Figure 3.

Performance comparison according to characteristics of knowledge bases. ROC curves of true and false positive rates for the analysis using (A) direct (YPD) and indirect (ChIP-chip) prior knowledge, (B) different sizes of knowledge bases with KBR SVM (solid) and KBR cutoff (dashed) methods, and (C) different qualities of knowledge bases by adding different amount of false positives to the knowledge base with KBR SVM (solid) and KBR cutoff (dashed) methods. True positive rate is the ratio of samples determined as true from the positive test set, and false positive rate is that from the negative test set