Skip to main content
. 2016 Dec 20;11:66. doi: 10.1186/s13062-016-0169-7

Fig. 4.

Fig. 4

Empirical False Discovery Rate (eFDR). We estimate an eFDR by dividing the number of differential expression calls in inter-site A-vs-A and C-vs-C comparisons by the number of calls in A-vs-C comparisons. The left panel shows the original results for q-value thresholding only (no additional processing or filters). In the middle panel, hidden confounders have been removed by svaseq. In the right panel, additional filters have also been applied. Plots for various methods of expression estimation are shown along the x-axis, with methods for differential expression calling indicated by colour