Handling computed copy numbers with spike-ins

[LukaP-BB]

Hi, I'm considering using COMPASS with my data, but I'm a bit reluctant to use it with the raw read counts as suggested in the readme. My data has spike-ins that I manually identify in my workflow, allowing me to compute and segment CNVs in my data. I end up with calls that are way less noisy than the raw data, that I would rather use.

Would using ploidy values work well with the algorithm ?
And is there a way to label the cells beforehand, at least the spike-ins, to help with the tree inference ?

Many thanks !

[e-sollier]

Hi,

If you already have copy numbers for each region and each cell, I would use in place of the raw read counts something like int(50*copy_number). Using the copy numbers directly won't work well because COMPASS assumes that the read counts come from a negative binomial distribution, so if you use copy numbers as read counts you would end up with very low read counts, which would be inferred to be rather unreliable.

Cells cannot be labelled. I don't know which protocol you are using, what do you mean by spike-in cells ?

[LukaP-BB]

I see, thanks for your reply !
In our protocol, we add "normal" cells to the mix, that serve as reference for the copy numbers. The first step of the bioinformatics workflow is to identify these cells to compute the ploidy using the dedicated method in the mosaic library.
These cells could be considered as the "root" of the phylogenic tree, since they have no CNAs or SNVs.

I'll give the 50*CN method a try and come back to you if I encounter any issues.