Sora Yoon

"Image-processing based detection of architectural stripes from chromatin conformation data"

Chromatin conformation capture techniques have revealed the existence of asymmetric structure in genome called architectural stripe. Former studies have suggested its role in transcription and recombination. However, the lack of a useful tool for stripe detection has made it difficult to study these architectural features systematically. The only existing tool called Zebra, developed for stripe detection, usually gives many false positive signals. Here, we devised a novel algorithm called Stripenn for stripe detection that implements the image processing technique. It can easily exclude the stripes of low quality using its scoring system. It is applicable for various 3D genomic contact data including Hi-C, HiChIP and micro-C. The stripe analysis with Hi-C and HiChIP data on immune cells showed that most of the stripes exist on the transcriptionally active regions and structural proteins are enriched at the anchor of the stripes. In addition, the topologically associating domains containing stripes were bound with significantly more structural proteins and enhancer marks and showed high DNA accessibility compared to non-stripy topologically associating domains. Genes that are highly expressed in these regions were enriched in immune responses unlike those in other regions. Finally, we found that the stripes were mostly conserved in T-cells of inbred strains of mice and the local change with the structural protein binding led to the large change in gene expression. In short, Stripenn enables the systematic study of architectural stripe and our finding adds more evidence to its role in cell-identity gene regulation.