The prognosis prediction performance.

<p>The comparison of AUC (A) and accuracy (B) for three datasets: Different coloring schemes and shape indicate three independent datasets (orange circle: German dataset; blue diamond: Barrier dataset; green square: GSE5206 dataset). TX_Y methods (X: top 500 or 1000 MDMs; Y: 10 or 18 reference tumors or Leave-One-Out method (LOO)). The filled symbols denote the mean of AUCs; The comparison of accuracies(C), sensitivities (D) and specificities (E) for prognosis prediction between our method and present methods with same datasets, including the LOO results from Lin07 (L) <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0033653#pone.0033653-Lin1" target="_blank">[3]</a>, Garman08 (G) <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0033653#pone.0033653-Garman2" target="_blank">[42]</a>, Barrier06 (B) <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0033653#pone.0033653-Barrier1" target="_blank">[5]</a>, and also the Barrier06's results obtained using 34 tumors (TS34), 18 tumors (TS18) or 10 tumors(TS 10) as the training set. The filled symbols are mean value. *The points in the dotted circle are the outcomes from the methods which were validated using makers discovered by the one and the same dataset.