東華大學圖書館 |

Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle./
Author:	Sanchez Castro, Miguel Angel.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
Description:	284 p.
Notes:	Source: Dissertations Abstracts International, Volume: 83-01, Section: B.
Contained By:	Dissertations Abstracts International83-01B.
Subject:	Animal sciences. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28323350
ISBN:	9798516077135

Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle.
Sanchez Castro, Miguel Angel.

Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 284 p.

Source: Dissertations Abstracts International, Volume: 83-01, Section: B.

Thesis (Ph.D.)--Colorado State University, 2021.

This item must not be sold to any third party vendors.

Female fertility is one of the most important economic drivers of cow-calf operations, however, the achievement of genetic improvement for female fertility traits is challenging due to the biological complexity of reproductive performance and the difficulties related to its statistical modeling. Among the traits relevant to beef cattle breeding practices, those related to key fertility events such as conception and calving are binary in nature. Traditional evaluations of binary traits involve the use of threshold models (TM) that convert categorical phenotypes to an underlying normally distributed range of genotypic values known as liabilities. Despite the successful influence that TM have had on genetic trends of categorically evaluated traits within livestock species, these models also have drawbacks. Among the most important weaknesses are their susceptibility to the extreme category problem (ECP) and their lack of flexibility to incorporate genomic information differently than using genomic relationship matrices whose inverse is difficult to obtain when the number of genotyped animals is high. These deficiencies of TM prevent them from comprehensively utilize all available phenotypic data and preclude their utilization in single-step genomic prediction methodologies based on marker effects models.Contrastingly, random regression models (RRM) have emerged as an attractive alternative for the evaluation of binary fertility traits in cattle due to their ability to overcome ECP problems and utilize all available information to produce more accurate results in comparison to TM. Furthermore, these models are flexible enough to accommodate any of the single-step genomic evaluation procedures that have been developed. Consequently, their extension to genomic evaluation procedures that avoid the need of inverting dense genomic relationship matrices such as the recently developed super-hybrid marker effects models, represents a novel approach to evaluate binary fertility traits in beef cattle. Traits like heifer pregnancy (HPG), first-service conception rate (FSCR) and stayability (STAY) constitute important elements of the breeding objective of beef cattle producers, therefore, they were selected as the traits to evaluate in this study. All the reproductive data utilized in this investigation was produced by the Angus cattle population of the John E. Rouse Colorado State University Beef Improvement Center (CSU-BIC). In general, this dissertation was divided in three different studies according to the physiological status of the females producing the phenotypic record (e.g., heifer vs. multiparous cows) and the number of instances that such phenotype can be recorded on the life of the animals (non-longitudinal vs. longitudinal).The first study involved the comparison of expected progeny differences (EPD) and genetic parameters obtained with TM and RRM in genetic evaluations of singly-observed heifer dichotomous fertility traits such as HPG and FSCR. Breeding and pregnancy ultrasound records of 4,334 Angus heifers (progeny of 354 sires and 1,626 dams) collected between 1992 to 2019 at the CSU-BIC were utilized. Observations for HPG and FSCR (1, successful; 0, unsuccessful) were defined by fetal age at pregnancy diagnosis performed approximately 130 d post artificial insemination (AI). Traditional evaluations for both traits were performed using univariate TM, whereas alternative evaluations were performed by regressing HPG (or FSCR) on age at first exposure (AFE) using linear RRM with Legendre Polynomials as the base function. Heritability (h2) estimates were 0.04 and 0.03 for HPG and FSCR using TM; whereas RRM derived h2 estimates were 0.02 and 0.006 for the average AFE for HPG and FSCR, respectively. Pearson and rank correlations between EPD obtained with each methodology were 0.97 and 0.96 for HPG, while for FSCR were 0.75 and 0.72, respectively. Regression coefficients from RRM predictions on those obtained with TM were 0.27 and 0.15 for HPG and FSCR, respectively. Differences in mean accuracies of prediction calculated at the average AFE were minimal between methodologies; however, RRM produced consistently higher accuracies than TM especially when considering young selection candidates. These results suggested that RRM genetic predictions for singly-observed fertility traits in beef heifers were feasible. More importantly, moderate to strong degrees of concordance were found between predictions obtained with both methodologies for both traits, implying that RRM could substitute for TM in genetic evaluations of heifer binary fertility traits.The second study focused on the comparison of EPD and genetic parameters yielded by TM and RRM in genetic evaluations for longitudinal binary fertility traits such as STAY and FSCR in multiparous Angus cows. Calving performance data, as well as, breeding and reproductive ultrasound records of Angus cows collected between 1990 to 2019 at the CSU-BIC were used for the study. Ten STAY endpoints defined as whether a cow calved at age 3, 4, and up to 12 yr given she calved as a 2-yr-old were assigned observations (1, successful; 0, unsuccessful). Similarly, ten FSCR age specific observations were assigned depending on the age of exposure of the females (ages ranged from 2 to 11 yr) and were defined by fetal age at pregnancy inspections performed approximately 130 d post-AI. Traditional evaluation for STAY was performed using a TM that only considered the success/failure of females reaching the age of 6 (STAY06), since this age is considered as the financial break-even point for cows within the beef industry. Conversely, given there is no specific age of interest for a multiparous cow to conceive in response to her first AI, the traditional evaluation for FSCR was performed using a repeatability TM. Alternative evaluations for both traits were performed by regressing each trait on its corresponding age specific endpoints using univariate linear RRM with Legendre Polynomials as the base function. Heritability (h2) estimates obtained for STAY were 0.10 and 0.04 for the TM and the RRM, respectively. In the case of FSCR, age was not a significant longitudinal descriptor for the trait; however, only with documentation purposes, h2 estimates were reported. For the TM the h2 estimate was 0.03 whereas for the RRM, heritabilities ranged between 0.02 to 0.05 for all the ages at exposure considered in the model. Pearson (rp) and Spearman's (rs) correlations between EPD obtained with each method for STAY were 0.84 and 0.86. For FSCR, correlations were calculated between the EPD obtained with the repeatability TM and each one of the age-specific EPD obtained with the RRM; therefore, results for the rp ranged between 0.70 to 0.99; whereas results for rs ranged between 0.69 to 0.99, depending on the age of exposure considered in the RRM. Although mean accuracies of prediction were higher using RRM than using TM for both traits, increments were much more relevant for STAY than for FSCR. The strong degrees of concordance found between predictions obtained with both methodologies for STAY, suggests that RRM could effectively substitute TM in genetic evaluations of this trait. For FSCR, no improvements were achieved by evaluating the trait using RRM, mainly due to the lack of influence that age had on the ability of cows to conceive in response to their first AI at any age.Finally, the third study had as objectives 1) to explore the feasibility of implementing single-step random regression super-hybrid models (ssRR-SHM) for the genomic evaluation of HPG, FSCR and STAY; 2) to assess the impact of differing data structures in the resulting genomic predictions of ssRR-SHM for all traits; 3) to identify quantitative trait loci (QTL) associated with the binary fertility traits contemplated in this dissertation.

ISBN: 9798516077135Subjects--Topical Terms:

3174829
Animal sciences.
Subjects--Index Terms:

Beef cattle

Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle.
LDR:11253nmm a2200397 4500 001 2351394
005 20221107085631.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798516077135
035 $a (MiAaPQ)AAI28323350
035 $a AAI28323350
040 $a MiAaPQ $c MiAaPQ
100 1 $a Sanchez Castro, Miguel Angel. $3 3690959
245 1 0 $a Random Regression Models and Their Impact in the Genetic Evaluation of Binary Fertility Traits in Beef Cattle.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 284 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-01, Section: B.
500 $a Advisor: Speidel, Scott E.;Thomas, Milton G.
502 $a Thesis (Ph.D.)--Colorado State University, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Female fertility is one of the most important economic drivers of cow-calf operations, however, the achievement of genetic improvement for female fertility traits is challenging due to the biological complexity of reproductive performance and the difficulties related to its statistical modeling. Among the traits relevant to beef cattle breeding practices, those related to key fertility events such as conception and calving are binary in nature. Traditional evaluations of binary traits involve the use of threshold models (TM) that convert categorical phenotypes to an underlying normally distributed range of genotypic values known as liabilities. Despite the successful influence that TM have had on genetic trends of categorically evaluated traits within livestock species, these models also have drawbacks. Among the most important weaknesses are their susceptibility to the extreme category problem (ECP) and their lack of flexibility to incorporate genomic information differently than using genomic relationship matrices whose inverse is difficult to obtain when the number of genotyped animals is high. These deficiencies of TM prevent them from comprehensively utilize all available phenotypic data and preclude their utilization in single-step genomic prediction methodologies based on marker effects models.Contrastingly, random regression models (RRM) have emerged as an attractive alternative for the evaluation of binary fertility traits in cattle due to their ability to overcome ECP problems and utilize all available information to produce more accurate results in comparison to TM. Furthermore, these models are flexible enough to accommodate any of the single-step genomic evaluation procedures that have been developed. Consequently, their extension to genomic evaluation procedures that avoid the need of inverting dense genomic relationship matrices such as the recently developed super-hybrid marker effects models, represents a novel approach to evaluate binary fertility traits in beef cattle. Traits like heifer pregnancy (HPG), first-service conception rate (FSCR) and stayability (STAY) constitute important elements of the breeding objective of beef cattle producers, therefore, they were selected as the traits to evaluate in this study. All the reproductive data utilized in this investigation was produced by the Angus cattle population of the John E. Rouse Colorado State University Beef Improvement Center (CSU-BIC). In general, this dissertation was divided in three different studies according to the physiological status of the females producing the phenotypic record (e.g., heifer vs. multiparous cows) and the number of instances that such phenotype can be recorded on the life of the animals (non-longitudinal vs. longitudinal).The first study involved the comparison of expected progeny differences (EPD) and genetic parameters obtained with TM and RRM in genetic evaluations of singly-observed heifer dichotomous fertility traits such as HPG and FSCR. Breeding and pregnancy ultrasound records of 4,334 Angus heifers (progeny of 354 sires and 1,626 dams) collected between 1992 to 2019 at the CSU-BIC were utilized. Observations for HPG and FSCR (1, successful; 0, unsuccessful) were defined by fetal age at pregnancy diagnosis performed approximately 130 d post artificial insemination (AI). Traditional evaluations for both traits were performed using univariate TM, whereas alternative evaluations were performed by regressing HPG (or FSCR) on age at first exposure (AFE) using linear RRM with Legendre Polynomials as the base function. Heritability (h2) estimates were 0.04 and 0.03 for HPG and FSCR using TM; whereas RRM derived h2 estimates were 0.02 and 0.006 for the average AFE for HPG and FSCR, respectively. Pearson and rank correlations between EPD obtained with each methodology were 0.97 and 0.96 for HPG, while for FSCR were 0.75 and 0.72, respectively. Regression coefficients from RRM predictions on those obtained with TM were 0.27 and 0.15 for HPG and FSCR, respectively. Differences in mean accuracies of prediction calculated at the average AFE were minimal between methodologies; however, RRM produced consistently higher accuracies than TM especially when considering young selection candidates. These results suggested that RRM genetic predictions for singly-observed fertility traits in beef heifers were feasible. More importantly, moderate to strong degrees of concordance were found between predictions obtained with both methodologies for both traits, implying that RRM could substitute for TM in genetic evaluations of heifer binary fertility traits.The second study focused on the comparison of EPD and genetic parameters yielded by TM and RRM in genetic evaluations for longitudinal binary fertility traits such as STAY and FSCR in multiparous Angus cows. Calving performance data, as well as, breeding and reproductive ultrasound records of Angus cows collected between 1990 to 2019 at the CSU-BIC were used for the study. Ten STAY endpoints defined as whether a cow calved at age 3, 4, and up to 12 yr given she calved as a 2-yr-old were assigned observations (1, successful; 0, unsuccessful). Similarly, ten FSCR age specific observations were assigned depending on the age of exposure of the females (ages ranged from 2 to 11 yr) and were defined by fetal age at pregnancy inspections performed approximately 130 d post-AI. Traditional evaluation for STAY was performed using a TM that only considered the success/failure of females reaching the age of 6 (STAY06), since this age is considered as the financial break-even point for cows within the beef industry. Conversely, given there is no specific age of interest for a multiparous cow to conceive in response to her first AI, the traditional evaluation for FSCR was performed using a repeatability TM. Alternative evaluations for both traits were performed by regressing each trait on its corresponding age specific endpoints using univariate linear RRM with Legendre Polynomials as the base function. Heritability (h2) estimates obtained for STAY were 0.10 and 0.04 for the TM and the RRM, respectively. In the case of FSCR, age was not a significant longitudinal descriptor for the trait; however, only with documentation purposes, h2 estimates were reported. For the TM the h2 estimate was 0.03 whereas for the RRM, heritabilities ranged between 0.02 to 0.05 for all the ages at exposure considered in the model. Pearson (rp) and Spearman's (rs) correlations between EPD obtained with each method for STAY were 0.84 and 0.86. For FSCR, correlations were calculated between the EPD obtained with the repeatability TM and each one of the age-specific EPD obtained with the RRM; therefore, results for the rp ranged between 0.70 to 0.99; whereas results for rs ranged between 0.69 to 0.99, depending on the age of exposure considered in the RRM. Although mean accuracies of prediction were higher using RRM than using TM for both traits, increments were much more relevant for STAY than for FSCR. The strong degrees of concordance found between predictions obtained with both methodologies for STAY, suggests that RRM could effectively substitute TM in genetic evaluations of this trait. For FSCR, no improvements were achieved by evaluating the trait using RRM, mainly due to the lack of influence that age had on the ability of cows to conceive in response to their first AI at any age.Finally, the third study had as objectives 1) to explore the feasibility of implementing single-step random regression super-hybrid models (ssRR-SHM) for the genomic evaluation of HPG, FSCR and STAY; 2) to assess the impact of differing data structures in the resulting genomic predictions of ssRR-SHM for all traits; 3) to identify quantitative trait loci (QTL) associated with the binary fertility traits contemplated in this dissertation.
520 $a Two types of genetic evaluations were implemented for each trait, the first type was a pedigree-based RRM that utilized Legendre polynomials as the base function in where the phenotype of interest was regressed on an appropriate age covariate. The second evaluation type was a ssRR-SHM that also used Legendre polynomials as the base function and regressed observations of the trait of interest on its appropriate age covariate, but that included random effects of marker and extra polygenic effects. Within each trait, four different data structure scenarios were created depending on the phenotypic performance of the genotyped and non-genotyped subsets of animals. The behavior of the genomic predictions was assessed through the calculation of Pearson and Spearman's correlations and the estimation of the regression coefficients of EPD obtained with the ssRR-SHM on those obtained with their corresponding pedigree-based RRM.Results of this study indicated that the implementation of ssRR-SHM for the genomic evaluation of singly-observed binary fertility traits like HPG and FSCR, as well as for the evaluation of a longitudinally recorded binary trait such as STAY was feasible. Nonetheless, an overestimation of genomic predictions occurred with these models when phenotypic records of pre-selected genotyped animals were included in the evaluation. Additionally, inaccurate imputation of genotypes for non-genotyped animals also impacted resulting genomic predictions, although this issue was restricted to this subgroup of animals only. In all cases, the removal of phenotypic records from preselected animals and the maintenance of closely related individuals in the pedigree ameliorated problems associated to overestimation of genomic predictions and improved correlations among genomically-enhanced and pedigree-based EPD for all traits. Regarding GWAS analyzes, the application of ssRR-SHM identified single nucleotide polymorphisms that resulted located either within or relatively close to genes that have been previously associated with important reproductive processes and fertility traits in cattle.
590 $a School code: 0053.
650 4 $a Animal sciences. $3 3174829
650 4 $a Genetics. $3 530508
650 4 $a Agriculture. $3 518588
653 $a Beef cattle
653 $a Binary
653 $a Fertility
653 $a Random regression model
653 $a Single-step genomic evaluation
653 $a Threshold model
690 $a 0475
690 $a 0369
690 $a 0473
710 2 $a Colorado State University. $b Animal Sciences. $3 3174842
773 0 $t Dissertations Abstracts International $g 83-01B.
790 $a 0053
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28323350