However, the exceedingly sparse nature of single cell epigenomic data sets present unique and significant computational challenges. All single-cell epigenomic methods are intrinsically sparse, as the total potential signal at a genomic locus is fundamentally limited by the copy number of DNA, thus generating 0, 1 or 2 reads from regulatory elements within a diploid genome. Methods developed for single cell RNA-seq have shown that measuring the dispersion of gene sets, such as Gene Ontology or co-expression modules, rather than individual genes can be a powerful approach for analyzing sparse data7. In this vein, and building on previous work from our group and others4,8,9, we have developed chromVAR, a versatile R package for analyzing sparse chromatin accessibility data by measuring the gain or loss of chromatin accessibility within sets of genomic features while controlling for known technical biases in epigenomic data (Supplementary Software). We show that chromVAR can be used to identify transcription factor (TF) motifs that define different cell types and vary within populations, providing a unique analytical toolkit for analysis of sparse epigenomic data.
