Prediction of Genomic Breeding Values for feed efficiency and related traits in pigs

TY - ABST

T1 - Prediction of Genomic Breeding Values for feed efficiency and related traits in pigs

AU - Do, Duy Ngoc

AU - Janss, Luc

AU - Strathe, Anders Bjerring

AU - Jensen, Just

AU - Kadarmideen, Haja

PY - 2014

Y1 - 2014

N2 - Improvement of feed efficiency is essential in pig breeding and selection for reduced residual feed intake (RFI) is an option. Accuracy of genomic prediction (GP) relies on assumptions of genetic architecture of the traits. This study applied five different Bayesian Power LASSO (BPL) models with different power parameters to investigate genetic architecture of RFI, to predict genomic breeding values, and to partition genetic variances for different SNP groups. Data were 1272 Duroc pigs with both genotypic and phenotypic records for RFI as well as daily feed intake (DFI). The gene mapping confirmed previously identified significant SNPs on 30-31Mb and 60-61Mb on chromosome 1 (SSC 1) for RFI and DFI, respectively, and found several novel SNPs on SSC 14 for RFI. Estimated genomic variance varied according to the power parameters. Prediction accuracy for BLP models ranged from 0.33-0.34 and 0.31-0.32 for DFI and RFI, respectively, and approx. 1.5% higher than that of GBLUP method. However , BPL models are more biased than GBLUP method for both traits and use of different power parameters has no effect on predictive ability of the models. Partitioning of genetic variance showed that SNP groups either by position (intron, exon, downstream, upstream and 5’UTR) or by function (missense and protein-altering) have similar average explained variance per SNP, except 3’UTR SNPs which explain approx. 3 times more variance then SNPs in the other groups. This study supports use of BPL models for both GWAS and indicates their potentials for genomic prediction. Further work includes applying other GP methods for RFI and DFI as well as extending these methods to feed efficiency related traits such as feeding behaviour and body composition traits.

AB - Improvement of feed efficiency is essential in pig breeding and selection for reduced residual feed intake (RFI) is an option. Accuracy of genomic prediction (GP) relies on assumptions of genetic architecture of the traits. This study applied five different Bayesian Power LASSO (BPL) models with different power parameters to investigate genetic architecture of RFI, to predict genomic breeding values, and to partition genetic variances for different SNP groups. Data were 1272 Duroc pigs with both genotypic and phenotypic records for RFI as well as daily feed intake (DFI). The gene mapping confirmed previously identified significant SNPs on 30-31Mb and 60-61Mb on chromosome 1 (SSC 1) for RFI and DFI, respectively, and found several novel SNPs on SSC 14 for RFI. Estimated genomic variance varied according to the power parameters. Prediction accuracy for BLP models ranged from 0.33-0.34 and 0.31-0.32 for DFI and RFI, respectively, and approx. 1.5% higher than that of GBLUP method. However , BPL models are more biased than GBLUP method for both traits and use of different power parameters has no effect on predictive ability of the models. Partitioning of genetic variance showed that SNP groups either by position (intron, exon, downstream, upstream and 5’UTR) or by function (missense and protein-altering) have similar average explained variance per SNP, except 3’UTR SNPs which explain approx. 3 times more variance then SNPs in the other groups. This study supports use of BPL models for both GWAS and indicates their potentials for genomic prediction. Further work includes applying other GP methods for RFI and DFI as well as extending these methods to feed efficiency related traits such as feeding behaviour and body composition traits.

M3 - Conference abstract for conference

T2 - 65th annual meeting of the European Federation of Animal Science.

Y2 - 25 August 2014 through 29 August 2014

ER -