logCPM and logTPM are not coherent

[violafanfani]

https://github.com/QuackenbushLab/NetSciDataCompanion/blob/ac0f21bf7639642a5578ddea25c0e3e465f62bd2/R/NetSciDataCompanion.R#L75C12-L75C31

For logCPMNormalization we are now checking that the passed class is a count matrix, but for logTPM we check that the object is an RSE. Can we revert/update the functions to be working in the same way?

[katehoffshutta]

Good catch @violafanfani. Which do you suggest that we support, or do you think it is important to be able to pass either matrix or RSE?

[violafanfani]

I think there might be a reason for this (maybe edgeR uses a matrix), I would try to use both if it's easy enough. But no strong preference on my side

[katehoffshutta]

Okay, I think we can at least accept both types as function arguments and throw an informative error asking the person to use the other type if needed. I will look into this.

[katehoffshutta]

Adding to this issue: currently, the return value logCPM$counts gives the raw counts from the input matrix, while logTPM$counts gives transformed counts based on recount3::transform_counts. I have described this in the quickstart guide as "expected behavior", but when this issue is addressed, we should consider adding a raw counts return to logTPM and/or a transformed counts return to logCPM.