Figure 1

Analysis of the false discovery rate of the signaling circuit test proposed. A) Results of the simulation experiment. A total of 1000 datasets, of n = 10, 20, 50 and 100 samples were simulated. All the probes for all the genes in all the nodes of the pathways had the same probability of being activated (0.1, 0.3, 0.5, 0.7 and 0.9). Normal distributed (mean = 0, SD = 0.05 noise was added to each probe). At the conventional p-value of 0.05 chosen for the test, the ratios of false positives are negligible. B) Results with a real dataset of pediatric acute myeloid leukemia (AML) which contains gene expression microarray data of 237 samples, which (in principle) should display a similar gene expression profile. We produced 1000 datasets, of n = 10, 20, 50 and 100 samples, by randomly subsampling the 237 samples dataset. Any of the datasets was divided into two equal parts and compared to each other with the test. Again, the number of false positives is negligible.