Differential histone occupation and modifications between ChIP-chip verified TFBSs and non-verified motif matching sites. Showing SWI4 as an example, most histone modifications (in different colors) are significantly different between ChIP-chip verified TFBSs (left boxes), which have binding motifs and are bound by TFs in ChIP-chip experiments, and non-verified motif matching sites (right boxes), which have matching motifs but are not bound by TFs.

Model parameters. (a) Stricter thresholds for target gene calling result in better predictions. (b) Combination of independent histone modification data sets can improve target predictions. Predictions for 203 TFs are evaluated by area under the receiver operator characteristic (ROC) curve (AUC) for data from Pokholok et al. [23] and Kurdistani et al. [22]. (c, d) Using histone modifications within a 500-bp (c) and 1, 000-bp (d) window upstream and downstream of ATG sites of target genes achieves a similar performance to using only upstream signals, and better performance to using only downstream signals. There was no significant performance difference between using 500-bp and 1, 000-bp window sizes. (e) Using histone modifications in intergenic regions has better predictive power than using those in ORF coding regions.

Target histone modification profiles and target-non-target differential modification profiles of TFs. (a) Target histone modification profiles comprising different normalized histone modification signals (columns) of TFs (rows). A target histone modification profile is the averaged histone modification signals of the TF's targets. TFs are clustered into histone-sensitive (blue bar) and -insensitive TFs (orange bar) using their target histone modification profiles. Histone-sensitive TFs have stronger histone modification signals. (b) Target-non-target differential modification profiles of TFs showing the discriminating power (t-statistic) of histone modifications to TF targets and non-targets. TFs are ordered the same as in (a). Histone-sensitive and -insensitive TFs have distinct differential modification profiles, indicating preferential histone modifications of TFs targets. (c) Correlation network of histone modifications in terms of TF differential modification profiles. Histone modification pairs with a correlation coefficient larger than 0.5 (red edges) or smaller than -0.5 (green edges) are connected. The network shows a high level of redundancy of histone modifications in differential modification profiles.

Chromatin modifications improve STE12 binding site prediction. (a, b) The AUC (a) and positive predictive value (b) of the histone-only model (HM), the PSSM-only model (PWM) and the combined model (PWM+ HM) for STE12 binding site prediction.

Chromatin modifications substantially improve TF target gene predictions. (a) ROC curves show improved TF target gene predictions using histone modifications. (b, c) Performance of prediction models for individual TFs, with PSSMs from Beer and Tavazoie [20] (b) and from Harbison et al. [5] (c). TFs are sorted by prediction performance using histone modifications and PSSMs (red bars).

Condition specificity of the chromatin model for TF target prediction. (a, b) ROC curves showing the performance of two chromatin models in predicting target genes of HSF1 (a) and the MSN2/4 complex (b). The model (model B) using histone modifications in the H₂O₂ condition is better at predicting target genes in the H₂O₂ condition (blue curve) than in YPD medium (model A; red curve), which indicates condition specificity of chromatin modifications and TF target genes.

Distinctions between histone-sensitive and -insensitive TFs. (a) Target genes of histone-sensitive TFs are better predicted using intergenic or upstream histone modifications than are those of histone-insensitive TFs. (b) Histone-sensitive TFs have a smaller number of target genes in the regulatory network than histone-insensitive TFs. (c) Histone-sensitive TFs have a larger number of interaction partners in the protein-protein interaction (PPI) network. (d) Higher mRNA expression levels of histone-sensitive TFs.