Genomic predictions based on haplotypes fitted as pseudo-SNP for milk production and udder type traits and SCS in French dairy goats.

The development of statistical methods aiming to improve the accuracy of genomic predictions is of utmost value for dairy goat breeding programs. In this context, the use of haplotypes, instead of individual SNP, could improve the accuracy of genomic predictions by better capturing the effect of causal variants, instead of relying solely on linkage disequilibrium with individual SNP. Haplotypes can be included in genomic evaluation models in various ways, such as fitting them as pseudo-SNP (i.e., haplotypes converted into biallelic SNP format). This can be easily incorporated in the software already available for single-step genomic predictions (ssGBLUP). Therefore, the aim of this study was to compare the predictive performances of ssGBLUP and weighted ssGBLUP (WssGBLUP) based on individual SNP or on haplotypes fitted as pseudo-SNP. Performance was compared in terms of accuracy, bias, and weights for SNP versus pseudo-SNP. Genomic predictions were performed on 5 milk production traits, 5 udder type traits, and somatic cell score (SCS). The training population was formed by 307 Alpine and 247 Saanen progeny-tested bucks, genotyped using the Illumina Goat SNP50 BeadChip (Illumina, San Diego, CA). The validation population included 205 Alpine and 146 Saanen young bucks. The accuracy of genomic predictions was evaluated in the validation population as the Pearson correlation between genomic estimated breeding values (GEBV), predicted based on various methods, and daughter deviation (DD) based on the official genetic evaluation of January 2016. Haplotype-based models were shown to improve the performance of genomic predictions for some traits. Gains in accuracy of up to +19% (0.310 to 0.368 for fat yield) in Alpine and up to +3% (0.361 to 0.373 for udder shape) in Saanen were observed with ssGBLUP. The ssGBLUP accuracies averaged across all traits and methods were equal to 0.467 (SNP) versus 0.471 (pseudo-SNP) in Alpine and 0.528 (SNP) versus 0.523 (pseudo-SNP) in Saanen. With WssGBLUP, gains in accuracy of up to 24% (0.298 to 0.370 for fat yield) in Alpine and 14% (0.431 to 0.490 for SCS) in Saanen were observed with WssGBLUP. Accuracies of WssGBLUP averaged across all traits and methods were equal to 0.455 (SNP and pseudo-SNP) in Alpine and 0.542 (SNP) versus 0.528 (pseudo-SNP) in Saanen. The average (±SD) slope of the regression of DD on GEBV for the validation animals, across all breeds, traits and scenarios, were equal to 0.82 ± 0.20 (SNP) and 0.83 ± 0.18 (pseudo-SNP) for ssGBLUP and 0.67 ± 0.16 (SNP) and 0.65 ± 0.16 (pseudo-SNP) for WssGBLUP, which suggest that haplotype-based models and ssGBLUPSNP were similarly biased. However, WssGBLUP was more biased than ssGBLUP, and its gains in accuracies were limited to milk production traits. Despite the fact that genomic predictions based on haplotypes require additional steps and time, the observed gains in GEBV predictive performance indicate that haplotype-based methods could be recommended for some traits.

Implications of SNP weighting on single-step genomic predictions for different reference population sizes.

We investigated the importance of SNP weighting in populations with 2,000 to 25,000 genotyped animals. Populations were simulated with two effective sizes (20 or 100) and three numbers of QTL (10, 50 or 500). Pedigree information was available for six generations; phenotypes were recorded for the four middle generations. Animals from the last three generations were genotyped for 45,000 SNP. Single-step genomic BLUP (ssGBLUP) and weighted ssGBLUP (WssGBLUP) were used to estimate genomic EBV using a genomic relationship matrix (G). The WssGBLUP performed better in small genotyped populations; however, any advantage for WssGBLUP was reduced or eliminated when more animals were genotyped. WssGBLUP had greater resolution for genome-wide association (GWA) as did increasing the number of genotyped animals. For few QTL, accuracy was greater for WssGBLUP than ssGBLUP; however, for many QTL, accuracy was the same for both methods. The largest genotyped set was used to assess the dimensionality of genomic information (number of effective SNP). The number of effective SNP was considerably less in weighted G than in unweighted G. Once the number of independent SNP is well represented in the genotyped population, the impact of SNP weighting becomes less important.