Sequence Polymorphisms Cause Many False cis eQTLs

Abstract

Many investigations have reported the successful mapping of quantitative trait loci (QTLs) for gene expression phenotypes (eQTLs). Local eQTLs, where expression phenotypes map to the genes themselves, are of especially great interest, because they are direct candidates for previously mapped physiological QTLs. Here we show that many mapped local eQTLs in genetical genomics experiments do not reflect actual expression differences caused by sequence polymorphisms in cis-acting factors changing mRNA levels. Instead they indicate hybridization differences caused by sequence polymorphisms in the mRNA region that is targeted by the microarray probes. Many such polymorphisms can be detected by a sensitive and novel statistical approach that takes the individual probe signals into account. Applying this approach to recent mouse and human eQTL data, we demonstrate that indeed many local eQTLs are falsely reported as “cis-acting” or “cis” and can be successfully detected and eliminated with this approach.

Detection, Validation, and Downstream Analysis of Allelic Variation in Gene Expression

“…As has been established for several years, SNP position and the number of SNPs overlapping the microarray probes have a significant impact on the discovery rate and the size of apparent expression differences for cis- (Schadt et al 2003;Alberts et al, 2005Alberts et al, , 2007Doss et al 2005;Chen et al 2009) and trans-modulated transcripts (Chen et al 2009). Our analysis found that the longer Illumina 50-mers have approximately the same sensitivity to sequence differences as the shorter Affymetrix 25-mer probes ( Figure S1 and Figure S2).…”

“…We used QTL Reaper and a more refined linkage analysis of individual probe data-the single 25-mers as opposed to the full probe set-and examined the consistency of linkage across individual probes and those generated by the Affymetrix full probe set. Alberts et al (2007) developed a different approach of identifying problematic probes by decomposing the signal provided by the probe set and analyzing the deviation of the individual probes from the entire set. A computational protocol that flags probes that may overlap SNPs is freely available in R (Alberts et al 2008).…”

“…While this study is by no means the first to point out the potential difficulty of sorting out sources of variation in array hybridization (Schadt et al 2003;Alberts et al 2005Alberts et al , 2007Alberts et al , 2008Doss et al 2005;Chen et al 2009), we have extended previous work in three ways. First, we provide a comparatively quantitative assessment of the severity of the problem for two of the preeminent array platforms (Affymetrix and Illumina) by exploiting several large expression data sets and allele-specific expression protocols.…”

Chemotherapeutic drug susceptibility associated SNPs are enriched in expression quantitative trait loci

“…It has been shown that SNPs in probes can produce false eQTL signals by affecting hybridization efficiency (31). We therefore downloaded SNP data from dbSNP (build 129, human genome assembly build 36) to identify probes in probe sets that hybridize to SNP-containing regions.…”

Ancestry-Related Differences in Gene Expression: Findings May Enhance Understanding of Health Disparities Between Populations

“…There were also other technical differences in the analysis. First, we made an attempt to address the potential confounding factor of SNPs in probes [23], while other studies did not take this potential problem into account in their analyses. Second, although, generally speaking, the microarray data are consistent with other experimental quantitative techniques and across microarray platforms [24,25], the Affymetrix GeneChips® and Illumina BeadArrays used in these studies have differences in probe-set design, gene coverage and probe-set replication.…”