97 / 2024-10-22 20:21:50

Analysis of 3D genome structure based on artificial intelligence

enhancer-promoter interactions, DNA replication timing, machine learning, acute lymphoblastic leukemia

Differentiating between the gene targets influenced by distal regulatory elements and those of other nearby transcribed genes is a challenging problem. However, identifying this distinction holds promise for revealing the underlying causal basis of complex diseases. Despite advances in recent high-throughput experimental methods that facilitate reconstruction of the regulatory landscape across the genome, it remains largely unclear to what extent DNA replication contributes to guiding enhancer-promoter interactions. Here, we present RepliChrom, a novel computational method that predicts enhancer-promoter interactions from the perspective of DNA replication timing, offering a unique approach to understanding three-dimensional chromatin conformation interactions. Assessing these features revealed replication timing conservation within the same cell types and specificity across different cell types. Our results across six cell types demonstrate that replication timing features effectively predict enhancer-promoter interactions. As a proof-of-principle, we applied RepliChrom to identify interactions from various chromatin conformation capture technologies, such as Hi-C, Hi-TrAC, ChIA-PET, and 5C. Moreover, we leveraged RepliChrom to screen for significant chromatin interactions in acute lymphoblastic leukemia samples, differentiating them precisely from normal samples. This work uncovers that replication timing signals shape the three-dimensional structure of fine-grained regulatory interactions.