| Full text | |
| Author(s): |
Boison, S. A.
;
Neves, H. H. R.
;
Perez O'Brien, A. M.
;
Utsunomiya, Y. T.
;
Carvalheiro, R.
;
da Silva, M. V. G. B.
;
Soelkner, J.
;
Garcia, J. F.
Total Authors: 8
|
| Document type: | Journal article |
| Source: | LIVESTOCK SCIENCE; v. 166, p. 14-pg., 2014-08-01. |
| Abstract | |
This study aimed at imputing non(un)-genotyped sires using a stepwise imputation approach that combines identity by descent (IBD) detection methods with other imputation algorithms. We also studied the effect of using actual or imputed genotypes of non-genotyped sires in estimating genomic relationships. Simulations and real data were used for the analysis. Fifty sire families were simulated and 23 sire families were derived from 995 Brazilian Nellore cattle genotyped with Illumina (R) Bovine HD (777,962 SNPs) SNP Chip. Un-genotyped sires were imputed using genotype information from progeny (5 or 10); progeny and grand offspring; a combination of progeny, mates of genotyped progeny and grand offspring; and the entire genotyped population. Stepwise imputation was done with an IBD detection method that uses simple inheritance rules (MERLIN) as a First step and subsequently with FImpute, MaCH or BEAGLE as the second step to infer genotypes that were not imputed unambiguously by MERLIN. The stepwise imputation procedure was compared to an approach that ignores the first step (MERLIN) but uses only prior pedigree information to impute non-genotyped animals. Imputation accuracy was assessed as percent of correctly called genotypes and the correlation between imputed and actual genotypes (in brackets). With real data, imputation accuracy ranged from 81.6% (0.856) to 97.4% (0.981) depending on the amount of genotyped information considered for the first step (MERLIN) and imputation algorithms used for the second step. Greater accuracies of imputing non-genotyped sires were obtained when the stepwise imputation approach was used with 10 genotyped offspring as the first step. The stepwise approach resulted in an increase of 1.2% (5 offsprings) and 4.7% (10 offsprings) in imputation accuracy. MaCH was more accurate in the second step, followed by FImpute then BEAGLE. Similar trends in imputation accuracy were observed for the simulated population. Generally, imputed genotypes were successfully used to estimate genomic relationships among close relatives but considerable bias was observed for true pairwise relationships of zero. In conclusion, high imputation accuracies can be achieved for non-genotyped animals when genotype information of 5 or 10 direct progeny is available for imputation. Performing preliminary IBD analysis and using non-ambiguous genotypes from that analysis in conventional imputation increased the imputation accuracy considerably. (C) 2014 Elsevier B.V. All rights reserved. (AU) | |
| FAPESP's process: | 11/16643-2 - Genome-wide association study for detection of chromosomal regions related to scrotal circumference in Nellore cattle |
| Grantee: | Yuri Tani Utsunomiya |
| Support Opportunities: | Scholarships in Brazil - Master |
| FAPESP's process: | 10/52030-2 - Genome wide association study for reproductive traits in zebu bulls (Bos indicus) using high-density SNP chip |
| Grantee: | José Fernando Garcia |
| Support Opportunities: | Regular Research Grants |