using a random regression model

Genetic prediction for first-service conception rate in Angus heifers using a random regression... Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic prediction for first-service conception rate in Angus heifers using a random regression model 1, Miguel A. Sánchez-Castro, Milton G. Thomas, R. Mark Enns, and Scott E. Speidel Department of Animal Sciences, Colorado State University, Fort Collins, CO 80523-1171 © The Author(s) 2020. Published by Oxford University Press on behalf of the American Society of Animal Science. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Transl. Anim. Sci. 2020.4:S43–S47 doi: 10.1093/tas/txaa094 predictions because edited data do not appropri- INTRODUCTION ately represent the population’s true variability First-service conception rate (FSCR) could be (Misztal et al., 1989). defined as the probability of a heifer to conceive Random regression models (RRM) represent in response to her first artificial insemination (AI) an alternative method to evaluate binary traits and maintain such pregnancy after the end of the and can incorporate data from contemporary breeding season. Economic implications of this groups with no variation (Golden et  al., 2018). trait in beef cattle heifers involve its relationship Nonetheless, since they were originally concep- with their development costs, as well as costs of tualized to analyze longitudinal traits, the effi- semen, synchronization protocols, estrus detec- cacy of RRM to evaluate traits with phenotypes tion and AI services (Horan et  al., 2005; Minick observed only once has not been deeply explored. Bormann et al., 2006). Furthermore, FSCR is also Relative to heifer fertility, these models have only related to differences in quality and value between been applied in genetic evaluations for heifer preg- AI-produced calves and natural service calves. nancy (HPG) in Red Angus cattle (Speidel et al., Additionally, heifers conceiving on their first ser - 2018). Therefore, objectives herein were to evalu- vice, calve earlier within the calving season, have ate the efficacy of a RRM genetic prediction for more chances to breed postpartum within a year, heifer FSCR by comparing the resulting EPD to and have more time to nurse and wean heavier those obtained using a traditional ATM, as well calves (Marshall et al., 1990). as to compare genetic parameters obtained with Given the binary nature of pregnancy out- each methodology in Angus cattle. comes after an AI, the method of choice to eval- uate FSCR has been an animal threshold model MATERIALS AND METHODS (ATM). However, such method has a lack of abil- The data used in the present study were ity to incorporate information of contemporary obtained from an existing database; therefore, the groups with no variation (e.g., extreme-category study was not subjected to animal care and use problem). When all observations within a categor- committee approval. ical fixed effect fall in one of the binary classes, typically, these observations are deleted from the data in order to calculate expected progeny differ- Data Collection and Editing ences (EPD). This can lead to distorted genetic Breeding and ultrasound records of 4,334 Angus heifers (progeny of 354 sires and 1,626 Corresponding author: scott.speidel@colostate.edu dams) collected from 1992 to 2019 at the Colorado Received March 13, 2020. Accepted June 22, 2020. State University Beef Improvement Center S43 Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Sánchez-Castro et al. S44 (CSU-BIC) were used for the study. Within each Additionally, FSCR was regressed on age at AI breeding year, heifers were estrus synchronized and using a linear RRM with Legendre polynomials subjected to AI only once, before they were exposed as the base function. The model in matrix form is to natural service sires approximately 3  wk after presented below: insemination. Observations for FSCR (1, successful; y = Xb + Zu + Qm + e 0, unsuccessful) were defined by fetal age obtained where y corresponded to a vector of binary obser- from ultrasound pregnancy exams performed 130 d vations of FSCR, b was a vector of unknown so- post-AI. Although FSCR in 12- to 15-mo-old heif- lutions for categorical fixed effects (contemporary ers is a singly observed phenotype, its expression is group, AI technician, and age of dam) and a linear likely to be dependent on age of onset of puberty. fixed regression of FSCR on age at AI, u corres- Consequently, age at AI was an important factor ponded to a vector of unknown solutions of animal included in all statistical models, and it was calcu- random additive genetic regression coefficients lated as the difference between an individual’s birth- (intercept and linear), m was a vector of unknown date and the date when they were subjected to AI. solutions for mating group random effects. X, Z, and Q were known incidence matrices relating ob- Statistical Analysis servations in y to both fixed and random effects, and e was the vector of unknown residual errors. Traditional EPD calculation for FSCR was per- Variances assumed for the models were: formed using a univariate BLUP threshold animal     model along with a probit link function to convert A ⊗ G 00   binary observations to an underlying normal distri-   m 0 I σ 0 Var =  m  bution. The model equation was as follows: 2 00 I σ y = Xb + Zu + Qm + e where A represented the Wright’s numerator rela- where y* corresponded to a vector of transformed tionship matrix, ⊗ was the Kronecker product, and observations of FSCR on the underlying scale, b G corresponds to a modified variance–covariance was a vector of unknown solutions for fixed ef- matrix of additive genetic random regression coef- fects that included breeding contemporary group ficients where the covariance between the intercept (defined as a combination between breeding year and the linear term was fitted to zero, given no and semen type), AI technician and age of dam, heifer had more than one observation for FSCR. 2 2 individual’s age at AI was included as a linear I , I , σ , and σ remained as described for the m e m n covariate, u corresponded to a vector of unknown previous model. For the estimation of the genetic solutions of animal random effects, and m was parameters, the entire pedigree from the CSU-BIC a vector of unknown solutions of mating group consisting of 13,983 individual animals, as well (e.g., inseminated on heat or during mass mating) as 967 and 3,699 unique sires and dams, respec- random effects. X, Z, and Q were known incidence tively, was used. The average inbreeding coefficient matrices relating observations in y* to both fixed of this pedigree was 0.009. Once the parameters and random effects, and e was the vector of un- were obtained, they were compared to each other known residual errors. For this model, variances and subsequently used to estimate EPD for FSCR were assumed to be: with each methodology. Resulting predictions were     compared through the estimation of Pearson (r ) Aσ 00 u p   and Spearman’s (r ) correlations. Additionally, a   s m 0 I σ 0 Var =  m  regression coefficient of EPD obtained with the 00 I σ RRM on those obtained with the ATM was esti- mated. Analyses were performed using ASREML where A corresponded to the Wright’s numerator 3.0 (Gilmour et al., 2009) and the Animal Breeder’s relationship matrix, I and I were identity matri- m n Tool Kit (Golden et al., 1992). ces whose orders were equal to the number of mating groups and observations, respectively. σ RESULTS AND DISCUSSION 2 2 , σ , and σ were the additive, mating group, and m e residual variances, respectively. In agreement with Summary statistics for FSCR and age at AI the specifications of a maximum a posteriori pro- are presented in Table  1. Forming contemporary bit threshold model, the residual variance was (σ ) groups by combining breeding year and semen type (i.e., conventional or sexed) resulted in a total of constrained to be equal to 1. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic evaluations for heifer fertility S45 Table 1. First-service conception rate and age at first AI summary statistics Item N Average SD Minimum Maximum First-service conception rate 4,334 0.46 0.50 0 1 Age a first AI , d 4,334 422.10 21.06 347 479 AI = artificial insemination. 44 unique groups with an average size of 98.5 indi- were considerably higher at the extremes of the viduals per group. The average FSCR in our study age prediction range than in the middle. As a pos- (0.46) was slightly smaller to the range from 0.53 sible explanation for such results, it has been pre- to 0.60 reported in other studies (Minick Bormann viously reported that a common artifact of RRM et al., 2006; Peters et al., 2013). Differences among using Legendre Polynomials as their base function reports could be attributed to the heterogeneity was their tendency to inflate the genetic variances of techniques applied for estrus detection and at the beginning and the end of the covariate data AI-technician’s expertise, since both factors have a range (Schaeffer and Jamrozik, 2008). This occurs strong influence on this trait. because RRM are sensitive to changes in data dis- The heritability (h ) estimate of FSCR obtained tribution, particularly, to reductions in the number with the ATM on the underlying scale was 0.03 ± of records associated with the covariate imple- 0.02, which agrees with a previous report in Angus mented (Brügemann et  al., 2013). Figure  2 shows heifers of 0.03  ± 0.03 (Minick Bormann et  al., the distribution of FSCR records associated with 2006). Even when other reports have indicated the ages at AI of the Angus heifers from the CSU- FSCR heritability estimates ranging between 0.18 BIC. The significant reductions in the number of to 0.22, such estimates have been obtained mainly observations registered at the extremes of the data in crossbred cattle (Dearborn et  al., 1973; Peters range could explain the substantial increases in h et  al., 2013). Conversely, heritability estimates for estimates for these ages. Similar data structures the resulting intercept and linear term of the RRM have led to comparable variations in h estimates for were 0.002 ± 0.012 and 0.138 ± 0.078, respectively. traits like days open and conception rate of dairy Transforming these RRM variance estimates, a cattle when implementing random regression tech- h of 0.005  ± 0.001 for FSCR at the average age niques (Yin et al., 2012; Brügemann et al., 2013). at AI (422 d) was observed. Although the previ- Regarding genetic predictions performed with ous estimate was less than that obtained from the each methodology, EPD summary statistics are ATM, variations in h across the range of ages at presented in Table  2. Results for the mean EPD AI contemplated in this study agree with the pre- were similar between models; however, a wider vious reports for this trait (Figure 1). Nonetheless, range in prediction values was observed with the it is important to acknowledge that h estimates ATM. The smaller variation observed within the Figure 1. Estimates of heritability for first-service conception rate at different ages of insemination in Angus heifers. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Sánchez-Castro et al. S46 Figure 2. Distribution of ages at artificial insemination of the Angus heifers at the Colorado State University Beef Improvement Center. RRM prediction could be explained by the six the trait is being analyzed within each statistical times smaller h estimate obtained with this meth- method (e.g., underlying vs. linear scale). In con- odology for the average age at AI. A  similar out- clusion, application of RRM for genetic predic- come was reported by Speidel et  al. (2018) when tions of traits with singly observed phenotypes applying RRM in the genetic prediction of HPG such as FSCR was feasible. Furthermore, although in Red Angus cattle. Pearson and Spearman’s cor- RRM predictions were moderately similar to those relations among EPD obtained with the ATM obtained with ATM, the substantial re-ranking and RRM were 0.63 and 0.60, respectively. These of individuals between methodologies suggested results suggested that even when predictions were that further research is required before considering moderately similar, a considerable re-ranking of possible substitutions of threshold evaluations by individuals occurred between both methodolo- RRM for use by the beef cattle industry. gies and different individuals would be chosen in genetic selection schemes. Sánchez-Castro et  al. IMPLICATIONS (2019) reported similar correlation magnitudes (e.g., r ~ 0.59 and r ~ 0.65) when comparing pre- This study compiled evidence about the fea- p s dictions for stayability in Angus cattle obtained sibility to apply random regression techniques in with ATM and RRM. For its part, Speidel et  al. genetic evaluations of binary traits like FSCR. (2018) reported higher correlations (e.g., r  = 0.87 Despite the genetic predictions obtained with both and r  = 0.89) in a study that compared genetic pre- methodologies were not perfectly correlated, it dictions for HPG obtained with ATM and RRM; is possible that using RRM as opposed to ATM however, authors recognized that such results were in large-scale genetic predictions for FSCR pro- higher than expected given the differences in the duce interesting benefits such as the generation of statistical models employed. age-specific genetic predictions or increased accu- Regression of predictions obtained with the racy of EPD due to their flexibility to include more RRM on the ATM revealed an underestimation of data. However, regardless of the statistical method- the genetic merit for FSCR by the RRM in com- ology employed, genetic progress by direct selection parison to ATM (β   =  0.095). This difference is on FSCR is expected to be slow due to the low her- likely to be related to the different scale in which itability of this reproductive trait. Table 2.  First-service conception rate EPD summary statistics according to the statistical method implemented Methodology N Average SD Minimum Maximum ATM 13,983 −0.18 1.59 −9.40 7.20 RRM 13,983 0.02 0.24 −1.03 1.29 EPD = expected progeny differences. ATM = animal threshold model. RRM = random regression model. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic evaluations for heifer fertility S47 Horan, B., J. F. Mee, P. O’Connor, M. Rath, and P. Dillon. 2005. ACKNOWLEDGMENTS The effect of strain of Holstein-Friesian cow and feeding The authors would like to acknowledge system on postpartum ovarian function, animal produc- tion and conception rate to first service. Theriogenology the Mexican National Council for Science and 63:950–971. doi:10.1016/j.theriogenology.2004.05.014 Technology (CONACYT) for providing the Marshall,  D.  M., W.  Minqiang, and B.  A.  Freking. 1990. scholarship that allowed the first author to con- Relative calving date of first-calf heifers as related duct research at Colorado State University and to production efficiency and subsequent repro- to the staff of the John E.  Rouse Colorado State ductive performance. J. Anim. Sci. 68:1812–1817. doi:10.2527/1990.6871812x University Beef Improvement Center for assistance Minick  Bormann,  J.  M., L.  R.  Totir, S.  D.  Kachman, in collecting data used in this study. This work is R. L. Fernando, and D. E. Wilson. 2006. Pregnancy rate supported by USDA National Institute of Food and first-service conception rate in Angus heifers. J. Anim. and Agriculture Hatch project COLO0607A, acces- Sci. 84:2022–2025. doi:10.2527/jas.2005-615 sion number 1006304 and COLO0681A and acces- Misztal,  I., D.  Gianola, and J.  L.  Foulley. 1989. Computing sion number 1010007. The authors declare that aspects of a nonlinear method of sire evaluation for cat- egorical data. J. Dairy Sci. 72:1557–1568. doi:10.3168/jds. they have no conflict of interest. S0022-0302(89)79267-5 Peters,  S.  O., K.  Kizilkaya, D.  J.  Garrick, R.  L.  Fernando, LITERATURE CITED J. M. Reecy, R. L. Weaber, G. A. Silver, and M. G. Thomas. 2013. Heritability and Bayesian genome-wide associa- Brügemann, K., E. Gernand, U. U. von Borstel, and S. König. tion study of first service conception and pregnancy in 2013. Application of random regression models to infer Brangus heifers. J. Anim. Sci. 91:605–612. doi:10.2527/ the genetic background and phenotypic trajectory of jas.2012-5580 binary conception rate by alterations of temperature×hu- Sánchez-Castro,  M.  A., M.  G.  Thomas, R.  M.  Enns, and midity indices. Livest. Sci. 157:389–396. doi:10.1016/j. S.  E.  Speidel. 2019. Stability of genetic predictions for livsci.2013.08.009 stayability using random regression models that include Dearborn, D. D., R. M. Koch, L. V. Cundiff, K. E. Gregory, end points beyond 6 yr of age. Trans. Anim. Sci. 3:1678– and G.  E.  Dickerson. 1973. An analysis of reproduc- 1682. doi:10.1093/tas/txz056 tive traits in beef cattle. J. Anim. Sci. 36:1032–1040. Speidel,  S.  E., R.  M.  Enns, and B.  L.  Golden. 2018. Use of doi:10.2527/jas1973.3661032x a random regression model for the evaluation of Heifer Gilmour, A. R., B. J. Gogel, B. R. Cullis, R. Thompson, and pregnancy in Red Angus cattle. In: Proc. 11th World D.  Buttler. 2009. ASReml user guide release 3.0. VSN Congr. Genet. Appl. Livest. Prod., Auckland, New International Ltd., Hemel Hempstead, UK. Zealand. Golden, B. L., W. M. Snelling, and C. H. Mallinckrodt. 1992. Schaeffer,  L.  R., and J.  Jamrozik. 2008. Random regression Animal breeder’s toolkit user’ s guide and reference man- models: A longitudinal perspective. J. Anim. Breed. Genet. ual. Agric. Exp. Stn. Tech. Bull. LTB92-2. Colorado State 125:145–146. doi:10.1111/j.1439-0388.2008.00748.x University, Fort Collins, CO. Yin, T., B. Bapst, U. U. V. Borstel, H. Simianer, and S. König. Golden,  B.  L., S.  Weerasinghe, B.  Crook, S.  Sanders, and 2012. Genetic parameters for Gaussian and categorical D.  J.  Garrick. 2018. A Single-step hybrid marker effects traits in organic and low input dairy cattle herds based on model using random regression for stayability in hereford random regression methodology. Livest. Sci. 147:159–169. cattle. In: Proc. 11th World Congr. Genet. Appl. Livest. doi:10.1016/j.livsci.2012.04.017 Prod., Auckland, New Zealand. Translate basic science to industry innovation http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Translational Animal Science Oxford University Press

Genetic prediction for first-service conception rate in Angus heifers using a random regression model

Download PDF
5 pages

Loading next page...
 
/lp/oxford-university-press/genetic-prediction-for-first-service-conception-rate-in-angus-heifers-fPMYuS5O2c

References (30)

Copyright
© The Author(s) 2020. Published by Oxford University Press on behalf of the American Society of Animal Science.
eISSN
2573-2102
DOI
10.1093/tas/txaa094
Publisher site
See Article on Publisher Site

Abstract

Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic prediction for first-service conception rate in Angus heifers using a random regression model 1, Miguel A. Sánchez-Castro, Milton G. Thomas, R. Mark Enns, and Scott E. Speidel Department of Animal Sciences, Colorado State University, Fort Collins, CO 80523-1171 © The Author(s) 2020. Published by Oxford University Press on behalf of the American Society of Animal Science. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Transl. Anim. Sci. 2020.4:S43–S47 doi: 10.1093/tas/txaa094 predictions because edited data do not appropri- INTRODUCTION ately represent the population’s true variability First-service conception rate (FSCR) could be (Misztal et al., 1989). defined as the probability of a heifer to conceive Random regression models (RRM) represent in response to her first artificial insemination (AI) an alternative method to evaluate binary traits and maintain such pregnancy after the end of the and can incorporate data from contemporary breeding season. Economic implications of this groups with no variation (Golden et  al., 2018). trait in beef cattle heifers involve its relationship Nonetheless, since they were originally concep- with their development costs, as well as costs of tualized to analyze longitudinal traits, the effi- semen, synchronization protocols, estrus detec- cacy of RRM to evaluate traits with phenotypes tion and AI services (Horan et  al., 2005; Minick observed only once has not been deeply explored. Bormann et al., 2006). Furthermore, FSCR is also Relative to heifer fertility, these models have only related to differences in quality and value between been applied in genetic evaluations for heifer preg- AI-produced calves and natural service calves. nancy (HPG) in Red Angus cattle (Speidel et al., Additionally, heifers conceiving on their first ser - 2018). Therefore, objectives herein were to evalu- vice, calve earlier within the calving season, have ate the efficacy of a RRM genetic prediction for more chances to breed postpartum within a year, heifer FSCR by comparing the resulting EPD to and have more time to nurse and wean heavier those obtained using a traditional ATM, as well calves (Marshall et al., 1990). as to compare genetic parameters obtained with Given the binary nature of pregnancy out- each methodology in Angus cattle. comes after an AI, the method of choice to eval- uate FSCR has been an animal threshold model MATERIALS AND METHODS (ATM). However, such method has a lack of abil- The data used in the present study were ity to incorporate information of contemporary obtained from an existing database; therefore, the groups with no variation (e.g., extreme-category study was not subjected to animal care and use problem). When all observations within a categor- committee approval. ical fixed effect fall in one of the binary classes, typically, these observations are deleted from the data in order to calculate expected progeny differ- Data Collection and Editing ences (EPD). This can lead to distorted genetic Breeding and ultrasound records of 4,334 Angus heifers (progeny of 354 sires and 1,626 Corresponding author: scott.speidel@colostate.edu dams) collected from 1992 to 2019 at the Colorado Received March 13, 2020. Accepted June 22, 2020. State University Beef Improvement Center S43 Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Sánchez-Castro et al. S44 (CSU-BIC) were used for the study. Within each Additionally, FSCR was regressed on age at AI breeding year, heifers were estrus synchronized and using a linear RRM with Legendre polynomials subjected to AI only once, before they were exposed as the base function. The model in matrix form is to natural service sires approximately 3  wk after presented below: insemination. Observations for FSCR (1, successful; y = Xb + Zu + Qm + e 0, unsuccessful) were defined by fetal age obtained where y corresponded to a vector of binary obser- from ultrasound pregnancy exams performed 130 d vations of FSCR, b was a vector of unknown so- post-AI. Although FSCR in 12- to 15-mo-old heif- lutions for categorical fixed effects (contemporary ers is a singly observed phenotype, its expression is group, AI technician, and age of dam) and a linear likely to be dependent on age of onset of puberty. fixed regression of FSCR on age at AI, u corres- Consequently, age at AI was an important factor ponded to a vector of unknown solutions of animal included in all statistical models, and it was calcu- random additive genetic regression coefficients lated as the difference between an individual’s birth- (intercept and linear), m was a vector of unknown date and the date when they were subjected to AI. solutions for mating group random effects. X, Z, and Q were known incidence matrices relating ob- Statistical Analysis servations in y to both fixed and random effects, and e was the vector of unknown residual errors. Traditional EPD calculation for FSCR was per- Variances assumed for the models were: formed using a univariate BLUP threshold animal     model along with a probit link function to convert A ⊗ G 00   binary observations to an underlying normal distri-   m 0 I σ 0 Var =  m  bution. The model equation was as follows: 2 00 I σ y = Xb + Zu + Qm + e where A represented the Wright’s numerator rela- where y* corresponded to a vector of transformed tionship matrix, ⊗ was the Kronecker product, and observations of FSCR on the underlying scale, b G corresponds to a modified variance–covariance was a vector of unknown solutions for fixed ef- matrix of additive genetic random regression coef- fects that included breeding contemporary group ficients where the covariance between the intercept (defined as a combination between breeding year and the linear term was fitted to zero, given no and semen type), AI technician and age of dam, heifer had more than one observation for FSCR. 2 2 individual’s age at AI was included as a linear I , I , σ , and σ remained as described for the m e m n covariate, u corresponded to a vector of unknown previous model. For the estimation of the genetic solutions of animal random effects, and m was parameters, the entire pedigree from the CSU-BIC a vector of unknown solutions of mating group consisting of 13,983 individual animals, as well (e.g., inseminated on heat or during mass mating) as 967 and 3,699 unique sires and dams, respec- random effects. X, Z, and Q were known incidence tively, was used. The average inbreeding coefficient matrices relating observations in y* to both fixed of this pedigree was 0.009. Once the parameters and random effects, and e was the vector of un- were obtained, they were compared to each other known residual errors. For this model, variances and subsequently used to estimate EPD for FSCR were assumed to be: with each methodology. Resulting predictions were     compared through the estimation of Pearson (r ) Aσ 00 u p   and Spearman’s (r ) correlations. Additionally, a   s m 0 I σ 0 Var =  m  regression coefficient of EPD obtained with the 00 I σ RRM on those obtained with the ATM was esti- mated. Analyses were performed using ASREML where A corresponded to the Wright’s numerator 3.0 (Gilmour et al., 2009) and the Animal Breeder’s relationship matrix, I and I were identity matri- m n Tool Kit (Golden et al., 1992). ces whose orders were equal to the number of mating groups and observations, respectively. σ RESULTS AND DISCUSSION 2 2 , σ , and σ were the additive, mating group, and m e residual variances, respectively. In agreement with Summary statistics for FSCR and age at AI the specifications of a maximum a posteriori pro- are presented in Table  1. Forming contemporary bit threshold model, the residual variance was (σ ) groups by combining breeding year and semen type (i.e., conventional or sexed) resulted in a total of constrained to be equal to 1. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic evaluations for heifer fertility S45 Table 1. First-service conception rate and age at first AI summary statistics Item N Average SD Minimum Maximum First-service conception rate 4,334 0.46 0.50 0 1 Age a first AI , d 4,334 422.10 21.06 347 479 AI = artificial insemination. 44 unique groups with an average size of 98.5 indi- were considerably higher at the extremes of the viduals per group. The average FSCR in our study age prediction range than in the middle. As a pos- (0.46) was slightly smaller to the range from 0.53 sible explanation for such results, it has been pre- to 0.60 reported in other studies (Minick Bormann viously reported that a common artifact of RRM et al., 2006; Peters et al., 2013). Differences among using Legendre Polynomials as their base function reports could be attributed to the heterogeneity was their tendency to inflate the genetic variances of techniques applied for estrus detection and at the beginning and the end of the covariate data AI-technician’s expertise, since both factors have a range (Schaeffer and Jamrozik, 2008). This occurs strong influence on this trait. because RRM are sensitive to changes in data dis- The heritability (h ) estimate of FSCR obtained tribution, particularly, to reductions in the number with the ATM on the underlying scale was 0.03 ± of records associated with the covariate imple- 0.02, which agrees with a previous report in Angus mented (Brügemann et  al., 2013). Figure  2 shows heifers of 0.03  ± 0.03 (Minick Bormann et  al., the distribution of FSCR records associated with 2006). Even when other reports have indicated the ages at AI of the Angus heifers from the CSU- FSCR heritability estimates ranging between 0.18 BIC. The significant reductions in the number of to 0.22, such estimates have been obtained mainly observations registered at the extremes of the data in crossbred cattle (Dearborn et  al., 1973; Peters range could explain the substantial increases in h et  al., 2013). Conversely, heritability estimates for estimates for these ages. Similar data structures the resulting intercept and linear term of the RRM have led to comparable variations in h estimates for were 0.002 ± 0.012 and 0.138 ± 0.078, respectively. traits like days open and conception rate of dairy Transforming these RRM variance estimates, a cattle when implementing random regression tech- h of 0.005  ± 0.001 for FSCR at the average age niques (Yin et al., 2012; Brügemann et al., 2013). at AI (422 d) was observed. Although the previ- Regarding genetic predictions performed with ous estimate was less than that obtained from the each methodology, EPD summary statistics are ATM, variations in h across the range of ages at presented in Table  2. Results for the mean EPD AI contemplated in this study agree with the pre- were similar between models; however, a wider vious reports for this trait (Figure 1). Nonetheless, range in prediction values was observed with the it is important to acknowledge that h estimates ATM. The smaller variation observed within the Figure 1. Estimates of heritability for first-service conception rate at different ages of insemination in Angus heifers. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Sánchez-Castro et al. S46 Figure 2. Distribution of ages at artificial insemination of the Angus heifers at the Colorado State University Beef Improvement Center. RRM prediction could be explained by the six the trait is being analyzed within each statistical times smaller h estimate obtained with this meth- method (e.g., underlying vs. linear scale). In con- odology for the average age at AI. A  similar out- clusion, application of RRM for genetic predic- come was reported by Speidel et  al. (2018) when tions of traits with singly observed phenotypes applying RRM in the genetic prediction of HPG such as FSCR was feasible. Furthermore, although in Red Angus cattle. Pearson and Spearman’s cor- RRM predictions were moderately similar to those relations among EPD obtained with the ATM obtained with ATM, the substantial re-ranking and RRM were 0.63 and 0.60, respectively. These of individuals between methodologies suggested results suggested that even when predictions were that further research is required before considering moderately similar, a considerable re-ranking of possible substitutions of threshold evaluations by individuals occurred between both methodolo- RRM for use by the beef cattle industry. gies and different individuals would be chosen in genetic selection schemes. Sánchez-Castro et  al. IMPLICATIONS (2019) reported similar correlation magnitudes (e.g., r ~ 0.59 and r ~ 0.65) when comparing pre- This study compiled evidence about the fea- p s dictions for stayability in Angus cattle obtained sibility to apply random regression techniques in with ATM and RRM. For its part, Speidel et  al. genetic evaluations of binary traits like FSCR. (2018) reported higher correlations (e.g., r  = 0.87 Despite the genetic predictions obtained with both and r  = 0.89) in a study that compared genetic pre- methodologies were not perfectly correlated, it dictions for HPG obtained with ATM and RRM; is possible that using RRM as opposed to ATM however, authors recognized that such results were in large-scale genetic predictions for FSCR pro- higher than expected given the differences in the duce interesting benefits such as the generation of statistical models employed. age-specific genetic predictions or increased accu- Regression of predictions obtained with the racy of EPD due to their flexibility to include more RRM on the ATM revealed an underestimation of data. However, regardless of the statistical method- the genetic merit for FSCR by the RRM in com- ology employed, genetic progress by direct selection parison to ATM (β   =  0.095). This difference is on FSCR is expected to be slow due to the low her- likely to be related to the different scale in which itability of this reproductive trait. Table 2.  First-service conception rate EPD summary statistics according to the statistical method implemented Methodology N Average SD Minimum Maximum ATM 13,983 −0.18 1.59 −9.40 7.20 RRM 13,983 0.02 0.24 −1.03 1.29 EPD = expected progeny differences. ATM = animal threshold model. RRM = random regression model. Translate basic science to industry innovation Downloaded from https://academic.oup.com/tas/article/4/Supplement_1/S43/6043879 by DeepDyve user on 22 December 2020 Genetic evaluations for heifer fertility S47 Horan, B., J. F. Mee, P. O’Connor, M. Rath, and P. Dillon. 2005. ACKNOWLEDGMENTS The effect of strain of Holstein-Friesian cow and feeding The authors would like to acknowledge system on postpartum ovarian function, animal produc- tion and conception rate to first service. Theriogenology the Mexican National Council for Science and 63:950–971. doi:10.1016/j.theriogenology.2004.05.014 Technology (CONACYT) for providing the Marshall,  D.  M., W.  Minqiang, and B.  A.  Freking. 1990. scholarship that allowed the first author to con- Relative calving date of first-calf heifers as related duct research at Colorado State University and to production efficiency and subsequent repro- to the staff of the John E.  Rouse Colorado State ductive performance. J. Anim. Sci. 68:1812–1817. doi:10.2527/1990.6871812x University Beef Improvement Center for assistance Minick  Bormann,  J.  M., L.  R.  Totir, S.  D.  Kachman, in collecting data used in this study. This work is R. L. Fernando, and D. E. Wilson. 2006. Pregnancy rate supported by USDA National Institute of Food and first-service conception rate in Angus heifers. J. Anim. and Agriculture Hatch project COLO0607A, acces- Sci. 84:2022–2025. doi:10.2527/jas.2005-615 sion number 1006304 and COLO0681A and acces- Misztal,  I., D.  Gianola, and J.  L.  Foulley. 1989. Computing sion number 1010007. The authors declare that aspects of a nonlinear method of sire evaluation for cat- egorical data. J. Dairy Sci. 72:1557–1568. doi:10.3168/jds. they have no conflict of interest. S0022-0302(89)79267-5 Peters,  S.  O., K.  Kizilkaya, D.  J.  Garrick, R.  L.  Fernando, LITERATURE CITED J. M. Reecy, R. L. Weaber, G. A. Silver, and M. G. Thomas. 2013. Heritability and Bayesian genome-wide associa- Brügemann, K., E. Gernand, U. U. von Borstel, and S. König. tion study of first service conception and pregnancy in 2013. Application of random regression models to infer Brangus heifers. J. Anim. Sci. 91:605–612. doi:10.2527/ the genetic background and phenotypic trajectory of jas.2012-5580 binary conception rate by alterations of temperature×hu- Sánchez-Castro,  M.  A., M.  G.  Thomas, R.  M.  Enns, and midity indices. Livest. Sci. 157:389–396. doi:10.1016/j. S.  E.  Speidel. 2019. Stability of genetic predictions for livsci.2013.08.009 stayability using random regression models that include Dearborn, D. D., R. M. Koch, L. V. Cundiff, K. E. Gregory, end points beyond 6 yr of age. Trans. Anim. Sci. 3:1678– and G.  E.  Dickerson. 1973. An analysis of reproduc- 1682. doi:10.1093/tas/txz056 tive traits in beef cattle. J. Anim. Sci. 36:1032–1040. Speidel,  S.  E., R.  M.  Enns, and B.  L.  Golden. 2018. Use of doi:10.2527/jas1973.3661032x a random regression model for the evaluation of Heifer Gilmour, A. R., B. J. Gogel, B. R. Cullis, R. Thompson, and pregnancy in Red Angus cattle. In: Proc. 11th World D.  Buttler. 2009. ASReml user guide release 3.0. VSN Congr. Genet. Appl. Livest. Prod., Auckland, New International Ltd., Hemel Hempstead, UK. Zealand. Golden, B. L., W. M. Snelling, and C. H. Mallinckrodt. 1992. Schaeffer,  L.  R., and J.  Jamrozik. 2008. Random regression Animal breeder’s toolkit user’ s guide and reference man- models: A longitudinal perspective. J. Anim. Breed. Genet. ual. Agric. Exp. Stn. Tech. Bull. LTB92-2. Colorado State 125:145–146. doi:10.1111/j.1439-0388.2008.00748.x University, Fort Collins, CO. Yin, T., B. Bapst, U. U. V. Borstel, H. Simianer, and S. König. Golden,  B.  L., S.  Weerasinghe, B.  Crook, S.  Sanders, and 2012. Genetic parameters for Gaussian and categorical D.  J.  Garrick. 2018. A Single-step hybrid marker effects traits in organic and low input dairy cattle herds based on model using random regression for stayability in hereford random regression methodology. Livest. Sci. 147:159–169. cattle. In: Proc. 11th World Congr. Genet. Appl. Livest. doi:10.1016/j.livsci.2012.04.017 Prod., Auckland, New Zealand. Translate basic science to industry innovation

Journal

Translational Animal Science – Oxford University Press

Published: Dec 1, 2020

There are no references for this article.