Abstract
Molecular QTL studies quantify whether genetic variants affect molecular traits, but non-linear effects including distributional patterns, variance, and interactions provide mechanistic insights beyond mean-level associations. Methods for detecting distributional effects have been developed for eQTL analysis, yet applications have focused on method demonstrations rather than large-scale biological discovery. We comprehensively mapped quantile, variance, and interaction QTLs across 34 data-set from 22 molecular contexts in >2,300 human brain donors, revealing that 48.7% of quantile QTLs (qQTLs) exhibit context-dependent regulation invisible to linear models, with enrichment at phenotypic extremes and in cell-type-specific regulatory elements, chromatin accessibility regions, and long-range chromosomal contacts. qQTL variants explained additional trait heritability beyond linear QTLs for brain-related traits. At Alzheimer's disease (AD) risk loci, qQTL analysis revealed complex regulatory architecture including variance effects at PITRM1, lower-quantile-specific effects at TMEM106B partially explained by APOE ε4 interactions, and coordinated epigenetic regulation at loci harboring CHRNE/SCIMP/RABEP1. Quantile-based transcriptome-wide association studies identified 34 AD risk genes and additional aging-related genes beyond standard TWAS, with enrichment in immune regulation and telomere maintenance pathways where distributional effects may reflect threshold-dependent mechanisms. Our non-linear QTL atlas and qTWAS resource enable characterization of context-dependent regulatory effects in complex disease genetics.