We estimated PCs from the gene expression values in the family panels, and progressively included the top PCs in the eQTL regression model as covariates until the number of transcripts mapped locally no longer increased. In the final models, we included the top 69 PCs in the MRCA (Affymetrix) expression data set and the top 61 PCs in the MRCE (Illumina) data set. Including data on batch and experimental conditions in the model did not change the findings, indicating that the PCs adjusted well for these factors. Overall, including PCs in the MRCA model identified eQTLs for 7406 probesets (6219 mapped locally) compared with 2432 probesets (1975 mapped locally) in the original data set (Table 1).
