We investigated six strategies: traditional and PC-adjusted expression analysis, each combined with three marker panels (Illumina 300K genotype, HapMap2 imputation, and 1000G imputation). For each strategy, we estimated the significance threshold for 5% FDR accounting for the number of all transcript–SNP pairs. We then counted the number of transcripts for which at least one SNP within 1 Mb exceeds that threshold as an indicator of the power of that strategy (Fig. 4).
