Tweaks (single cell phyloset and more)

[nifets]

still WIP

left to do:

• add phylostrata labels to x axis of strata distribution plot
• cache pseudobulking in sc obj
• cache TAI computation in sc obj
• parallelise computation of pseudobulking and TAI

[LotharukpongJS]

Hi @nifets ,
While doing an unrelated analysis, I thought it could be good to multiply the expression values by 2 (or something between 2 and 10) prior to rounding for the rlog or vst transformation. This would preserve more information for the lower expression values, especially TPM.

myTAI/R/plot_signature_transformed.R

Lines 23 to 24 in 5826fa6

	ct[["vst"]] <- \(x) DESeq2::vst(round(x, digits=0))
	ct[["rlog"]] <- \(x) DESeq2::rlogTransformation(round(x, digits=0))

DESeq2::vst(round(x*1, digits=0))

DESeq2::vst(round(x*2, digits=0))

DESeq2::vst(round(x*10, digits=0))

where x is expression data of class "matrix" "array". For the histogram, I am just showing the fourth column.

Let me know what you think :)

Best,
Sodai

[nifets]

Hi @nifets , While doing an unrelated analysis, I thought it could be good to multiply the expression values by 2 (or something between 2 and 10) prior to rounding for the rlog or vst transformation. This would preserve more information for the lower expression values, especially TPM.

myTAI/R/plot_signature_transformed.R

Lines 23 to 24 in 5826fa6

	ct[["vst"]] <- \(x) DESeq2::vst(round(x, digits=0))
	ct[["rlog"]] <- \(x) DESeq2::rlogTransformation(round(x, digits=0))

DESeq2::vst(round(x*1, digits=0))

DESeq2::vst(round(x*2, digits=0))

DESeq2::vst(round(x*10, digits=0))

where x is expression data of class "matrix" "array". For the histogram, I am just showing the fourth column.

Let me know what you think :)

Best, Sodai

Hi Sodai, yes we can definitely do that I think!

[nifets]

Okay, I think I am ready to merge this to master. I have made quite a few changes to the single cell object. The main one is that it no longer wraps around Seurat, but uses the count matrix directly. So we are decoupled from Seurat now (but the user can still build a sc phyloset from seurat if they choose to). The TAI computation for single cell is now cached. I also tried to cache the pseudobulking but I couldn't find an elegant way to do it with S7. I might come back to this in the future. If the user wants to, they can precache the pseudobulk by setting p@.pseudobulk_cache[[identity]] for the grouping they want to cache. I think it would be nice to expand the vignette for single cell or maybe make a new single cell vignette that goes over every feature of the object. I think it's in a stable place now.

And generally I think we can probably push this to CRAN relatively soon. Maybe there are some things which could benefit from more documentation and/or useful messages. I have added back the messages in plot signature but maybe there are other places that would benefit. And I'm still not 100% happy with the single cell plot signature plots, I think we could make them more readable still.