mitochondrial percentage calculation and other metrics

[kinali]

Dear all,

I do not have any sc-atac cell data analysis experience and I am trying to follow the tutorial.
I would like to learn how "mitochondrial" feature is calculated in meta.data and how can I calculate mitochondrial fragment percentage by using it.

I see general cut-off numbers for quality metrics such as fraction_of_fragments_in_peaks or tss_enrichment_score on this website:
https://portal.hubmapconsortium.org/docs/assays/atacseq
and I am wondering is there any general cut-off for the peak_region_fragments.

Also, I am wondering is there any documentation that explains features in the meta.data clearly? Although I checked the Signac paper and R documentation I cannot understand the below bold features and how should I filter them like what should be the max, min, or expected values.

[1] "orig.ident" "nCount_peaks"
[3] "nFeature_peaks" "total"
[5] "duplicate" "chimeric"
[7] "unmapped" "lowmapq"
[9] "mitochondrial" "passed_filters"
[11] "cell_id" "is__cell_barcode"
[13] "TSS_fragments" "DNase_sensitive_region_fragments"
[15] "enhancer_region_fragments" "promoter_region_fragments"
[17] "on_target_fragments" "blacklist_region_fragments"
[19] "peak_region_fragments" "peak_region_cutsites"
[21] "nucleosome_signal" "nucleosome_percentile"
[23] "TSS.enrichment" "TSS.percentile"
[25] "pct_reads_in_peaks" "blacklist_ratio"

I would appreciate any help, thank you!

[timoast]

Many of the metrics you've listed here come from the 10x Cellranger metadata, see the 10x website for documentation: https://support.10xgenomics.com/single-cell-atac/software/pipelines/latest/output/singlecell

I am wondering is there any general cut-off for the peak_region_fragments.

The specific cutoff values to use will depend on your dataset, and should be determined through a supervised analysis. In general, you should aim to remove barcodes with a low number of total counts, which could suggest that they originate from empty droplets or that they have too few counts to provide useful information. You may also remove barcodes that have many more counts than most other barcodes, which could suggest they originate from some artefact (multiplets, dead cells).