There have been many different efforts to improve dimensionality reduction methods for scATAC-seq data, particularly considering that it is a relatively underexplored datatype when compared to scRNA-seq. With many different options, it can be somewhat confusing to follow the differences and pros/cons behind each method. The purpose of this post is to:
In contrast to most single-cell RNA-seq analyses, there are many regular tasks when working with single-cell ATAC-seq (scATAC-seq) data, such as peak calling and plotting pileups, that require the BAM file (or something similar) containing some representation of the actual reads. Unlike bulk ATAC-seq data, in which your samples are static and will not change, “samples” in single-cell datasets will consist of groups of similar cells, typically arrived at via clustering. This is often an exploratory and iterative process, so your “samples” can be subject to frequent changes. You might even want to group your cells by other metadata fields entirely.
Read More