Tejaas

Tejaas identifies trans-acting expression quantitative trait loci (trans-eQTLs) by aggregating small genetic effects across many genes to detect distal regulatory variants that influence gene expression.

Key Features:

• L2-regularized reverse multiple regression: Performs L2-regularized 'reverse' multiple regression by regressing each single nucleotide polymorphism (SNP) against all gene expression levels simultaneously.
• k-nearest-neighbor confounder removal: Uses a non-linear, unsupervised k-nearest-neighbor method to remove confounders from expression data.
• Aggregation of small trans-effects: Aggregates evidence across many genes to increase power for detecting small-effect trans-eQTLs while mitigating strong expression correlations and correlated p-values.
• Addresses multiple-testing burden and tissue-specificity: Specifically targets challenges of small effect sizes, extensive multiple-testing, and tissue-specific regulation in trans-eQTL discovery.
• Empirical discovery in GTEx: Predicted 18,851 unique trans-eQTLs across 49 tissues using Genotype-Tissue Expression (GTEx) data.
• Regulatory enrichment and disease overlap: Identified trans-eQTLs are enriched in open chromatin regions, enhancers, and other regulatory elements and overlap with disease-associated SNPs.
• Relevance to expression heritability: Targets trans-eQTLs that collectively account for a large fraction (≥70%) of expression heritability.

Scientific Applications:

• Trans-eQTL discovery: Detects distal regulatory variants that affect expression of multiple genes across the genome.
• Mapping tissue-specific regulation: Maps tissue-specific trans-regulatory effects across 49 GTEx tissues to study context-dependent gene regulation.
• Linking genetic variants to disease: Identifies trans-eQTLs that overlap disease-associated SNPs to elucidate mechanisms underlying complex diseases.
• Annotating regulatory elements: Associates SNPs with open chromatin, enhancers, and other regulatory elements implicated in transcriptional regulation.

Methodology:

Applies L2-regularized 'reverse' multiple regression regressing each SNP against all gene expressions and a non-linear unsupervised k-nearest-neighbor method to remove confounders.