DGW: an exploratory data analysis tool for clustering and visualisation of epigenomic marks

Novel technologies such as ChIP-Seq and DNAse-Seq have enabled scientists to gather ever increasing amounts of data on epigenetic modifications in various cell types and conditions. Epigenomic marks (e.g. histone marks) are often distributed over large (several Kb) genomic regions, and display non-trivial structures, such as multimodality and plateaus, which may be indicative of biologically relevant features, such as nucleosome displacement or interaction with cofactors. Standard clustering and visualisation techniques developed for microarrays are not immediately transferable to the epigenomic scenario though: peaks have different lengths, the data is digital, and the noise distribution is not fully understood. Here, we propose a simple method for hierarchical clustering of epigenomic marks based on Dynamic Time Warping, a popular technique in signal processing which locally stretches/ compresses two signals in order to best match their shape. We implement the method in an open source Python package, and demonstrate its working on simultaneous clustering of multiple histone marks.