Volume 14, Issue 4 (12-2023)
Res Anim Prod 2023, 14(4): 102-113 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Ghafouri-Kesbi F. (2023). Assessing the Performance of Ridge Regression Method-6 in Genomic Evaluation of Discrete Threshold Traits with Additive and Dominance Genetic Architecture. Res Anim Prod. 14(4), 102-113. doi:10.61186/rap.14.42.102
URL: http://rap.sanru.ac.ir/article-1-1371-en.html
URL: http://rap.sanru.ac.ir/article-1-1371-en.html
Department of Animal Science, Faculty of Agriculture, Bu-Ali Sina University, Hamedan, Iran
Abstract: (1289 Views)
Extended Abstract
Background: One of the important issues in genomic selection is estimating the effect of markers. In recent years, various methods have been proposed to estimate marker effects, each providing genomic breeding values with different levels of accuracy. One commonly used method in genomic evaluation is ridge regression (rrBLUP), which has been applied in various studies to predict genomic breeding values. Recently, modifications to the parameters of the rrBLUP method have led to the development of a variant known as Ridge Regression method 6 (RR-m6) to address regression problems. However, this method has not yet been utilized in the genomic evaluation of threshold traits with additive and dominant genetic architectures, and its performance in this context remains unknown. Therefore, this research aimed to compare the prediction performance of the RR-m6 method with other common methods of genomic evaluation.
Methods: A genome consisting of 10 chromosomes, each containing 1,000 bi-allelic single nucleotide polymorphisms (SNPs), was simulated at a heritability level of 0.5. All quantitative trait loci (QTLs) were assigned additive genetic effects, modeled by a gamma distribution. Two scenarios for the number of QTLs were considered: 1% and 10% of the total number of SNPs (100 and 1,000 QTLs, respectively). Additionally, in different scenarios, 0%, 50%, and 100% of QTLs were assigned dominance effects. Genomic breeding values were estimated using RR-m6, rrBLUP, GBLUP, BayesA, regression tree (RT), Random Forest (RF), and boosting. Indicators from the LR method, including prediction accuracy, bias, and dispersion (inflation) of genomic breeding values, were used to analyze the breeding values estimated by different methods. Furthermore, the computing time and memory required to execute the codes of each method on the CPU were calculated.
Results: The results indicated that using a purely additive model when genetic dominance effects contributed to the phenotypic variation of the trait led to decreased accuracy and increased bias and dispersion of the genomic breeding values. The extent of these effects depended on the number of QTLs exhibiting dominance. Compared to other methods, RR-m6 demonstrated excellent performance, as it consistently yielded the highest accuracy and the lowest bias and dispersion across all studied scenarios, although differences were not significant in most cases when compared to BayesA. In terms of computational speed, the RR-m6 method was the fastest and required less memory than other methods for analysis.
Conclusion: The findings suggest that the RR-m6 method predicts genomic breeding values with high accuracy and is efficient in terms of computing time and memory requirements, making it a viable option for the genomic evaluation of threshold traits.
Methods: A genome consisting of 10 chromosomes, each containing 1,000 bi-allelic single nucleotide polymorphisms (SNPs), was simulated at a heritability level of 0.5. All quantitative trait loci (QTLs) were assigned additive genetic effects, modeled by a gamma distribution. Two scenarios for the number of QTLs were considered: 1% and 10% of the total number of SNPs (100 and 1,000 QTLs, respectively). Additionally, in different scenarios, 0%, 50%, and 100% of QTLs were assigned dominance effects. Genomic breeding values were estimated using RR-m6, rrBLUP, GBLUP, BayesA, regression tree (RT), Random Forest (RF), and boosting. Indicators from the LR method, including prediction accuracy, bias, and dispersion (inflation) of genomic breeding values, were used to analyze the breeding values estimated by different methods. Furthermore, the computing time and memory required to execute the codes of each method on the CPU were calculated.
Results: The results indicated that using a purely additive model when genetic dominance effects contributed to the phenotypic variation of the trait led to decreased accuracy and increased bias and dispersion of the genomic breeding values. The extent of these effects depended on the number of QTLs exhibiting dominance. Compared to other methods, RR-m6 demonstrated excellent performance, as it consistently yielded the highest accuracy and the lowest bias and dispersion across all studied scenarios, although differences were not significant in most cases when compared to BayesA. In terms of computational speed, the RR-m6 method was the fastest and required less memory than other methods for analysis.
Conclusion: The findings suggest that the RR-m6 method predicts genomic breeding values with high accuracy and is efficient in terms of computing time and memory requirements, making it a viable option for the genomic evaluation of threshold traits.
Type of Study: Research |
Subject:
ژنتیک و اصلاح نژاد دام
Received: 2023/03/23 | Accepted: 2023/07/31
Received: 2023/03/23 | Accepted: 2023/07/31
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
