Destin: toolkit for single-cell analysis of chromatin accessibility

Abstract

Summary Single-cell assay of transposase-accessible chromatin followed by sequencing (scATAC-seq) is an emerging new technology for the study of gene regulation with single-cell resolution. The data from scATAC-seq are unique—sparse, binary and highly variable even within the same cell type. As such, neither methods developed for bulk ATAC-seq nor single-cell RNA-seq data are appropriate. Here, we present Destin, a bioinformatic and statistical framework for comprehensive scATAC-seq data analysis. Destin performs cell-type clustering via weighted principle component analysis, weighting accessible chromatin regions by existing genomic annotations and publicly available regulomic datasets. The weights and additional tuning parameters are determined via model-based likelihood. We evaluated the performance of Destin using downsampled bulk ATAC-seq data of purified samples and scATAC-seq data from seven diverse experiments. Compared to existing methods, Destin was shown to outperform across all datasets and platforms. For demonstration, we further applied Destin to 2088 adult mouse forebrain cells and identified cell-type-specific association of previously reported schizophrenia GWAS loci. Availability and implementation Destin toolkit is freely available as an R package at https://github.com/urrutiag/destin. Supplementary information Supplementary data are available at Bioinformatics online.

scBFA: modeling detection patterns to mitigate technical noise in large-scale single cell genomics data

“…While single cell genomic data from different modalities such as scATAC-seq have similar data structure as scRNA-seq data, the analysis tools and pipelines developed to date for those two technologies are largely mutually exclusive. Here we show that scBFA generalizes to other single cell genomic modalities and outperforms existing methods for cell type identification for scATACseq datasets as well, even those that take advantage of auxiliary data such as distance to transcription start sites 35 . We expect our results to generalize to other single cell genomic modalities such as single-cell methylation or histone modification data.…”

“…We compared scBFA against PCA, Binary PCA, Scasat 34 , Destin 35 and scABC 36 , Scasat and Destin are scATAC-seq analysis tools primarily designed to identify cell types and differential accessibility analysis. Both methods treat dimensionality reduction as prior step before further clustering distinct cell types.…”

“…We performed dimensionality reduction and cell type classification experiments on several scATAC-seq datasets, analogous to our scRNA-seq analyses above. We benchmarked scBFA against PCA, Binary PCA, Scasat 34 , Destin 35 and scABC 36 . scBFA systematically outperformed all other methods in our benchmark datasets ( Fig.…”

“…We followed the scATAC-seq pipeline for processing and aligning reads used by the Destin method 35 , obtained from GitHub at https://github.com/urrutiag/destin on April 22, 2019. This preprocessing pipeline yielded 2779, 576, and 960 BAM files for GSE96769, GSE74310 and GSE107816, respectively.…”

scATAC-pro: a comprehensive workbench for single-cell chromatin accessibility sequencing data

“…Both chromVar [10] and single-cell regulome analysis toolbox, SCRAT [11] work with preprocessed data and only report loss or gain of chromatin accessibility on a set of predefined genomic regions, which ignores a large amount of information in the data. Detection of cell-type specific difference in chromatin accessibility, Detin [12], single cell accessibility based clustering, scABC [13] and cisTopic [14], focus on identifying cell populations and/or differential accessible regions given the processed data like bam files or in peak-by-cell count matrix.…”