31.2: eQTL Basics

Cis-eQTLs

The use of whole genome eQTL analysis has separated eQTLs into two distinct types of manifestation. The first is a cis-eQTL (Figure 31.3) in which the position of the eQTL maps near the physical position of the gene. Because of proximity, cis-eQTL effects tend to be much stronger, and thus can be more easily detected by GWAS and eQTL studies. Often, these function as promoters of certain polymorphisms, affect methylation and chromatin conformation (thus increasing or decreasing access to transcription), and can manifest as insertions and deletions to the genome. Cis-eQTLs are generally classified as variants that lie within 1 million base pairs of the gene of interest. However, this is indeed an arbitrary cutoff and can be altered by an order of magnitude, for instance.

Trans-eQTLs

The second distinct type of eQTL is a trans-eQTL (Figure 31.4). A trans-eQTL does not map near the physical position of the gene it regulates. Its functions are generally more indirect in their effect on the gene expression (not directly boosting or inhibiting transcription but rather, affecting kinetics, signaling path- ways, etc.). Since such effects are harder to determine explicitly, they are harder to find in eQTL analysis; in addition, such networks can be extremely complex, further limiting trans-eQTL analysis. However, eQTL analysis has led to the discovery of trans hotspots which refer to loci that have widespread transcriptional effects [11].

Perhaps the biggest surprise of eQTL research is that, despite the location of trans hotspots and cis-eQTLs, no major trans loci for specific genes have been found in humans [12]. This is probably attributed the current process of whole genome eQTL analysis itself. As useful and widespread whole genome eQTL analysis is, we find that genome-wide significance occurs at \(p=5 \times 10^{-8}\) with multiple testing on about 20,000 genes. Thus, studies generally use an inadequate sample size to determine the significance of many trans-eQTL associations, which start with priors of very low probability to begin with as compared to cis-eQTLs [4]. Further, the bias reduction methods described in earlier sections deflate variance, which is integral to capture the microtrait associations inherent in trans loci. Finally, non-normal distributions limit the statistical significance of associations between trans-eQTLs and gene expression[4]. This has been slightly remedied by the use of cross-phenotype meta-analysis (CPMA)[5] which relies on the summary statistics from GWAS rather than individual data. This cross-trait analysis is effective because trans-eQTLs affect many genes and thus have multiple associations originating from a single marker. Sample CPMA code can be found in Tools and Resources.

However, while trans loci have not been found, trans-acting variants have been found. Since it can be inferred trans-eQTLs affect many genes, CPMA and ChIP-Seq can be used to detect such cross-trait variants. Indeed, 24 different significant trans-acting transcription factors were determined from a group of 1311 trans-acting SNP variants by observing allelic effects on populations and target gene interactions/connections.