Calculated Normalization Error (E) for single and multiple reference genes. Normalization error values (E) were calculated as the ratio of the ratio of two reference genes in two different samples as described in Vandesompele et. al.  with modifications (see materials and methods) and are shown as cumulative percent distribution for each sample type. The graph shows independent measurements in Whole Blood and PBMC: Whole Blood, PPIB vs. PGK1-blue solid line; PBMC, RPLP0 vs. PGK1-red dashed line. These overlap with the comparisons to the calculated geometric mean (GM) of the two most stable reference genes: Whole Blood, PGK1 vs. GM (PPIB and EAR)-black dashed line or PBMC, PGK1 vs. GM (RPLP0 and PPIB)-grey dashed line. For these within-tissue comparisons, the fold error for 90% of the samples is 1.7 or less. In the cross-tissues analysis where PBMC was compared to Whole Blood, the fold error for 90% of the samples is much higher at 2.7. The fold-difference between PBMC vs. Whole Blood normalized by PGK1 is shown as a solid green line, the geometric mean of PPIB and PGK1 is a dashed purple line, and the geometric mean of YWHAZ and RPLP0 is shown as a solid black line.