Fig. 1

Schema of the methodology used for detecting genome segments with an epigenetic pattern resembling a sample classification. For a given sample, different epigenomic markers are quantified in a genome-wide fashion (a). For a given region of the genome, all these markers are collapsed into a single epigenetic state (colors) using ChromHMM (b). This is done for hundreds of different samples (c). These samples can be classified according with different criteria (into three groups in this example: brown, pink and blue). The epigenetic pattern of a given segment of the genome (e) is compared with an equivalent pattern representing this sample classification (f) using a “mutual information” based approach. One of the patterns is shuffled thousands of times in order to generate a null distribution of MI scores from where to extract a p-value for the MI score of the segment of interest (g). The process is repeated for all other windows in the human genome (d). The genome segments with the highest MI values and significant p-values are taken as those related to the sample classification (h). Lung clipart source: Wikimedia Commons (http://commons.wikimedia.org/)