Non-additive genetic effects are well known to play an important role in the phenotypic expression of complex traits, such as fertility and reproduction. In this study, a genome scan was performed using 41,640 single nucleotide polymorphism (SNP) markers to identify genomic regions associated with epistatic (additive-by-additive) effects in fertility and reproduction traits in Holstein cattle. Nine fertility and reproduction traits were analysed on 5825 and 6090 Holstein heifers and cows with phenotypes and genotypes, respectively. The Marginal Epistasis Test (MAPIT) was used to identify SNPs with significant marginal epistatic effects at a chromosome-wise 5% and 10% false discovery rate (FDR) level. The −log10(p) values were adjusted by the genomic inflation factor (λ) to correct for the potential bias on the p-values and minimize the possible effects of population stratification. After adjustments, MAPIT enabled the identification of genomic regions with significant marginal epistatic effects for heifers on BTA5 for age at first insemination, BTA3 and BTA24 for non-return rate (NRR); BTA16 and BTA28 for gestation length (GL); BTA1, BTA4 and BTA17 for stillbirth (SB). For the cow traits, MAPIT enabled the identification of regions on BTA11 for GL, BTA11 and BTA16 for SB and BTA19 for calf size (CZ). An additional approach for mapping epistasis in a genome-wide association study was also proposed, in which the genome scan was performed using estimates of epistatic values as the input pseudo-phenotypes, computed using single-trait animal models. Significant SNPs were identified at the chromosome-wise 5% and 10% FDR levels for all traits. For the heifer traits, significant regions were found on BTA7 for AFS; BTA12 for NRR; BTA14 and BTA19 for GL; BTA19 for calving ease (CE); BTA5, BTA24, BTA25 and in the X chromosome for SB; BTA23 and in the X chromosome for CZ and in the X chromosome for the number of services (NS). For the cow traits, significant regions were found on BTA29 and in the X chromosome for NRR, BTA11, BTA16 and in the X chromosome for SB, BTA2 for GL, BTA28 for CZ, BTA19 for calving to first insemination, and in the X chromosome for NS and first insemination to conception. The results suggest that the epistatic genetic effects are likely due to many loci with a small effect rather than few loci with a large effect and/or a single SNP marker alone do not capture the epistatic effects well. The genomic architecture of fertility and reproduction traits is complex, and these results should be validated in independent dairy cattle populations and using alternative statistical models.
We estimated genetic parameters for two pork production and six litter performance traits of Landrace, Large White, and Duroc pigs reared in Japan. Pork production traits were average daily gain from birth to end of performance testing and backfat thickness at end of testing (46,042 records for Landrace, 40,467 records for Large White, and 42,920 records for Duroc). Litter performance traits were number born alive, litter size at weaning (LSW), number of piglets dead during suckling (ND), survival rate of piglets during suckling (SV), total piglet weight at weaning (TWW), and average piglet weight at weaning (AWW) (27,410, 26,716, and 12,430 records for Landrace, Large White, and Duroc, respectively). ND was calculated as the difference between LSW and litter size at start of suckling (LSS). SV was calculated as LSW/LSS. AWW was calculated as TWW/LSW. Pedigree data for Landrace, Large White, and Duroc breeds contained 50,193, 44,077, and 45,336 pigs, respectively. Trait heritability was estimated via single-trait analysis and genetic correlation between two traits was estimated via two-trait analysis. When considering the linear covariate of LSS in the statistical model for LSW and TWW, for all breeds, the heritability was estimated to be 0.4–0.5 for pork production traits and below 0.2 for litter performance traits. Estimated genetic correlation between average daily gain and backfat thickness was small, ranging from 0.057 to 0.112, and those between pork production traits and litter performance traits were negligible to moderate, ranging from −0.493 to 0.487. A wide range of genetic correlation values among the litter performance traits was estimated, while that between LSW and ND could not be obtained. The results of genetic parameter estimation were affected by whether the linear covariate of LSS was included in the statistical model for LSW and TWW or not. This finding implies the necessity of carefully interpreting the results according to the choice of statistical model. Our results could give fundamental information on simultaneously improving productivity and female reproductivity for pigs.