پژوهشهای علوم دامی ایران (Dec 2023)
Comparison of Different Genomic Relationship Matrices for Multibreed Weighted Single Step GWAS
Abstract
Introduction: Genomic best linear unbiased prediction (GBLUP) and Bayesian methods are used for genomic selection (GS) and genome wide association study (GWAS) in animal and plant breeding. The main objective of GWAS is detection of quantitative trait loci (QTL) that is affecting trait. The GBLUP assumes equal variance for all markers but the Bayesian methods assumes specific variance for each marker. When trait are affecting by major QTLs, Bayesian methods have the benefit of marker selection. In weighted GBLUP (WGBLUP), disparate variance of marker specific weights for weighting of all markers are used. If only a deduction of animals is genotyped, single-step WGBLUP (WssGBLUP) can be used. The weighting factors were calculated using marker effect derived Bayesian methods or iteratively based on single step marker effect. In multibreed genomic evaluation, there are several specific genetic structures into a genomic relationship matrix (G). The block wise genomic relationship matrix (BG) consist of several specific relationship blocks for each and pair breeds. BG can more accurately calculated relationships among animals than G matrix in multibreed genomic evaluations. The aim of this study is comparison G and BG and weighted BG (WBG) in weighted single step GWAS.Materials and Methods:To conduct our study, we initially simulated two distinct populations, labeled as A and B, utilizing the QMsim software. The simulation involved the creation of two chromosomes, each spanning a length of two morgans. Within each chromosome, we simulated 2500 single nucleotide polymorphisms (SNPs). Subsequently, four traits were simulated, each possessing heritabilities of 0.05 and 0.3, along with varying numbers of quantitative trait loci (QTLs) set at 50 and 500. Following the simulation, we calculated the genetic value (G), the breeding value given by markers (BG), and the weighted breeding value given by markers (WBG) using SNP genotypes for all animals in the study. This comprehensive approach allowed us to evaluate and analyze the genetic and breeding values associated with the simulated traits across the populations. Genomic relationship matrices were used for single step GWAS (SSGWAS) analysis for each trait. 10 iterations was considered for single step SNP effect analyses. Moreover, the SNP effects were obtained by BayesB approach. BayesB effects was used for calculated weighting factors in WBG. Accuracies of methods and number of identified SNPs with explained genetic variance higher than 1% were reported. Results and Discussion: using G and BG and WBG in SSGWAS led to identify 14, 16 and 21 SNP with higher than one percentage variance explained, respectively. Moreover, convergent accuracies of WssGBLUP using G and BG and WBG were 0.36, 0.39 and 0.43, respectively. WssGBLUP using WBG could be converged faster than using G and BG. Furthermore, accuracy of WssGBLUP using WBG was significantly more than using G and BG. Multibreed GWAS is led to increase power of model because phenotypic information is severely increase. In multibreed GWAS, relationships among breeds usually are rare or zero but there are several locations among breeds that shared among them and should use those for genomic relationship calculation. In WBG and BG could be accurately calculate pair breeds genomic relationships using sharing pair breeds genomic locations. Principal component analyses showed that WBG was let to strongly increase genomic relationship among animals that is led to improved power of WssGWAS. Conclusion: According to recent studies, multibreed genomic evaluation with the WssGBLUP can improve the accuracy of multibreed genomic evaluation, and the results of our study showed that for multibreed genomic evaluation and WssGWAS with the WssGBLUP , instead of the genomic relationship matrix (G), BG or WBG genomic relationship matrices are the better to use.
Keywords
