human autosomes list of Amulet vs fragment file

[Erfan1369]

Dear Amulet Developers,

First of all, thank you for developing such a great tool!

I have a simple question about the human_autosomes.txt file, which includes a list of chromosomes 1 to 22. When I look into the fragment file from the Cell Ranger output, I see additional contigs listed. Should I include those in the human_autosomes.txt file as well, or exclude them?

primary_contig=1

primary_contig=10

primary_contig=11

primary_contig=12

primary_contig=13

primary_contig=14

primary_contig=15

primary_contig=16

primary_contig=17

primary_contig=18

primary_contig=19

primary_contig=2

primary_contig=20

primary_contig=21

primary_contig=22

primary_contig=3

primary_contig=4

primary_contig=5

primary_contig=6

primary_contig=7

primary_contig=8

primary_contig=9

primary_contig=MT

primary_contig=X

primary_contig=Y

primary_contig=KI270728.1

primary_contig=KI270727.1

primary_contig=GL000009.2

primary_contig=GL000194.1

primary_contig=GL000195.1

primary_contig=GL000219.1

primary_contig=KI270734.1

primary_contig=GL000213.1

primary_contig=GL000218.1

primary_contig=KI270731.1

primary_contig=KI270721.1

primary_contig=KI270726.1

primary_contig=KI270711.1

primary_contig=KI270713.1

Thank you for your assistance!

[ajt986]

I would continue with just chromosomes 1-22 so that the read distributions reflect chromosomes where there is an expected 2 copies of each for standard usage of AMULET.

[Erfan1369]

I will follow your suggestion!
Thanks for your super quick response!

[Erfan1369]

Hi again @ajt986

I have the output from Amulet for my samples, and I noticed something unusual in the data. In the image below, a fixed p-value or q-value appears repeatedly across different cells (barcodes). Is it normal to observe this, or could it be due to technical issues?

Additionally, how do you determine the appropriate cutoff (p-value or q-value) to filter out cells identified as doublets?

[ajt986]

Essentially these cells have the same number of overlaps > 2 and will have the same p-value as a result. The p-values are calculated based on cumulative Poisson distribution probabilities. The q-value is the FDR adjustment of the p-values for multiple comparisons, so if the p-values are the same, then the q-values will be too.

q-values less than 0.05 or 0.01 (FDR < 5% or 1% respectively) are the high confident doublets/multiplets in the data and these thresholds are oftern used for other tools/applications. I personally use a q-value threshold less than 0.05, but a threshold of 0.01 is reasonable too.