Reconstruction of enhancer–target networks in 935 samples of human primary cells, tissues and cell lines

Kevin Yip and colleagues report a method for determining the target genes of enhancers in specific cells and tissues by combining global trends across many samples with sample-specific information, and considering the joint effect of multiple enhancers. They apply their method to reconstruct enhancer–target networks in 935 samples of human primary cells, tissues and cell lines. We propose a new method for determining the target genes of transcriptional enhancers in specific cells and tissues. It combines global trends across many samples and sample-specific information, and considers the joint effect of multiple enhancers. Our method outperforms existing methods when predicting the target genes of enhancers in unseen samples, as evaluated by independent experimental data. Requiring few types of input data, we are able to apply our method to reconstruct the enhancer–target networks in 935 samples of human primary cells, tissues and cell lines, which constitute by far the largest set of enhancer–target networks. The similarity of these networks from different samples closely follows their cell and tissue lineages. We discover three major co-regulation modes of enhancers and find defense-related genes often simultaneously regulated by multiple enhancers bound by different transcription factors. We also identify differentially methylated enhancers in hepatocellular carcinoma (HCC) and experimentally confirm their altered regulation of HCC-related genes.