CellBender genes removed?

[giorgiatosoni]

Hi,

Thank you very much for the additional features added to the new release, I find the CellBender vignette very useful!

I was wondering what would be a way to see which genes are removed (nFeature_Diff) in the dual assay. The dual assay plotting is already very useful to understand if CellBender is doing a great job, however, it would be useful to have a quick way to understand what Cellbender did and what was considered as ambient RNA.

Thanks!

Giorgia

[samuel-marsh]

Hi Giorgia,

Yes totally makes sense. Working on function to do just that. Hope to push it to dev branch before end of the week.

Best,
Sam

[samuel-marsh]

Hi Giorgia,

Ok the functions are now present in develop branch (v1.0.0.0007).
Here's quick overview.

# Already present function (but now renamed).
# This will add two columns to meta.data slot (nFeature_Diff and nCount_Diff)
# These will be the numerical differences in features per cell and counts per cell between the two assays
dual_seurat <- Add_CellBender_Diff(seurat_object = dual_seurat, raw_assay_name = "RAW", cell_bender_assay_name = "RNA")

head(dual_seurat@meta.data)
                       orig.ident nCount_RNA nFeature_RNA nCount_RAW nFeature_RAW nFeature_Diff nCount_Diff
WT1_AAACCCAAGCGACCCT-1        WT1       9844         3423      10032         3481            58         188
WT1_AAACCCAAGGAGGCAG-1        WT1       9248         3477       9431         3531            54         183
WT1_AAACCCACACACCAGC-1        WT1       3207         1699       3472         1852           153         265
WT1_AAACCCACATCTGTTT-1        WT1      14328         3930      14572         3993            63         244
WT1_AAACCCAGTATTCTCT-1        WT1      12034         3911      12223         3958            47         189
WT1_AAACCCAGTCTTGGTA-1        WT1      11722         3933      11967         4024            91         245



# New Function #1
# Return data.frame with rowSums for each gene in dataset (features will be rownames)
# 4 columns Raw Counts, Cell Bender Counts, Count Difference, % Change
# by default data.frame returns sorted by % change.

diff_df <- CellBender_Feature_Diff(seurat_object = dual_seurat, raw_assay = "RAW", cell_bender_assay = "RNA")

head(diff_df)
        Raw_Counts CellBender_Counts Feature_Diff Pct_Diff
Gm42418     256834              3198       253636 98.75484
Uqcr11        1315                20         1295 98.47909
Gm48099        277                 5          272 98.19495
Tmsb4x        6010               116         5894 98.06988
Rps21         2597                61         2536 97.65114
Ndufa4        3960                94         3866 97.62626

tail(diff_df)
           Raw_Counts CellBender_Counts Feature_Diff Pct_Diff
Zfy2                6                 6            0        0
Gm21865            11                11            0        0
mt-Atp8            74                74            0        0
AC168977.1          8                 8            0        0
AC132444.5          9                 9            0        0
CT868723.1          9                 9            0        0



# New Function #2
# Visualize the results of `CellBender_Feature_Diff`
# Can plot either genes changed by XX% or a specific number of genes.
# optional parameters change change feature labeling and other plot parameters
# more optional parameters than those shown below.  Use `?CellBender_Diff_Plot` to see manual entry.

# plot default (features > 25% change)
p1 <- CellBender_Diff_Plot(feature_diff_df = diff_df)
# Only plot features changed more than 50%
p2 <- CellBender_Diff_Plot(feature_diff_df = diff_df, pct_diff_threshold = 50)
# Plot the top 500 changed genes regardless of the % change
p3 <- CellBender_Diff_Plot(feature_diff_df = diff_df, num_features = 500, pct_diff_threshold = NULL)
# Only plot text labels for the top 10 changed features
p4 <- CellBender_Diff_Plot(feature_diff_df = diff_df, num_labels = 10)
# Don't add text feature labels
p5 <- CellBender_Diff_Plot(feature_diff_df = test, labels = FALSE)
# Label a specific set of genes (Note these genes must be presented in the filtered (pct_diff_threshold or num_features data).
p6 <- CellBender_Diff_Plot(feature_diff_df = test, custom_labels = "Malat1")

library(patchwork)
(p1 + p2 + p3) / (p4 + p5 + p6)

Let me know if you have any issues with new functions and I can reopen issue.
Best,
Sa,

[giorgiatosoni]

That's so useful! Thank you very much! :)

[samuel-marsh]

No problem!! Definitely useful functions to have!

CellBender v3 (currently in alpha) includes some of this in output html summary but nice to have in R when creating objects etc.