Figure 4

Results of TSN classification on cancer datasets. Results of 100 rounds of 5-fold cross validation over a range of N = {2,3,4} where the number of differentially expressed probes is different for each value of N {16,10,9}. This yields approximately the same number of possible combinations for each value of N (~120), illustrating how classification accuracy can be determined by the permutation itself, not just the number of combinations available. Results shown include accuracies of fixed values of N as well as the dynamic N algorithm described in the methods section. Statistical differences were calculated using the nonparametric Kruskal-Wallis one-way analysis of variance by ranks, and a p-value < 0.05 was considered significant. If bars share the same letter they are not statistically different. The datasets are derived from[2] and represent a wide range of cancers. Significance plots for all nine cancer datasets are in Additional file1: Figure S4.