Shared and Disease-Specific Monocyte eQTLs Across Inflammatory Conditions

Background: Expression quantitative trait loci (eQTLs) carry a disproportionate share of autoimmune disease heritability, yet whether genetic regulation differs across inflammatory diseases within one cell type is unknown. Cross-disease comparison of monocyte eQTLs has not been performed with harmonized processing and power-calibrated specificity, and the ten-fold range of cohort sample sizes makes naive comparisons vulnerable to false specificity.Methods: We harmonized monocyte cis-eQTL summary statistics from five public resources spanning healthy, lupus-enriched, influenza-stimulated, and inflammatory bowel disease contexts; quantified power and false specificity by simulation; and decomposed effect sharing with multivariate adaptive shrinkage (mashR) and a variance-based specificity index.Results: In simulation, naive per-condition significance calls misclassify 72.6% of truly shared eQTLs as disease-specific; cross-condition borrowing reduces this to 13.0%, and a high-confidence specificity criterion to 0.005%. A three-condition pilot run (1,671,712 gene–variant pairs) passed all quality-control gates, and the full-scale seven-condition run (3,277,117 pairs) confirmed the patterns at scale across the healthy, SLE-cohort, and influenza axes: data-driven covariance structures dominated the fitted mixture (combined weight 0.9471), 2,737,644 pairs (83.5% of the strong set) remained significant under the joint model, and 16.87% of pairs were high-confidence specific under the tested-mask-aware criterion (16.90% after adjudication-layer MHC exclusion), within the pre-registered 20% ceiling under both configurations and in all 18 cells of the adjudication sensitivity grid, with MHC-region and 17q21.31 genes top-ranked. Magnitude sharing was highest within studies (within-study mean 0.859 versus cross-study mean 0.345; Randolph non-infected versus influenza-exposed 0.953, Perez classical versus non-classical 0.899). Pre-registered arbitration diagnostics (PCA and hierarchical clustering of the posterior mean matrix) grouped conditions by study of origin rather than by health-versus-disease status; under the pre-registered scenario-C rule, cross-cohort disease-specificity claims therefore contract to within-study paired contrasts. In the paired Randolph cohort—the single design in which disease state is unconfounded with cohort—influenza exposure and infection left the monocyte eQTL landscape nearly intact: sign concordance between the non-infected and infected states was 99.7% among jointly significant pairs, magnitude sharing was 0.89–0.95, and effect amplitudes were attenuated with infection (median posterior-effect ratio, non-infected over infected, 1.31), while 71,107 pairs (2.2% of the strong set) were high-confidence specific to the influenza axis. Conclusions: This study establishes and validates a power-calibrated framework for cross-disease regulatory genetics at full scale. Because disease context is confounded with cohort, ancestry, and platform in the current design and the pre-registered discriminating diagnostics show study-of-origin grouping, cross-cohort disease-specificity claims contract to within-study paired contrasts under the pre-registered rule. Within the one paired disease-state design available, influenza exposure and infection produce graded amplitude modulation on an almost completely shared regulatory backbone rather than wholesale regulatory rewiring; whether healthy-reference maps systematically miss disease-context regulation beyond such amplitude effects cannot be decomposed from cohort effects here. Disease-stratified heritability partitioning and locus-level nomination are pre-registered and in progress.