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

Abstract

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.