UCell support instead of AddModuleScore? about scpubr HOT 12 CLOSED

Hi @cnk113,

Thanks for the interest and the feedback! Happy to hear that you like the package.

This is indeed a very interesting addition. I guess you mean using it as scoring method in SCpubr::do_EnrichmentHeatmap() no?

I will have a look at it and put in the To-Do list!

Best,
Enrique

from scpubr.

Hey Enrique,

Yes that's the function I was thinking of.

Best,
Chang

from scpubr.

Hi @cnk113,

I have implemented the method in the last commit. Could you please check it out and tell me if it works for you as intended?

devtools::install_github("enblacar/SCpubr", ref = "d6e7617a9c4e7817da8bf9bfde73eb7634991625")

For this, use the following code:

SCpubr::do_EnrichmentHeatmap(sample = sample, 
                             input_gene_list = genes, 
                             ncores = 1, 
                             flavor = "UCell", 
                             storeRanks = TRUE)

With it, I got the following plot:

While with Seurat's AddModuleScore:

If it also works as intended for you, then I will do some more aesthetic tweaks and I will add it to the next CRAN update!

A couple of notes though:

• You need to have R version higher or equal to 4.2.0.
• You need UCell installed.

Looking forward to your feedback!

Enrique

from scpubr.

Hey Enrique,

I'm getting this error when I run it without the UCell parameter:
SCpubr::do_EnrichmentHeatmap(sample = first, group.by = 'clusters', input_gene_list = sigs)

Error in `[.data.frame`([email protected], , col_name) : 
  undefined columns selected

Do you know why this occurs?

Best,
Chang

from scpubr.

Hi @cnk113,

Hmm, interesting. Does this happen, under the same parameters, when flavor = "UCell"? Can you maybe double-check if group.by is a metadata variable without NAs and sigs is a named vector?

I will have a look also to the code, but so far it works for me in any combination I provide.

Best,
Enrique

from scpubr.

Ah my named lists had characters that would change in the UCell names due to naming issues. It works!

from scpubr.

So something to note, I was wondering if we could get max enrichment percentile cutoff over the group.by metadata? This way we can see the most enriched sig per cluster?
As you can see here the other clusters don't look as enriched due to an outlier and I think the module scores are relative values so I don't see an issue with doing the cutoffs per cluster?

from scpubr.

Hi @cnk113,

Glad to hear it worked for you!

I do not fully understand what you are trying to suggest here, but it gave me another idea that might work as well as what you say. In the last commit, I implemented a min.cutoff and max.cutoff parameter, which will let you subset the color scale to set the min or max values to a given one. The values have to be comprised in the absolute max or min values in the enrichment matrix.

This is a change affecting all the clusters at the same time, not a single one. But this is basically the aim of the visualization, to put the enrichment in a heatmap and compare between clusters.

Here you have some examples:

# Original heatmap.
SCpubr::do_EnrichmentHeatmap(sample = sample, 
                             input_gene_list = genes, 
                             ncores = 1, 
                             flavor = "UCell", 
                             storeRanks = TRUE)

# Set a max value for the scale.
SCpubr::do_EnrichmentHeatmap(sample = sample, 
                             input_gene_list = genes, 
                             ncores = 1, 
                             flavor = "UCell", 
                             storeRanks = TRUE,
                             max.cutoff = 0.5)

# Set a min value for the scale.
SCpubr::do_EnrichmentHeatmap(sample = sample, 
                             input_gene_list = genes, 
                             ncores = 1, 
                             flavor = "UCell", 
                             storeRanks = TRUE,
                             min.cutoff = 0.5)

# Set a min and a max value for the scale
SCpubr::do_EnrichmentHeatmap(sample = sample, 
                             input_gene_list = genes, 
                             ncores = 1, 
                             flavor = "UCell", 
                             storeRanks = TRUE,
                             min.cutoff = 0.2,
                             max.cutoff = 0.5)

Please install the commit using:

devtools::install_github("enblacar/SCpubr", ref = "52429ab0e0dc1d99450f6807bb840f392a577e08")

And be aware this is a dev version so the code might break. Let me know if this works for you! The changes will appear in the next CRAN update.

Best,
Enrique

from scpubr.

Ok this works for me! Thanks for the quick turnaround.

Best,
Chang

from scpubr.

Ah one more thing, could you perhaps have cutoffs also applicable to your other heatmaps like pathways/TF?

from scpubr.

Hi @cnk113,

Happy to hear it worked for you! I have added min.cutoff and max.cutoff behaviour to all the rest of heatmap functions.

You can access this by:

devtools::install_github("enblacar/SCpubr", ref = "c6e4f7bbc635d49ec349c7b2c4832a5e144745e1")

All the changes will become part of the next CRAN update.

Closing this Issue for now. Thanks for your feedback!
Enrique

from scpubr.

All the changes have been implemented in SCpubr v1.0.3, currently available on CRAN.

Best,
Enrique

from scpubr.