Promoter activity classification performance. Receiver-Operator Curves (ROC) were calculated for different 13-bin based classifiers. We made pair wise classifications between active promoters, inactive promoters and randomly selected genomic regions. The performance can be measured by the area under the curve (AUC), where AUC = 0.5 corresponds to a random guessing. Active promoters are easier to distinguish from random genomic background than inactive promoters. Adding additional features (dinucleotide content, DNA methylation status), or only using the positional distribution of marks did not have a substantial impact.