GWAS & Selection signatures

Single-step GWAS with APY and its application to a large pig population

---

Abstract

Using many genotyped individuals to increase the power of genome-wide association studies (GWAS) to identify QTL comes with computational challenges. In this study, we used large-scale pig populations to assess the performance of the method for single-step GWAS (ssGWAS) that uses the algorithm for proven and young (APY) with an approximated prediction error (co)variance (PEV) matrix. The approach avoids inverting the left-hand side of the mixed model equations to compute SNP p-values, allowing the inclusion of many genotyped animals in the analysis without imposing computational constraints. Data included average daily gain (ADG) and backfat thickness (BFT) from purebred maternal lines A and B (n = 485,891) and terminal lines C and D (n = 404,351) from PIC, as well as wean-to-finish mortality (MOR; 1 = alive, 2 = dead) on crossbred progeny from lines C and D (n = 726,039). Genotypes on 32,901 SNP were available for 653,199 (lines A and B) and 700,572 (lines C and D) pigs. Two-trait linear mixed models involving ADG with BFT and ADG with MOR were used. Fixed effects included the contemporary group (with farm, sex, off-test year, and off-test week), covariates of the growth test end-weight (BFT only) and parity (MOR only), while random effects were litter, additive genetics, and residual. To assess the effect of the APY core size, we tested six different sizes: first, we defined two baseline cores explaining 98% or 99% of the variance in G, C98, and C99, respectively; then, we tested doubling or tripling the baseline cores (2C98, 3C98, 2C99, and 3C99). Analyses utilized the BLUPF90 family of software. Results from C98 were the same as those from C99, but C99 yielded relatively better GWAS resolution and had higher power to identify small-effect QTL. C99 is thus recommended if its size is computationally manageable. No QTL were identified for MOR, but potentially pleiotropic regions were observed on SSC5 and 7. Major QTL were identified on SSC1 and SSC17 for ADG and on SSC1 and SSC2 for BFT. Multi-line GWAS successfully reproduced results for single lines and identified more small-effect QTL on SSC7 for ADG in terminal lines, suggesting that the method is robust to handle larger data from multiple genetic sources. Core sizes 2C (2C98 or 2C99) and 3C (3C98 or 3C99) improved GEBV accuracy relative to respective baseline cores, C98 or C99, but GWAS signals were similar across core sizes. Overall, ssGWAS with APY using an approximated PEV matrix enables efficient use of large pig datasets, which enhances GWAS power and resolution while eliminating computational challenges.

Keywords: 2026

How to Cite:

Bermann, M., Bussiman, F., Chen, C., Hidalgo, J., Holl, J., Kayondo, F. & Lourenco, D., (2026) “Single-step GWAS with APY and its application to a large pig population”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2272158. doi: https://doi.org/10.31274/wcgalp.23411

Rights: 1