Finally, if we assume that many of the cis eQTLs in the AA and HA were caused by PS and consequently excluded by eigenvector adjustment, the false trait-SNP associations should bear the following properties: (1) the gene expression trait should be differentially expressed between subpopulations, and (2) the SNP allele frequency should be different between the subpopulations.
An sRNA with RPM value ≥0.6 was regarded as expressed, and an sRNA identified as expressed in more than 25 of the 98 IMF2s was regarded as an sRNA expression trait (s-trait).
Correspondingly, the expression trait value would also be different between the two subgroups.
We subsequently fit linear mixed models to each expression trait (20,400) and putative QTL position.
For eQTL analyses, the gene expression trait values were first pre-adjusted with gender as a covariate in a linear model and the gender adjusted gene expression trait residuals used as the dependent variable in the QTL mapping.
Strikingly, we found no cases where both ER and GR of an expression trait mapped to the same marker, using the genome-wide significance cutoffs described above.
A nominal p-value was used to test the null hypothesis of no eQTL at each position and expression trait combination.
If M1 is selected as the best model we conclude that the data suggests that the gene expression trait is supported as causal for the clinical trait [5].
The summary value for each expression trait (as described in [5]) was used in the generation of the eQTLs upon which this analysis was performed.
For each GWAS locus, we examined whether the lead/tag SNP correlated with any expression trait in GHS data and when a significant association was found, we checked whether the expression trait was significantly associated with the risk factor under consideration.
In addition, our models are constructed to account for the contribution of nearby trans-acting QTL that occur on the same chromosome as the expression trait of interest, thus producing linkage that mimics cis-acting QTL.
