Error of FDR and S estimators. Comparisons were made between two groups of 2, 3 or 5 replicates. The first group is composed of synthetic data sets made from the experimental replicates Wt-t0a and Wt-t0b, in which 1000 genes (500 increased and 500 decreased) are changed (p <= 0.10), and the second group is composed of synthetic data sets without extra variation. Ten comparisons with different combinations of data sets were made, and the resulting mean errors of FDR (upper panel) and S (lower panel) for several values of the estimator (5 to 90%, by steps of 5%) are plotted (error is computed by substracting the estimated value from the real value and expressed in relative percentage point difference). Replicates 2, 3 and 5 are marked, respectively, by blue circles, green triangles and red squares.