Fig. 1

Overview of the HiC-Reg framework. HiC-Reg makes use of 14 datasets: ten chromatin marks and four datasets related to transcription factor (TF) binding. The TF-binding datasets are factor-specific ChIP-seq datasets and sequence-specific motif counts of a TF in accessible regions. A 5 kb genomic region is represented by a vector of aggregated signals of either chromatin marks, accessibility, or TF occupancy. A pair of regions in HiC-Reg is represented using one of three types of features: PAIR-CONCAT, WINDOW, and MULTI-CELL. HiC-Reg uses Random Forests regression to predict the contact counts between pairs of genomic regions. Once trained, HiC-Reg takes as input a feature vector for a pair of regions and gives as output a predicted contact count for that pair (e.g., Count: 4).