ENCODE TF ChIP-seq data analysis

We downloaded 2,216 ChIP-seq experiment data from the ENCODE Project. The list of the data is in Supplementary Table S8. The data were lifted over from hg19 to hg38. We found overlapping peaks on four different categories: (1) 500bp upstream the promoter region of pcRNA-associated coding genes, (2) 500bp upstream promoter region of pcRNAs, (3) pcRNA genomic loci, and (4) pcRNA genomic loci but not overlapping with promoter region. To understand the correlation of TF binding patterns in the four categories, we made a binary matrix per category that consists of rows of TFs and columns of pcRNA/coding genes. Hence, the matrix contains connections between TF and pcRNA/associate coding genes. The matrix of category 2 is clustered by Euclidian Distance. To check the extent to which promoter sharing or proximity determines TFBS correlation, we also separated the clustered heat-map in the pcRNA bidirectional transcript (BIDIR) subgroup to the other subgroups (Non-BIDIR). To directly compare the TF binding patterns between each category, the other three matrices were sorted by the same order of the clustered matrix. We used the MatLab function corr2 to calculate r-value between category (1) and (2). We performed Monte Carlo simulation to calculate the p-value and test the significance of the r-value.