Expression Quantitative Trait Locus (eQTL)

eQTLs are classified by their genomic distance from the target gene, distinguishing local from distal regulatory mechanisms.

• Cis-eQTLs: Located near the target gene (typically within 1 Mb of the transcription start site). They often map to promoter or enhancer regions and directly affect transcription factor binding.
• Trans-eQTLs: Located far from the target gene, often on a different chromosome. They typically act by altering the expression of a transcription factor or regulatory RNA that then controls the target gene.
• Mechanistic Insight: Cis-eQTLs are more abundant and have larger effect sizes, making them the primary focus for fine-mapping causal variants from GWAS loci.

A defining feature of eQTLs is their dependence on biological context, reflecting the cell-type-specific nature of gene regulation.

• Tissue-Specific eQTLs: A variant may associate with gene expression in liver tissue but show no effect in brain tissue, due to differential chromatin accessibility.
• Context-Responsive eQTLs: Some eQTLs only manifest under specific conditions, such as after immune stimulation or drug treatment, revealing latent regulatory potential.
• GTEx Project: The Genotype-Tissue Expression project has cataloged eQTLs across 49 human tissues, demonstrating that the majority of regulatory variants are shared across tissues, but the strongest effects are often tissue-specific.

eQTL mapping is a massive multiple-testing problem requiring rigorous statistical control to link millions of variants to thousands of genes.

• Linear Regression: The standard approach tests the additive effect of genotype dosage (0, 1, 2 minor alleles) on normalized gene expression levels.
• Matrix eQTL: An R package that performs billions of association tests in seconds by using efficient matrix operations, bypassing slow per-gene loops.
• Multiple Testing Burden: A typical study tests 10^6 SNPs × 2 × 10^4 genes = 2 × 10^10 associations. False Discovery Rate (FDR) control via the Benjamini-Hochberg procedure is mandatory.
• Covariates: Models must include principal components to correct for population stratification and known technical factors like batch effects.

The primary translational value of eQTLs is identifying the causal gene and direction of effect at non-coding GWAS risk loci.

• Colocalization Analysis: Statistical methods like COLOC and FINEMAP assess whether the same causal variant drives both the eQTL signal and the GWAS disease association.
• Mechanistic Triangulation: If a SNP increases the expression of Gene X (eQTL) and also increases disease risk (GWAS), Gene X is implicated as a causal mediator of the disease.
• Drug Target Validation: This approach has successfully identified SORT1 as a lipid-modulating target and IL6R as a coronary artery disease target, providing genetic support for therapeutic development.

An alternative to total expression eQTL mapping that provides a highly sensitive internal control for regulatory effects.

• Principle: In a heterozygous individual, the two alleles of a gene are exposed to the same cellular environment. Any deviation from a 50:50 expression ratio indicates a cis-regulatory effect linked to one of the haplotypes.
• Allele-Specific Expression (ASE): ASE analysis uses heterozygous coding SNPs as tags to measure transcript abundance from each chromosomal copy.
• Advantage: ASE is robust to trans-acting variation and environmental confounders, providing a direct readout of cis-regulatory activity with higher statistical power than total expression mapping.

Single-cell RNA-seq has revolutionized eQTL analysis by resolving regulatory effects to specific cell types without physical sorting.

• Cell-Type Resolution: Traditional bulk tissue eQTLs can be masked or diluted by heterogeneous cell populations. Single-cell eQTLs identify effects specific to rare cell types like dendritic cells or regulatory T cells.
• Dynamic eQTLs: Pseudotime trajectory analysis can reveal eQTLs that only affect gene expression during a specific stage of cellular differentiation or activation.
• Computational Methods: Tools like SAIGE-QTL and cellRegMap use mixed models to account for the sparsity and zero-inflation inherent in single-cell count data while controlling for individual-level relatedness.