Linking cis-regulatory regions using transcription factor binding signatures

Linking cooperatively functioning cis-regulatory elements (CREs), specifically enhancers and promoters, is a challenging task. Current strategies include correlation of expression of RNA transcribed from the CREs, experimentally measured chromatin interactions (Promoter Capture Hi-C) or machine learning based computational predictions. However, all three approaches require the availability of experimental data, which is sparse for most cells and tissues. We propose a new similarity metric to link enhancers to their target promoters based on transcription factor (TF)-binding “signatures”. TF-binding signatures are binary string representations (e.g. 0011001...), where each position indicates binding (“1”) or not (“0”) of a TF to a CRE. We apply a cosine similarity metric to enhancer-promoter pairs linked in published studies involving CRISPRi-FlowFISH, co-expression (FANTOM), or experimental tiling-deletion (CREST-seq). We find a significant difference between TF signature similarities of linked promoter-enhancer pairs compared to unlinked pairs. Furthermore we observe that TF-binding similarity scores are CRR specific. Based on the results, new directions are proposed that may allow further improvement towards a reliable mapping of interacting CREs across the genome.Science, Faculty ofGraduat