QC parameters change after merging

[driesdewit]

Hi,

I am currently performing ATAC QC on 10X Multiome Objects (12 samples).
I have performed ATAC QC on sample level, using nCount_ATAC, FRIP, nucleosome signal etc.

However, after merging the different sample objects (according to the Merge vignette, using the GRanges reduce function), the number of fragments, and thus, my QC parameters change.

For example I had set a lower limit of 500 for nCount_ATAC, but after merging, barcodes that had previously passed that QC parameter now have less than 500.

What would be your advice? I am used to doing QC steps on sample level in scRNAseq, but would it here be necessary to repeat QC filtering after merging?

Thanks a lot!

[shayanjl]

This is expected for scATAC data. After merging, peaks or fragments can be recalculated in a way that changes derived QC metrics such as nCount_ATAC, so values may no longer match the sample-level numbers exactly. In practice, it is usually better to do the main QC at the sample level first, then merge, and only re-check QC after merging if you want to confirm that no cells now fall into problematic ranges. I would not treat the post-merge shift as automatically meaning the original filtering was wrong, but I would be consistent about which stage defines your final thresholds.

[driesdewit]

Thanks a lot!