Stripenn detects architectural stripes from chromatin conformation data using computer vision

Abstract

Architectural stripes tend to form at genomic regions harboring genes with salient roles in cell identity and function. Therefore, the accurate identification and quantification of these features is essential for the understanding of lineage-specific gene regulation. Here, we present Stripenn, an algorithm rooted in computer vision to systematically detect and quantitate architectural stripes from chromatin conformation measurements of various technologies. We demonstrate that Stripenn outperforms existing methods, highlight its biological applications in the context of B and T lymphocytes, and examine the role of sequence variation on architectural stripes by studying the conservation of these features in inbred strains of mice. In summary, Stripenn is a computational method which borrows concepts from widely used image processing techniques for demarcation and quantification of architectural stripes.