Non-coding GWAS โ From Associated Signals to Regulatory Mechanisms
Why Is This Important?
Many disease-associated signals identified in GWAS lie outside protein-coding exons. To interpret these signals, one must move beyond merely identifying the nearest gene to the most significant SNP and systematically narrow down the actual causal variant candidates, along with the affected cell types, regulatory elements, and target genes.
The non-coding region does not imply a lack of function; it may contain enhancers, promoters, insulators, and various RNA regulatory elements. However, since a single annotation does not prove function, statistical, chromatin, expression, and perturbation evidence are integrated stepwise.
Core Concepts
The GWAS lead variant serves as a proxy representing the association signal at a locus. Because neighboring variants are co-inherited through linkage disequilibrium (LD), the variant with the smallest p-value is not guaranteed to be the true functional variant. Fine-mapping should account for ancestry-specific LD differences, imputation quality, and multiple independent signals to construct a credible set with high causal probability.
A credible set is not a definitive list of answers but a probabilistic candidate set dependent on the model and LD reference. When sample sizes are small or LD is strong, the number of candidates increases; furthermore, if assumptions regarding the number of causal variants are incorrect, the set may be distorted. Data from other ancestries and independent cohorts can help narrow down these candidates.
Steps for contextualizing regulatory signals
We verify whether candidate variants overlap with open chromatin or enhancer marks in relevant tissues. While motif alterations and allele-specific accessibility suggest potential transcriptional regulatory mechanisms, overlap alone does not prove that the locus is functional in the disease state; cell type and stimulation conditions must align.
eQTL analysis and colocalization assess the likelihood that a locus harbors shared signals for both expression and disease traits. The target gene may be distal rather than the nearest one, and a single enhancer may interact with multiple genes.
What the 3C-based methods measure
3C measures the contact frequency between two selected loci after cross-linking and ligating fixed chromatin. 4C identifies genome-wide partners from a single viewpoint, while Hi-C captures a genome-wide contact matrix. Variants such as capture-C and promoter capture Hi-C also exist to enrich specific regions.
These values serve as proxies for the frequency with which two DNA segments were in close proximity within a cell population. They are influenced by genomic distance, restriction site distribution, GC content, mappability, sequencing depth, and normalization procedures. Loops observed in bulk Hi-C data may not represent static structures present in all cells.
Small Example
Suppose a credible set for a disease locus contains an enhancer variant that alters accessibility of the risk allele in relevant immune cells, and this enhancer contacts the promoter of gene B located 200 kb away. If the eQTL signal for gene B colocalizes with the same association signal, a coherent hypothesis is formed.
Nevertheless, physical contact alone is not sufficient to establish regulatory causality. The causal chain is strengthened only when precise perturbation of the variant or enhancer leads to changes in gene B expression and disease-relevant cellular phenotypes in the expected direction, and these effects are replicated across independent experiments.
Evidence Ladder
The specificity of mechanistic claims increases along the sequence: GWAS association โ fine-mapped candidate โ relevant-cell chromatin annotation โ allele-specific molecular effect โ contact/eQTL linkage โ perturbation โ phenotype. While not all studies need to reach the final stage, it is essential to clearly indicate the extent of validation achieved.
Common Misconceptions
- The lead SNP is not synonymous with the causal variant.
- Enhancer overlap does not specify the target gene.
- Chromatin contact is not equivalent to active regulation.
- Overlap between eQTLs and GWAS signals does not automatically prove shared causal signals.
- Effects observed in reporter assays may differ from those in the native chromatin context.
Limitations
Obtaining directly relevant disease tissues is difficult; using surrogate cells may alter the biological context. Fine-mapping and contact maps are influenced by ancestry, cellular state, and analytical resolution. Results should be reported with the genome build, LD reference, cell type, assay, and pipeline version used.
Reading in Context
Expression-associated and colocalization analyses are covered in the eQTL section, open chromatin measurement is detailed in ATAC-seq, and reference coordinates and allele representation are addressed in the Reference genome concept.