scART: recognizing cell clusters and constructing trajectory from single-cell epigenomic data

Abstract

SUMMARYThe development of single-cell assay for transposase-accessible chromatin using sequencing data (scATAC-seq) has allowed the characterization of epigenetic heterogeneity at single-cell resolution. However, the sparse and noisy nature of scATAC-seq data poses unique computational challenges. To address this, we introduce scART, a novel bioinformatics tool specifically designed for scATAC-seq data analysis. scART utilizes analytical methods highly stable for processing sparse and noisy data, such as k-nearest neighbor (KNN) imputation, Term Frequency-Inverse Document Frequency (TF-IDF) weighting scheme, and the cosine similarity metric to identify underlying cellular heterogeneity in scATAC-seq data. It accurately and robustly identifies cell identities, particularly in data with low sequencing depth, and constructs the trajectory of cellular states. As a demonstration of its utility, scART successfully reconstructed the development trajectory of the embryonic mouse forebrain and uncovered the dynamics of layer-specific neurogenesis. scART is available at GitHub.