D. Biagiotti, F.F. Silva, L.S. Sena, F.C.B. Sousa, G.V. Santos, L.A.S. Figuereido Filho, J.S. Moura, A.S. Júnior, N.P.S. Santos, J.L.R. Sarmento
Published: February 28, 2024
Genet. Mol. Res. 23(1): GMR19229
DOI: https://doi.org/10.4238/gmr19229
Cite this Article:
D. Biagiotti, F.F. Silva, L.S. Sena, F.C.B. Sousa, G.V. Santos, L.A.S.Figuereido Filho, J.S. Moura, A.S. Júnior, N.P.S. Santos, J.L.R. Sarmento (2024). Comparison of Bayesian methods for genomic prediction for resistance to worm infection in meat sheep. Genet. Mol. Res. 23(1): GMR19229. https://doi.org/10.4238/gmr19229
About the Authors
D. Biagiotti, F.F. Silva, L.S. Sena, F.C.B. Sousa, G.V. Santos, L.A.S. Figuereido Filho, J.S. Moura, A.S. Júnior, N.P.S. Santos, J.L.R. Sarmento
Corresponding Author
F.C.B. Sousa
Email: fabiana.cristina1@hotmail.com
ABSTRACT
The objective of this study was to compare the performance of various Bayesian methods for genomic prediction of characteristics indicative of resistance to worms in Santa Inês sheep. Phenotypic records were collected from 271 animals belonging to six breeders in the states of Piauí and Maranhão, five from commercial herds and one from the conservation center of the research unit of Embrapa Central-North in Campo Maior, Piauí. Phenotypic records were collected from 271 animals of the Santa Inês sheep breed belonging to six breeders located Sub-região Centro-Norte do Brasil in the states of Piauí and Maranhão, five coming from commercial herds and one from herds at the conservation center of the federal research unit Embrapa Central-North (Campo Maior, Piauí). Phenotypic records of Strongyloides sp. eggs in the feces (STE), log transformed fecal egg count (LFEC), FAMACHA score of the ocular conjunctiva (FAM), and body condition score (BCS) were used. All animals were genotyped using the OvineSNP50 BeadChip (Illumina Inc.). After quality control, 44,548 SNP markers and all the DNA samples remained for further analyses. The following models were tested to estimate the effects of markers: Bayesian ridge regression (BRR), Bayes A, Bayes B, Bayes C, and Bayesian least absolute shrinkage and selection operator (BLASSO). The correlations between Genomic Breeding Values (GEBVs) and observed breeding values were calculated and used as indicators of prediction accuracy of the genomic models. We also calculated the accuracy of the pedigree-based BLUP for comparison. Variance components, heritability, and GEBvs were calculated using the BRR model. The BRR model was considered best, due to its prediction accuracy and because this model used the lowest number of parameters. Accuracy gains higher than 60% were obtained using Bayesian models in comparison to the pedigree-based model. The heritability estimates were 0.560, 0.242, 0.253, and 0.244 for STE, LFEC, FAM, and BCS, respectively. The Bayesian models showed similar performance for prediction accuracy and significantly outperformed the pedigree-based model. The BRR model is the most recommended for genomic selection for the traits evaluated.
Key words: Cross-validation, Estimated breeding value, Fecal egg count, Heritability