Researchers have combined genome-wide association studies with a suite of functional genomic tools to dissect the genetic architecture of body weight and body mass index in pigs, and they report that folding the resulting functional variants back into prediction models lifted genomic prediction accuracy by up to 21 percent. The study, published in Genetics Selection Evolution, analyzed 1226 Shanxia Long Black pigs and traced body size signals to specific regulatory elements, effector tissues, and candidate genes, offering a template for functionally informed breeding of complex livestock traits.

Body weight at slaughter age and body mass index, calculated as body weight in kilograms divided by the square of body length in meters, are central determinants of meat yield and productivity in swine. Both traits are controlled by many genes of small effect, and most of the responsible variation maps to non-coding regions of the genome rather than to protein-coding sequences. Although genome-wide association studies have repeatedly mapped quantitative trait loci for these traits, pinpointing the actual causative genes and variants has remained difficult, which in turn limits the effectiveness of genomic prediction in breeding programs.

The team, led by Naibiao Yu and Yuanmei Guo along with colleagues at their institution, began by genotyping the pigs on a 50K SNP chip and imputing their genotypes to whole-genome sequence density using a population-specific reference panel. After quality control and imputation, more than 14.6 million SNPs were available for analysis. Imputation quality was high, with a concordance rate of 99.16 percent and a squared correlation of 0.89 between genotyped and imputed genotypes. All animals were raised at the same farm in Jiangxi province, China, in six production batches, and were slaughtered at body weights of 90 to 125 kilograms at ages of 210 to 240 days.

The genome-wide association analysis, conducted with a linear mixed model in GEMMA, identified eight genomic regions reaching genome-wide significance for body weight, with lead SNPs on pig chromosomes SSC1, SSC2, SSC3, SSC7, SSC11, and SSC12. For body mass index, four significant regions were detected on SSC1, SSC2, SSC3, and SSC17. Quantile-quantile plots indicated minimal population stratification, with genomic inflation factors of 1.051 for body weight and 1.031 for body mass index. Notably, signals for both traits converged on a shared region at roughly 52.8 megabases on SSC2, a locus containing a cluster of olfactory receptor genes including OR2T27, OR2T6, and OR2T1.

To narrow the association signals to likely causal variants, the researchers applied linkage disequilibrium analysis and SuSiE fine-mapping within 500-kilobase windows centered on each lead SNP. Variants with posterior inclusion probabilities above 0.1 were retained as candidates, yielding 29 candidate functional variants for body weight and 12 for body mass index. In total, the study delineated 10 quantitative trait loci with 45 high-confidence candidate functional variants. These loci, however, explain only a small share of the genetic variance, about 20.76 percent for body weight and 10.27 percent for body mass index, underscoring the strongly polygenic nature of the traits. SNP-based heritability estimates were 0.23 for body weight and 0.37 for body mass index.

The functional annotation stage drew on chromatin state maps for 14 pig tissues, deep-learning predictions from a Basenji convolutional neural network adapted to the pig genome, and high-throughput chromosome conformation capture data. The body weight locus on SSC2 overlapped an open chromatin region in the cerebellum, pointing to brain regulatory activity, and the olfactory receptor genes at this locus are thought to modulate feeding behavior, with feed intake a plausible intermediate linking the locus to growth. Body mass index loci showed enhancer activity across multiple tissues: the SSC1 locus overlapped an adipose enhancer, the SSC3 locus encompassed brain-active enhancers, and the SSC17 locus coincided with a weak liver enhancer.

Hi-C data, generated from backfat tissue and combined with published brain and liver contact maps, allowed the team to connect regulatory variants to target genes through topologically associating domains and chromatin loops. In adipose tissue, the SSC1 body mass index locus contained an enhancer within a domain spanning ZER1, and the lead SNP sat within that gene, a nomination corroborated by gene-based association testing. The SSC17 locus connected through bridging loops to HAO1 in the liver, a gene highly expressed in hepatocytes. Taken together, the multi-omics analyses converged on ZER1, KLHL29, and HAO1 as high-confidence candidate genes, with OR2T27 as a putative candidate on SSC2.

The candidate genes have intriguing parallels in human genetics. ZER1 encodes a substrate adaptor for a Cullin-2 ubiquitin ligase complex, is highly expressed in the brain, and has been associated in human studies with LDL cholesterol, body mass index, and anorexia nervosa. KLHL29 has been linked in pig GTEx analyses to circulating LDL cholesterol, and human studies connect the orthologous region to lipid metabolism and obesity. HAO1 encodes a peroxisomal enzyme active in fatty acid oxidation, and human genome-wide studies associate it with type 2 diabetes susceptibility. The authors note that although BMP2, a known regulator of skeletal development, lies near HAO1, the metabolic signals at the locus prioritize HAO1 for body composition traits, though minor contributions from BMP2 cannot be excluded.

The predictive payoff came when the prioritized variants were incorporated as fixed effects into genomic prediction models. In the Shanxia Long Black pigs, a composite estimator combining candidate variant effects with genome-wide polygenic effects, termed QTL-fixed GEBV, achieved correlations of 0.61 with observed body weight and 0.60 with observed body mass index, significantly outperforming standard GBLUP, which reached only 0.41 and 0.48 respectively. In simulated breeding populations with 60K, 140K, and 660K marker densities, the same approach performed best at higher marker densities, reaching correlations of 0.85 with true breeding values at both 140K and 660K. Across scenarios, the functionally informed framework improved accuracy by up to 21 percent over conventional genomic prediction.

The authors are careful to frame the work as a methodological advance rather than a selection directive. They caution that improved prediction accuracy alone does not justify immediate application in breeding programs, since major-effect alleles may persist under selection through antagonistic pleiotropy, meaning production gains could carry costs to unmeasured fitness traits. They also emphasize that establishing definitive causality for the candidate genes will require experimental validation, such as CRISPR-Cas9 allelic editing and tissue-specific perturbations, and that single-cell multi-omics approaches will be needed to resolve cell-type-specific regulatory mechanisms obscured in bulk tissue data. Even so, the study demonstrates that integrating regulatory genomics into genomic prediction can enhance both accuracy and biological interpretability, offering a scalable strategy for genetic improvement of complex traits in livestock.

Subject of Research: Functional genomic dissection of body weight and body mass index traits in pigs and functionally informed genomic prediction

Article Title: Functional genomic dissection and prediction of body size traits in pigs

Article References: Yu, N., Cui, D., Xie, L., Tang, X., Xiong, S., Zhang, Y., He, R., Li, L., Xiao, S., & Guo, Y. (2026). Functional genomic dissection and prediction of body size traits in pigs. Genetics Selection Evolution, 58(1), Article 55. https://doi.org/10.1186/s12711-026-01078-1

Image Credits: AI Generated

DOI: 10.1186/s12711-026-01078-1

Keywords: pigs, body weight, body mass index, GWAS, fine-mapping, genomic prediction, Hi-C, chromatin states, Basenji, ZER1, KLHL29, HAO1