DNA methylation is one of biology’s most conserved regulatory mechanisms. Enzymes called DNA methyltransferases attach methyl groups to cytosine bases, producing 5-methylcytosine, a mark that can be faithfully copied through cell division yet remains responsive to developmental and environmental cues. In humans and classic model organisms, methylation has been studied intensively, underpinning phenomena as diverse as X-chromosome inactivation, genomic imprinting, and the maintenance of cellular identity. In domestic animals, however, the picture has remained fragmentary. Most livestock studies have relied on differential methylation analysis — simply comparing methylated regions between groups of animals with different phenotypes. Such comparisons identify correlations but cannot say which epigenetic differences are actually driven by genetic variation, nor which ones matter for traits.

To close that gap, a team led by Yiting Wang, Zitao Chen, and Qishan Wang generated and integrated an extraordinary dataset from pig skeletal muscle: 185 whole-genome bisulfite sequencing samples, 148 RNA-seq transcriptomes, and 8 three-dimensional chromatin interaction maps, alongside whole-genome sequencing data. Fifty-six of the bisulfite samples were newly generated for the study, drawn from Jinhua pigs raised in Zhejiang Province and Duroc pigs raised in Anhui Province, while 129 came from public repositories. The bisulfite approach converts unmethylated cytosines to uracil during library preparation, allowing methylation levels to be read out base by base. The team sequenced on Illumina NovaSeq 6000 instruments, verified conversion efficiency with unmethylated lambda DNA spike-in controls, and ultimately profiled more than 5.25 million individual CpG sites and roughly 1.25 million regional CpG blocks across the porcine genome.

A crucial methodological hurdle was obtaining reliable genotypes for samples that had only bisulfite data. The researchers called SNPs directly from the bisulfite reads using Bis-SNP, then imputed them to sequence-level density with Beagle, using the PHARP v3 pig haplotype reference panel. Benchmarking against matched whole-genome sequencing data from twenty Duroc animals showed that, at a stringent dosage R-squared threshold above 0.9, imputed genotypes achieved nearly 90 percent precision and over 93 percent overall concordance with the reference calls. After merging with WGS-derived variants, the final dataset contained more than 12.2 million high-quality SNPs — a foundation robust enough for quantitative trait locus mapping at scale.

The centerpiece of the analysis is a block-resolved methylation quantitative trait locus, or meQTL, atlas. Rather than treating each CpG site in isolation, the team used the wgbstools toolkit to segment the genome into co-methylated blocks — typically a few hundred base pairs long — and mapped genetic variants within one megabundle of each block’s start position using tensorQTL. The results were striking. Independent cis-meQTL signals numbered 49,801 at single CpG sites and 448,028 at blocks, implicating 67,333 genetically regulated CpGs and 436,630 blocks. Cross-cohort validation on chromosome 1 showed that roughly 85 to 87 percent of associations detected in the newly generated cohort replicated in the public cohort with concordant effect directions. Allele-specific methylation analysis provided independent confirmation: block pairs showing recurrent allele-specific methylation were more than twice as likely to harbor meQTLs, with 98.5 percent of overlapping signals agreeing in allelic direction.

Heritability estimates underscored the biological meaning of these signals. CpGs with detectable meQTLs showed a mean cis-heritability of about 0.41, and regulated blocks about 0.25, whereas unregulated sites and blocks hovered near 0.04 to 0.05. Regions governed by multiple independent genetic variants were progressively more heritable, and rarer alleles tended to show larger effects on methylation. Enrichment analyses placed meQTL variants preferentially in promoters, enhancers, and transcription start site-proximal regions, and sequence-based prediction models suggested the strongest effects on chromatin accessibility in skeletal and cardiac muscle cells. Roughly 19 percent of meQTL variants were predicted to alter transcription factor binding motifs — including motifs for ARNT::HIF1A, HAND2, ZEB1, and SOX10 — although the authors caution that these remain computational predictions rather than experimentally validated binding changes.

Connecting methylation to gene expression, the team performed expression quantitative trait methylation mapping, pairing each gene with CpGs and blocks within one megabase of its transcription start site. They identified 14,726 significant site-gene pairs and 4,637 block-gene pairs, with effects distributed almost evenly between positive and negative directions — a pattern the authors attribute to the mixture of promoter, gene body, enhancer, and distal contexts captured by the analysis rather than a contradiction of the canonical repressive role of promoter methylation. To establish that these methylation-expression links share a genetic basis, the researchers deployed colocalization, HyPrColoc, and summary-data-based Mendelian randomization. Together these approaches supported 15,602 site-gene and 86,318 block-gene relationships, with thousands backed by multiple methods. Hi-C chromatin contact maps then supplied spatial context: some distal CpG-gene pairs shared the same topologically associating domain, and a handful — such as a CpG more than 500 kilobases from SIPA1L2 — were physically connected by the same chromatin loop.

The leap to complex traits came from integrating the meQTL atlas with genome-wide association statistics for 91 growth, carcass, and meat quality traits from PigBiobank. Complementary methods — colocalization, SMR, epigenome-wide association studies using genetically predicted methylation, and HyPrColoc — yielded 73,580 methylation-trait associations, with 8,097 supported by at least two approaches. Several loci stood out. A CpG in the promoter of NEDD4L, a gene implicated in muscle development and insulin resistance, associated strongly with average daily gain and backfat thickness. A block within the HSPB3 promoter, spanning three enhancers, associated with lean cut percentage, loin muscle area, and loin depth. At the MPC2 locus, a single variant, rs346166613, drove both site-level and block-level methylation signals linked to backfat thickness, with MPC2 highly expressed in adipocytes. The team also built methylation-based genetic scores from genotype-predicted methylation; these achieved moderate out-of-sample prediction, peaking for backfat thickness in Duroc pigs at a correlation of 0.353 and about 6 percent predictive variance, and combining them with conventional polygenic scores improved prediction beyond either alone.

Perhaps the most compelling result is a full regulatory cascade validated at the bench. Multi-trait colocalization and mediation Mendelian randomization identified 12,171 CpG-gene-trait triplets, with 2,702 showing significant mediation — often with more than 90 percent of the genetic effect on the trait apparently transmitted through methylation and expression. One prioritized pathway centered on a CpG block near SOCS3, a modulator of JAK-STAT signaling previously linked to myogenic differentiation. The block, its methylation, SOCS3 expression, and average daily gain all shared genetic signals, and Hi-C data placed the block and the gene in the same TAD. Using CRISPR-dCas9-TET1 targeted demethylation in immortalized porcine muscle satellite cells, the researchers reduced methylation at the block, observed increased SOCS3 protein by western blot, and measured enhanced cell proliferation and altered cell-cycle distribution — proof of concept that the computationally prioritized locus has real regulatory teeth.

The authors are careful about scope. The study covers only skeletal muscle; colocalization and Mendelian randomization establish convergent genetic evidence, not definitive causality; and functional editing validated just one locus among thousands of candidates. Methylation-based scores, being genotype-derived, capture genetic rather than environmental epigenetic information. Still, the resource — publicly queryable through the PSMEA database — and the prioritization framework, which distilled 6,894 candidate genes across 84 traits into Gold, Silver, and Bronze tiers, offer breeders and biologists a powerful new lens. By showing that a genetically predictable slice of the methylome carries trait-relevant, mechanistically interpretable information, the study transforms DNA methylation from a correlational curiosity into a working layer of livestock genomics, with implications that could eventually ripple from the pig pen to human medicine.

Subject of Research: Genetic regulation of DNA methylation and its role in complex trait variation in pig skeletal muscle

Article Title: Integrative genomics decoding DNA methylation-mediated genetic control of complex traits in pigs

Article References: Wang, Y., Chen, Z., Fan, M., Liu, C., Gu, J., Yu, P., Jin, W., Zhang, Z., Xie, S., Yang, T., Jiang, J., Zhang, Z., Wang, Z., ZHang, Z., Fang, L., Pan, Y., & Wang, Q. (2026). Integrative genomics decoding DNA methylation-mediated genetic control of complex traits in pigs. Journal of Advanced Research. https://doi.org/10.1016/j.jare.2026.10.001

Image Credits: AI Generated

DOI: 10.1016/j.jare.2026.10.001

Keywords: DNA methylation, meQTL, pig genomics, epigenomics, skeletal muscle, Mendelian randomization, colocalization, Hi-C, CRISPR-dCas9-TET1, methylation-based genetic scores, livestock breeding, SOCS3

The scale of the underlying data collection is considerable. Between 2007 and 2018, the team recruited 250 patients with colon cancer and isolated tumor samples together with matched adjacent normal tissue from each patient. Rather than relying on the methylation arrays that dominate most epigenome-wide association studies, the researchers applied three complementary technologies to the same samples: whole-genome sequencing to read out the complete genetic sequence, whole-genome bisulfite sequencing to measure methylation at essentially every cytosine in the genome, and RNA sequencing to quantify gene expression. This trio of measurements, gathered from both tumor and normal tissue, allowed the team to connect genetic variation, methylation state, and gene activity within the same individuals, an integration that is rarely achieved at this depth in cancer genomics.

The headline result is a catalog of staggering size. By testing which single nucleotide polymorphisms, or SNPs, are statistically associated with methylation levels at nearby CpG sites, the team identified 48,429,542 cis-acting SNP-CpG associations spanning 553,834 meCpGs across the genome, all passing an extremely stringent significance threshold of p less than 2 times 10 to the power of minus 12. In other words, more than half a million methylation sites in the colon cancer genome show measurable genetic control from variants located nearby on the chromosome. This is precisely the kind of resource that has been missing for East Asian populations, where meQTL studies at single-base resolution have been scarce, even though colorectal cancer incidence patterns and genetic architecture can differ from those of European-ancestry cohorts that have dominated the field.

One of the most striking patterns in the data is pleiotropy: the same methylation site is often tugged at by multiple independent genetic variants. Sixty-three percent of the CpGs linked to meQTLs were associated with more than one lead variant, suggesting that methylation at a given cytosine is rarely governed by a single genetic switch. Instead, the epigenetic state of a CpG appears to emerge from the combined influence of several regulatory elements, each contributing its own effect. This layered architecture has practical consequences for interpreting disease genetics, because a variant that nudges methylation slightly may only matter in combination with others, and a GWAS signal attributed to one gene may in fact operate through several methylation targets simultaneously.

The study also connects genetic control to the methylation changes that define cancer itself. Colon tumors differ from normal colon tissue in characteristic patterns of methylation gain and loss, concentrated in differentially methylated regions, or DMRs. When the researchers overlaid their meQTL map onto these tumor-specific DMRs, they found that 16.8 percent of lead meQTLs sit inside regions whose methylation is dysregulated between tumor and normal tissue. That overlap implies a mechanistic bridge: inherited variants may predispose certain genomic locations to the methylation errors that accumulate during tumorigenesis, effectively pre-marking vulnerable sites in the genome long before cancer develops.

To trace the chain of causation from DNA variant to methylation to gene expression, the team turned to colocalization analysis, a statistical method that asks whether two association signals, one for methylation and one for gene expression, are likely driven by the same causal variant rather than by two different variants that merely sit close together on the chromosome. This analysis yielded 14,104 CpG-gene pairs that appear to share a common causal variant. Even more revealing were the mediation analyses built on top of these pairs: the researchers estimated that meQTLs mediate 51 percent of the SNP effects on gene expression. In plain terms, for roughly half of the genetic influences on gene activity that the team could quantify, the effect appears to flow through methylation rather than acting directly on transcription machinery. That makes DNA methylation not merely a passive marker of cancer biology but a major conduit through which inherited sequence variation shapes tumor gene expression.

Pathway analysis of the genes and methylation sites under genetic control pointed to several biological systems with obvious relevance to cancer. The team found epigenetic regulation of pathways involved in translation, RNA polymerase function, and DNA repair. Each of these is a plausible lever on tumor behavior: altered translation rates are a hallmark of many cancers, RNA polymerase activity determines how aggressively genes are transcribed, and defective DNA repair is a classic route to the mutation burden that drives malignancy. The finding that these pathways are subject to inherited, methylation-mediated regulation suggests that genetic susceptibility to colon cancer may operate in part by tuning the epigenetic setting of core cellular processes rather than by altering protein-coding sequences themselves.

Perhaps the most clinically resonant result concerns known colorectal cancer risk loci. Genome-wide association studies have identified dozens of genomic regions linked to colorectal cancer risk, but for most of them the causal gene and mechanism remain unknown. The new meQTL map covers 86.6 percent of previously identified CRC GWAS risk loci, and 40 percent of those covered loci showed significant colocalization with methylation signals. This means that for a substantial fraction of known risk regions, there is now a concrete, testable hypothesis about mechanism: the risk variant acts by altering methylation at nearby CpG sites, which in turn changes gene expression. For researchers hunting the functional consequences of GWAS hits, this resource converts a long list of anonymous risk loci into a ranked set of variant-CpG-gene triads to interrogate experimentally.

The study also fills an important demographic gap. Most large-scale meQTL resources, such as those derived from the Biobank Japan consortium or European-ancestry cohorts, reflect populations whose variant frequencies and linkage disequilibrium patterns differ from those of other groups. By building the catalog in an East Asian cohort of colon cancer patients, the team has created a reference that should improve the transferability of genetic risk prediction and functional interpretation to populations that have been underrepresented in cancer genomics. The authors describe the resulting dataset as a comprehensive resource integrating genetic, epigenomic, and transcriptomic information on colon cancer, revealing co-regulation of methylation and gene expression at single-base resolution across the genome.

The broader significance of the work lies in how it reframes the relationship between genetics and epigenetics in cancer. For years, DNA methylation changes in tumors were often treated as secondary consequences of transformation, downstream noise accompanying the driving mutations. This study argues for a more structured picture: a large portion of the tumor methylome is under direct genetic control, that control is pleiotropic and distributed, and it channels a substantial share of inherited risk into altered gene expression through defined molecular pathways. With the full dataset openly available as an open-access publication, other teams can now overlay their own GWAS results, test candidate variants for methylation-mediated effects, and design experiments around the specific CpG-gene links the map exposes. As multi-omics cohorts grow and similar maps are built for other cancers and populations, the colon cancer meQTL atlas stands as a template for turning epigenomic noise into mechanistic insight, one methylated cytosine at a time.

Subject of Research: Genetic regulation of DNA methylation in colon cancer through genome-wide meQTL mapping

Article Title: Genome-wide meQTL mapping reveals genetic control of DNA methylation in colon cancer

Article References: Hou, H., Zhou, Y., Chen, X., Luo, S., Yang, B., Peng, L., Tang, X., Yang, H., Wu, X., Liu, D., & Wu, K. (2026). Genome-wide meQTL mapping reveals genetic control of DNA methylation in colon cancer. Genome Medicine. https://doi.org/10.1186/s13073-026-01766-z

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01766-z

Keywords: colon cancer, DNA methylation, meQTL, epigenetics, whole-genome sequencing, whole-genome bisulfite sequencing, GWAS, colorectal cancer, gene expression, multi-omics, eQTL, cancer genomics