Figure 6

Consistency of findings when different replicates or numbers thereof are analyzed. Ten reduced data sets were constructed by removing all possible combinations of 1 or 2 replicates of LPS-stimulated DC from the data set. The plot shows the cumulative frequency at which the probe sets are consistently selected in the 16iDC+LPS and in the ten reduced data sets by the following methods: the resampling approach applied to PLGEM-STN (filled squares), the permutation strategy applied to CLASSIC-STN (filled circles) and the SAM method (open triangles). In case of PLGEM, a single model was fitted on the common 16iDC baseline data set. The cumulative frequencies are normalized with respect to the total number of probe sets identified by the corresponding method.