Using large-scale whole-genome sequence data for single-step genomic predictions in maternal and terminal pig lines

Abstract

Background: Whole-genome sequence (WGS) data harbor causative variants that may not be present in the regular SNP chip data. The objective of this study was to investigate the impact of using preselected variants from WGS for single-step genomic predictions in maternal and terminal pig lines with up to 1.8k sequenced and 104k imputed sequenced animals per line. Methods: Two maternal and four terminal lines were investigated for eight and seven traits, respectively. The number of sequenced animals ranged from 1,365 to 1,491 in maternal lines and 381 to 1,865 in terminal lines. Imputation occurred within each line, and the number of animals imputed to sequence ranged from 66k to 76k in maternal lines and 29k to 104k in terminal lines. Two preselected SNP sets were generated based on genome-wide association study (GWAS). Top40k included the SNP with the lowest p-value in each of 40k genomic windows; ChipPlusSign included significant variants integrated into the regular porcine SNP chip. Single-step genomic predictions with equal or different SNP variances using those SNP sets were compared to the regular porcine SNP chip. Results: In maternal lines, ChipPlusSign, and Top40k showed, on average, 0.62%, and 4.9% increased accuracy compared to the regular porcine SNP chip. The greatest changes were for fertility traits with Top40k, where the initial accuracy based on the SNP chip was low. However, for terminal lines, Top40k resulted in a loss of accuracy of 1% on average. Only ChipPlusSign provided a positive, albeit small, gain (0.85%). Assigning different variances for SNP slightly improved accuracies when using variances obtained from BayesR; however, the increase was inconsistent across the lines and traits. Conclusions: The benefit of using sequence data depends on the line, size of the genotyped population, and how the WGS variants are preselected. When WGS is available on hundreds of thousands of animals, the advantage of sequence data is present but limited in maternal and terminal pig lines.

Multi-line ssGBLUP evaluation using preselected markers from whole-genome sequence data in pigs

“…We investigated the average daily feed intake (ADFI), average daily gain (ADG), backfat thickness (BF), ADG recorded in crossbred animals (ADGX), and BF recorded in crossbred animals (BFX) in three terminal pig lines named TL1, TL2, and TL3. These three terminal lines were chosen from the seven lines analyzed in a previous study by Jang et al (2022a) and Ros-Freixedes et al (2022). These three lines were chosen because of their large numerical size, completeness of pedigree, and availability of phenotypic data for terminal traits.…”

“…Two scenarios were considered in this study: SLE and MLE. The same data were considered for SLE in each line as described in Jang et al (2022a). For the MLE scenario, all the data from every single line were combined.…”

“…Five different preselected SNP panels were created from WGS for genomic predictions. Two of them were as described in Jang et al (2022a); Ros-Freixedes et al (2022): 1) Top40k and 2) ChipPlusSign. Three additional preselected panels were designed for the current study: 3) TopSign, 4) LDTags, and 5) AllComb.…”