'pct_reads_in_peaks' and 'a fraction of reads in the peak' by FRiP

[akhst7]

Hi,

FRiP generates a 'fraction of reads in the peak', times 100 of which should be equivalent to 'pct_reads_in_peaks' described under Computing QC Metrics in Analyzing PBMC scATAC-seq or am I wrong on this ?

I used pbmc_granulocyte_sorted_3k_atac_fragments.tsv.gz from 10x and generated a total.fragment counts by following;

total.fragments<-CountFragments("~/Downloads/fragments/pbmc_granulocyte_sorted_3k_atac_fragments.tsv.gz")

From there, I used FRiP to get the 'fraction of reads in the peak' as follows,

pbmc <- FRiP(
object = pbmc,
assay = 'ATAC',
total.fragments = 'total.fragments'
)

A summary of the resulting 'fraction of reads in the peak' is following;

summary(pbmc$FRiP)
    Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
  0.0025   0.8414   1.2725   3.8284   1.8448 378.7273 

As you can see, a distribution looks ZINB, and if I use the QC criterium regarding 'pct_reads_in_peaks' (Cells with low values (i.e. <15-20%)) in 'Analyzing PBMC scATAC-seq' vignette, there are very few cells left to play with.

Looking at the FRiP function,

frip <- peak.counts/total_fragments_cell

FRiP generate a fraction of peaks (where read pileups are) in fragments per a single cell rather than a fraction of reads in fragments per cell.

Am I missing anything here ? Any pointer will be appreciated.

[akhst7]

Ok I figures this out.

[Susan920715]

Hi, Can you tell me whether FRiP generates a 'fraction of reads in the peak', times 100 of which should be equivalent to 'pct_reads_in_peaks' .I have the same confusions, thankyou verymuch

[akhst7]

@Susan920715 , FRiP will give you the 'fraction of reads in the peak' if you feed it the right peak.count and total_fragments_cell. The problem, not explained well in the vignette is that very definitions of total peak, read, and fragment counts differ if you are dealing with 10X data analyzed by CellRanger ARC or ATAC. In my case, I thought that the read and fragment counts (vs cut-site counts) generated by CR-ARC/ATAC were the same but in fact they were different. For CR-ARC/ATAC, you can take a look at definitions of these in the Summary Metrics page, as 10x changed summary metadata (again!!). If you can, you should calculate this manually with using a right set ofpeak and total frag counts I think using cut-site counts rather than fragment counts in the peaks may be more suitable describing the number of Tn5 accessible open chromatin.

[Susan920715]

I see ~ Thank you very much!!