how to merge different peak sparse matrices

Question

[anlau62]

Hi,
I used 10X cellranger atac software to analysis scATAC-seq data,
but the software could only analysis one sample per time,
and produced separate filtered peak matrices for different tissues.

I want to do cell cluster upon all cells and and color different tissues.
So I should first to concat multiple sparse matrices, and I found some problems like the picture showing,
although the peak ids were different, but I thought they were the same peak.
If I concat the two matrices roughly, it could cause something wrong.
So maybe I should process something firstly, for example merging the two peaks to one peak id.

But I did not find useful information about how to merge scATAC sparse matrices,
how to judge which peaks should to be merged?
which peaks should to be filtered?
Bulabula……

Are there some papers or softwares to do it?

Hope someone help me to shoot this problem.

Answer 1

[aditi.qamra]

hey - you can't directly merge these matrices since the peaks are not exactly the same. 1) make a consensus peakset from the peak files called by 10x 2) get readcounts in the consensus peakset from all cells of all your samples -- that would be your final matrix of interest

[anlau62]

hi, thanks very much. can you suggest me some reference materials?