scHiCyclePred: a deep learning framework for predicting cell cycle phases from single-cell Hi-C data using multi-scale interaction information

Abstract

While scRNA-seq offers gene expression snapshots, it misses the spatial context of chromatin organization crucial for cell cycle regulation. Single-cell Hi-C, capturing chromatins three-dimensional (3D) architecture, fills this void, revealing interactions between genomic regions that transcript-only data might overlook. We introduce scHiCyclePred, a model that utilizes single-cell Hi-Cs multi-scale interaction data to predict cell cycle phases by extracting chromatins 3D features. This fusion-prediction model integrates three feature sets into a unified vector. Remarkably, scHiCyclePred outperforms methods like NAGANO and CIRCLET and traditional machine learning techniques across various metrics. Our insights into 3D chromatin dynamics during the cell cycle further underscore its utility. By offering a more comprehensive view of cell cycle dynamics through chromatin structure, scHiCyclePred stands to significantly advance our understanding in cellular biology and holds potential to catalyze breakthroughs in disease research. Access scHiCyclePred at github.com/HaoWuLab-Bioinformatics/scHiCyclePred.