Estimates of the actual relationship between half-sibs in a pig population using two contrasting approaches

MUNILLA, S.; FORNERIS, N.S.; RANEY, N.E.; GARCÍA BACCINO, C.A.; VITEZICA, Z.G.; BATES, R.O.; CANTET, R.J.C.; LEGARRA, A.; ERNST, C.W.; STEIBEL, J.P.

Tandil

Congreso; 39° Congreso Argentino de Producción Animal ? Asociación Argentina de Producción Animal.; 2016

Asociación Argentina de Producción Animal

IntroductionGenomic relationships based on markers capture the actual instead of the expected (based on pedigree) proportion of genome shared identical by descent (IBD). Several methods exist to estimate genomic relationships. They differ in terms of the statistical model used and the assumptions made. Here, we implement two contrasting approaches to estimate the actual relationship between genotyped half-sibs from a pig population. The first method is based on pedigree and marker information (Elsen et al, 2009) and the second one only considers marker data (VanRaden, 2008). We also assess the methods? sensitivity to changes in the quantities of molecular information.Material and MethodsGenotypes from an outbred pig experimental population raised at the Michigan State University Swine Teaching and Research Farm, East Lansing, Michigan were used. This population comprised 411 genotyped individuals distributed in three generations: 19 from F0, 56 from F1, and 336 from F2. All in all, 6,704 F2 pairs of half-sibs were available for the study. Only the autosomal genome was considered and two minor allele frequencies (MAF) thresholds were set to generate two genomic data sets with different SNP density. Initially we excluded the variants with a MAF below 1%, giving as a result a genomic data set (DMAF_0.01) containing 39,122 markers for each animal. In a more extreme case, a MAF threshold of 20% was set to create a second data set (DMAF_0.20) with 25,957 markers per animal. Actual relationships between half-sibs were estimated by two approaches. First, we used the method proposed by Elsen et al (2009) implemented in the QTLMap software (Filangi et al, 2010). In the second approach we used the first method presented by VanRaden (2008) using the allele frequencies from the F0 generation. Results and DiscussionFigure 1 shows the differences in mean and variance of the estimated actual relationships between methods. For Elsen et al?s method the mean ranged from 0.2513 (DMAF_0.20) to 0.2516 (DMAF_0.01), in close agreement with the theoretical expectation (0.25). VanRaden?s method displayed a significantly higher mean, ranging from 0.2685 (DMAF_0.20) to 0.2713 (DMAF_0.01). This is mainly due to the pig population structure considered in this study and its effect over the calculated allele frequencies, proving the sensitivity of the method to the population frequencies. According to the formula of Hill (1993), the expected standard deviation (SD) was 0.0373. The method of Elsen et al produced the closest SD ranging from 0.0287 (DMAF_0.20) to 0.0290 (DMAF_0.01), whereas the method of VanRaden displayed larger values (0.0918 for DMAF_0.20 and 0.0927 for DMAF_0.01). Figure 2 shows the regression of the estimates produced by each method using DMAF_0.20 on the same estimates from DMAF_0.01. There is a remarkably good agreement between both scenarios for both methods, with a correlation of approximately 0.99, showing that both methods are robust when using different MAF thresholds.ConclusionsBased on the mean and variance of the theoretical distribution of IBD genome sharing, our results showed that the method by Elsen et al (based on pedigree and marker data) performed better. The estimates were close to the theoretical values and the method was robust to changes in the MAF threshold. ReferencesELSEN, J.-M., et al. 2009. Genet. Sel. Evol., 41, 50.FILANGI, O. et al 2010 Proceedings of World Congress on Genetics Applied to Livestock Production. 1?6. Hill, W. G. 1993. Heredity. 71: 652-653.VANRADEN, P.M. 2008. J. Dairy Sci. 91:4414?4423.