Scientists have charted how DNA shapes the chemistry of young pigs’ blood, identifying 41 significant associations between genetic variants and serum metabolites that consolidate into 18 distinct metabolic quantitative trait loci, or mQTLs. The study, published in Genetics Selection Evolution, analyzed 213 three-way crossbred pigs that were approximately five weeks old, a life stage that had not previously been examined with this combination of serum metabolomics and genome-wide association mapping in pigs. Seven of the 18 loci overlap with findings from human studies, while the remainder represent new entries in the growing catalog of genetic sites that influence metabolite levels in livestock.
The pigs were three-way crossbreds with Duroc as the paternal line and a Landrace by Yorkshire maternal line, all born and raised under conventional Danish commercial conditions across two farms. Serum was collected once per pig at roughly five weeks of age, about two weeks after weaning, during September and October 2019. Of the 218 pigs initially enrolled, 213 passed genotype quality control and entered the final analyses. The cohort included 119 males and 94 females, and antibiotic treatment of diarrhea within the two weeks before sampling was recorded and included as a covariate in the genetic models.
Metabolite profiling was carried out with a targeted liquid chromatography–mass spectrometry workflow that detected and annotated 248 serum metabolites, of which 231 were confirmed at the highest annotation confidence level using authentic reference standards. The panel spanned amino acids and their derivatives, lipids including glycerophospholipids and sphingolipids, fatty acids and organic acids, nucleobases, nucleosides and nucleotides, carnitines and acylcarnitines, carbohydrates and sugar acids, bile acids, and benzenoid compounds. Beyond the metabolites themselves, the researchers computed all possible pairwise metabolite ratios, yielding 30,628 derived traits. Ratios can capture product–substrate relationships, conjugation states, or shared regulation by enzymes and transporters, and they can also dampen shared technical and physiological noise such as dilution effects or the time since an animal last ate.
Genetically, the animals were genotyped with a 50,697-SNP chip, and the team then imputed genotypes to whole-genome sequence level using reference panels of the corresponding Danish breeds, in two steps that passed through a high-density intermediate array. After quality filtering, 14,968,568 autosomal variants entered the second stage of association testing. The genome-wide association studies used additive linear mixed models with a leave-one-chromosome-out approach, adjusting for sex, herd, recent antibiotic treatment, age, and body weight, and additionally for mass spectrometry batch in the ratio analyses. Significance thresholds were set separately for metabolites and ratios by correcting for the effective number of independent variants and traits, arriving at 9.14 × 10⁻⁸ for metabolites and 4.29 × 10⁻⁹ for ratios.
The first screening pass, using the SNP chip data, flagged associations for 15 metabolites and 24 metabolite ratios. A second pass with the imputed sequence-level genotypes, restricted to chromosomes harboring significant signals, refined the findings, and a conditional analysis that included lead variants as covariates revealed one additional independent association for a metabolite level on the same chromosome. When trait-specific association regions shared a lead variant or overlapped, they were merged, producing six complex mQTLs that each explained more than one metabolic trait and twelve independent mQTLs for single metabolites or ratios. One region was associated with up to nine metabolic phenotypes, and within multi-trait regions the lead variants for different traits were in high linkage disequilibrium, consistent with a single underlying association signal rather than distinct loci.
Not every shared region told a simple story. The phenotypic correlations among traits controlled by the same mQTL ranged from weak, around 0.2, to extremely strong, above 0.95. In one region, mQTL 7, the association showed an asymmetric pattern: the lead variant identified for the ratio between acetyl-tryptophan and formyl-methionine did not reach genome-wide significance for acetyl-tryptophan alone, whereas the lead variant for acetyl-tryptophan was significantly associated with the ratio. The authors interpret such patterns as a shared genetic signal with unequal effect sizes. Intriguingly, the biochemical character of the traits offered hints about mechanism even where phenotypic covariance was limited: all traits linked to mQTL 7 contained an acetyl group, and all traits linked to mQTL 15 contained a glutamyl group.
The imputed sequence data also allowed the team to hunt for candidate causative mutations. Variants reaching genome-wide significance were examined with a variant effect predictor, retaining only those classified as having at least moderate impact on a gene. This yielded 88 significant SNPs with predicted functional effects, distributed across 46 genes within 11 mQTL regions; the remaining seven mQTLs contained no variants meeting these criteria. Cross-referencing the mQTL regions with the human GWAS Catalog, and retaining only overlaps where genes identified in human studies sat within the corresponding pig region, produced candidate genes for seven mQTLs encompassing 30 unique genes. Several of these genes are already tied to human disorders: argininosuccinate synthase 1 with citrullinemia, kynurenine 3-monooxygenase with Huntington’s and Alzheimer’s disease, gamma-glutamyltransferase 1 with glutathionuria, and N-acetyltransferase 8 with chronic kidney disease.
Through an integrative prioritization strategy that combined manual assessment of enzyme- and transporter-encoding genes with human catalog evidence and variant effect predictions, the researchers proposed putative causal genes for 14 of the 18 mQTLs, predominantly enzymes and transporters. Two examples illustrate the approach. The region associated with 3-hydroxyoctanoic acid, capryloyl-glycine, and 2-hydroxycapric acid harbors variants with predicted functional impact on acyl-CoA thioesterase 13, an enzyme whose activity is expected to influence multiple downstream acyl-CoA-derived metabolites, matching the observed pleiotropic pattern. Another region affecting circulating deoxyuridine levels contains variants with predicted impact on uridine phosphorylase 2, an enzyme with broad substrate specificity that includes deoxyuridine; notably, uridine was measured in the dataset but showed no association with this region, suggesting substrate-specific sensitivity within pyrimidine nucleoside metabolism.
The study is not the first of its kind in pigs—a 2025 analysis of a large cohort of purebred pigs at slaughter age demonstrated that some genetically influenced metabotypes are conserved between humans and pigs while others are unique to pigs—but it is the first to profile serum metabolites in early-age crossbred animals. Direct comparison with that earlier work was constrained by limited overlap in measured metabolites, only 32 of 248, yet two metabolites, citrulline and kynurenine, were consistently associated with the same mQTL regions across both studies, which the authors take as evidence of robust and reproducible genetic regulation at those loci. A complementary earlier study of the fetal pig plasma metabolome had shown that the genetic architecture of the porcine metabolome is stage-dependent and can differ markedly between fetal and finishing stages, underscoring the value of sampling at multiple ages.
The authors are candid about the limits of their findings. Extensive long-range linkage disequilibrium and the relatively low genetic heterogeneity of pig populations mean that imputation often produces clusters of variants with identical p-values, and there is a risk of false positives in candidate gene assignment; predicted variant effects alone cannot establish causality, and associations may also arise from regulatory mechanisms that alter gene expression rather than coding changes. At least nine promising candidate genes contained no coding variants with predicted impact, pointing toward regulatory mutations and motivating the integration of expression quantitative trait loci and chromatin accessibility data in future work. Four mQTLs remained without a prioritized candidate gene, possibly because causal genes do not conform to the enzyme-and-transporter expectation, since human studies suggest roughly six percent of causal mQTL genes encode neither. The team suggests that future studies incorporate transcriptomics and whole-genome sequencing to refine gene prioritization, and positions the results as a foundation both for incorporating metabolomics into pig breeding programs and for translational research into human health and nutrition.
Subject of Research: Genetic loci controlling serum metabolite levels in young crossbred pigs
Article Title: Expanding the pig metabolomics QTL catalog
Article References: Ibragimov, E., Morsing, M. K., Rydal, M. P., Cai, Z., Stenlund, H., Nielsen, J. P., Fredholm, M., & Karlskov-Mortensen, P. (2026). Expanding the pig metabolomics QTL catalog. Genetics Selection Evolution, 58(1), Article 75. https://doi.org/10.1186/s12711-026-01095-0
Image Credits: AI Generated
DOI: 10.1186/s12711-026-01095-0
Keywords: pig genomics, metabolomics, GWAS, mQTL, serum metabolites, crossbred pigs, candidate genes, metabolite ratios, livestock genetics, translational model, Genetics Selection Evolution, Expanding
News Source: Juliet Wilcox. (October 28, 2026). Genetic map of the young pig serum metabolome reveals 18 metabolic loci. Scienmag.
