All data are downloaded from ENCODE project (https://www.encodeproject.org/)¹. And wig file, quantile normalization and peak calling were performed as described in fig1. Cell type specific sub networks were constructed as described in Fig 6. Cell identity genes and identity drivers were predicted as described in Fig 5 and Fig 6. Pathway enrichment analysis was performed using (DAVID 6.8 https://david.ncifcrf.gov/)⁶. For heatmap and hierarchy cluster analysis, top 500 putative identity genes were collected and unioned together first, and then heatmap were created by using morpheus from Broad institute (https://software.broadinstitute.org/morpheus/) by using the default parameters.