New to ATAC-seq single cell

Question

[zizigolu]

Hello

I have some ATAC-seq single cell

I know I should start with quality check of fastq files, alignment with bowtie2, sorting with samtools

I think the next is peak calling which can be done by MACS2 or Genrich

Seurat does need fragment file and raw read counts matrix

I want to ask you fragment.tsv file and raw read counts matrix are the output files of which software and which step in ATAC-seq single cell??

Google made me more confused

Thank you

Answer 1

[Friederike]

The steps you described are typically done for bulk ATAC-seq. If you want to treat your scATAC-seq samples as bulk ATAC-seq samples, you can do the steps you described above and you won't need Seurat, which is specialized for scATAC-seq.

For keeping the single-cell information in the data set, you will need to use a slightly different set of tools. If your data was generated on the 10X Chromium platform, you could make use of 10X' CellRanger pipeline, either for scATAC-seq alone or, in case you also have scRNA-seq to go with the scATAC-seq, you can try their ARC pipeline. Both pipelines will result in the types of files that the Seurat people will refer to.

If your data was not generated on a 10X Chromium platform, you will need to play around with other tools. There are several papers about scATAC-seq processing that list tools and their capabilities: Baek et al. (2020, Chen et al. (2019). SnapATAC seems to try to offer a one-for-all pipeline and ArchR also starts with BAM files.

[zizigolu]

Sorry

I have two questions please

1- Can not I simply use a chip seq pipleline like using MASC2 peak caller (using different argument specific to single cell) instead of what you kindly mentioned here?

2- If my data is 10X Chromium platform, I again can use ArchR / SnapATAC or just for non 10X Chromium platform data I can use these tools?

[Friederike]

1- Can not I simply use a chip seq pipleline like using MASC2 peak caller (using different argument specific to single cell) instead of what you kindly mentioned here?

As I wrote above: if you want to treat your scATA-seq data as bulk ATAC-seq data (i.e. you ignore the individual cell barcodes and simply sum up all reads per genome bin), then MACS2 could work. If you want to retain the single-cell information, try to do the thought experiment of how to run MACS2. You'd essentially have to create BAM files for every single cell and run MACS2 on every single file. Typical scATAC-seq experiments with thousands of cells would become pretty unwieldy this way (plus I doubt that MACS2 could handle peak calling with only 200-500K reads instead of the usual >1mio reads per BAM file).

If my data is 10X Chromium platform, I again can use ArchR / SnapATAC or just for non 10X Chromium platform data I can use these tools?

If you have the FASTQ files, you can probably use ArchR/SnapATAC, too.