Figure 3.

The accuracies of models built with different features. (A), (B) and (C) show kappa, SS and 1 - SP for models built with different features, respectively. ChIP: ChIP-chip data, pre: predefined motifs, auto: automatically identified motifs using AlignACE. The first cross-validation method was used for models in autol and the second cross-validation method was used for models in auto2 (see text). White bar: average accuracy of models for true DEGs. Grey bar: average accuracy of models for random genes. The error bars represent individual 95% confident interval for the means. (D) and (E) show kappa values for models built with different types of features for stress conditions and cell cycle conditions, respectively. X-axis: kappa for models built with predefined motifs; y-axis: kappa for models built with a combination of ChIP-chip data and predefined motifs.