Expression Quantitative Trait Loci

eQTLs are classified by their genomic proximity to the target gene:

• Cis-eQTLs: Located near the gene they regulate, typically within 1 megabase of the transcription start site. These act on local chromatin to directly modulate promoter or enhancer activity.
• Trans-eQTLs: Located far from the target gene, often on different chromosomes. These typically encode transcription factors or signaling molecules that indirectly influence expression.

Cis-eQTLs generally have larger effect sizes and are more reproducible across studies.

A defining feature of eQTLs is their context-dependency. A genetic variant may significantly alter gene expression in one tissue but have no effect in another.

• This is driven by tissue-specific chromatin accessibility and transcription factor availability.
• Resources like the GTEx project have mapped eQTLs across 49+ human tissues.
• Tissue-specific eQTLs are critical for pinpointing the causal cell type in genome-wide association studies for complex diseases.

eQTLs operate by disrupting specific molecular interactions:

• Transcription Factor Binding: Variants in promoters or enhancers can create or destroy binding motifs, altering RNA polymerase II recruitment.
• Chromatin State: Variants can shift local nucleosome positioning or histone modification landscapes, changing DNA accessibility.
• mRNA Splicing: Exonic or intronic variants, known as sQTLs, can create cryptic splice sites or disrupt canonical donor/acceptor sites.
• RNA Stability: Variants in 3' UTRs can disrupt microRNA binding sites, altering transcript degradation rates.

eQTL mapping requires paired genotype and transcriptome data from a population:

• Genotyping: Microarrays or whole-genome sequencing identify single nucleotide polymorphisms across individuals.
• RNA Quantification: RNA-seq or expression arrays measure transcript abundance, often normalized using TPM or FPKM.
• Association Testing: Linear regression or Spearman correlation tests the relationship between each variant's dosage and the expression of each gene, typically using tools like Matrix eQTL or FastQTL.
• Multiple Testing Correction: A stringent Bonferroni or Benjamini-Hochberg correction is applied to control false discovery across millions of variant-gene pairs.

eQTLs provide a mechanistic bridge between non-coding GWAS variants and disease. Colocalization analysis statistically tests whether the same causal variant drives both a GWAS trait signal and an eQTL signal.

• Methods like COLOC and SMR use Bayesian frameworks to distinguish true pleiotropy from linkage disequilibrium artifacts.
• A significant colocalization suggests the disease risk is mediated through altered gene expression in a specific tissue.
• This identifies causal genes and drug targets from the hundreds of candidates in a GWAS locus.

Modern sequence-based models can predict eQTL effects in silico without experimental data:

• Models like Enformer and Basenji2 take raw DNA sequence as input and predict expression tracks across tissues.
• In silico mutagenesis systematically mutates each nucleotide and measures the predicted expression change, generating a variant effect score.
• These predictions correlate with experimentally measured eQTL effect sizes and can prioritize functional variants within large linkage disequilibrium blocks.
• This approach extends eQTL analysis to rare variants and cell types not accessible in large cohort studies.

A computational perturbation method used to interpret eQTL effects. ISM systematically mutates every nucleotide in an input sequence and measures the predicted change in a model's output. This reveals the causal regulatory motifs driving an eQTL signal. For a variant associated with expression, ISM can confirm whether the alternate allele disrupts a transcription factor binding site, providing mechanistic insight beyond statistical linkage.

A high-throughput experimental technique that directly tests the regulatory activity of thousands of synthesized DNA sequences. MPRA provides the gold-standard functional validation for eQTLs predicted computationally. By measuring transcribed barcodes, it quantifies whether a variant allele truly alters enhancer or promoter activity, bridging the gap between statistical association and causal molecular biology.

Systematic non-biological variations introduced by differences in sample processing, reagent lots, or sequencing platforms. These confounders can create spurious eQTLs if not corrected. Tools like ComBat-Seq use negative binomial regression to adjust for known technical covariates while preserving biological variability, a critical preprocessing step before performing eQTL mapping or training predictive models.

A machine learning strategy where a model pre-trained on a data-rich task—like predicting chromatin accessibility from DNA—is fine-tuned for a related task, such as gene expression prediction. This leverages learned universal regulatory features to improve eQTL detection in tissues with limited samples. Self-supervised pretraining on unlabeled genomes creates a generalizable foundation that encodes the syntax of cis-regulation.