Genome-wide association studies have shown that most common trait-associated variants fall into non-coding genomic regions and likely alter gene regulation (Maurano et al., 2012; Nica et al., 2010). This has motivated large-scale studies to catalog candidate regulatory variants (quantitative trait loci; QTLs) associated with various molecular phenotypes (i.e. quantitative molecular traits with a genomic location) across various populations (Lappalainen et al., 2013), cell (Fairfax et al., 2012) and tissue types (GTEx Consortium, 2015; Ongen et al., 2014). Mapping QTLs in this context usually consists of finding statistically significant associations between phenotype quantifications and nearby genetic variants; task commonly undertaken using linear regressions (GTEx Consortium, 2015). Alternative approaches have also been developed to increase discovery power by accounting for confounding factors (Fusi et al., 2012), integrating functional annotations (Gaffney et al., 2012), leveraging allelic imbalance (van de Geijn et al., 2015) or aggregating measurements across multiple tissues (Flutre et al., 2013). In practice, this requires millions of association tests in order to scan all possible phenotype-variant pairs in cis (i.e. variants located within a specific window around a phenotype), resulting in
