Skip to content

Commit

Permalink
update documentations
Browse files Browse the repository at this point in the history
  • Loading branch information
@zktuong
zktuong committed Jun 24, 2022
1 parent 780100c commit 7240a53
Show file tree
Hide file tree
Showing 3 changed files with 20 additions and 133 deletions.
2 changes: 1 addition & 1 deletion R/plot_cpdb2.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ plot_cpdb2 <- function(cell_type1, cell_type2, scdata, idents, means, pvals, dec
desiredInteractions = NULL, interaction_grouping = NULL, edge_group_colors = NULL,
node_group_colors = NULL, version3 = FALSE, return_df = FALSE, ...) {
if (class(scdata) == "Seurat") {
stop("Sorry not supported yet. Please use a SingleCellExperiment object.")
stop("Sorry not supported. Please use a SingleCellExperiment object.")
}
if (length(separator) > 0) {
sep = separator
Expand Down
2 changes: 1 addition & 1 deletion R/plot_cpdb3.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ plot_cpdb3 <- function(cell_type1, cell_type2, scdata, idents, means, pvals, dec
alpha = 0.5, edge_colors = NULL, grid_colors = NULL, show_legend = TRUE, legend.pos.x = 20,
legend.pos.y = 20, ...) {
if (class(scdata) == "Seurat") {
stop("Sorry not supported yet. Please use a SingleCellExperiment object.")
stop("Sorry not supported. Please use a SingleCellExperiment object.")
}
if (length(separator) > 0) {
sep = separator
Expand Down
149 changes: 18 additions & 131 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -102,29 +102,6 @@ if ```genes``` and ```gene.family``` are both not specified, the function will t
Specifying ```keep_significant_only``` will only keep those that are p<0.05 (which you can try to adjust with ```p.adjust.method```).


### plot_cpdb3
Generates a chord diagram inspired from CellChat's way of showing the data!

```R
library(ktplots)
data(kidneyimmune)
data(cpdb_output2)

p <- plot_cpdb3(cell_type1 = 'B cell', cell_type2 = 'CD4T cell|MNPd',
scdata = kidneyimmune,
idents = 'celltype', # column name where the cell ids are located in the metadata
means = means2,
pvals = pvals2,
deconvoluted = decon2, # new options from here on specific to plot_cpdb3
keep_significant_only = TRUE,
standard_scale = TRUE,
remove_self = TRUE
)
p
```

![plot_cpdb3](exampleImages/plot_cpdb3.png)

### plot_cpdb2
Generates a circos-style wire/arc/chord plot for cellphonedb results.

Expand Down Expand Up @@ -215,119 +192,29 @@ test <- plot_cpdb2(cell_type1 = "CD4_Tem|CD4_Tcm|CD4_Treg", # same usage style a
```
![plot_cpd2](exampleImages/plot_cpdb2.png)

## New feature

### compare_cpdb/plot_compare_cpdb
These functions piggy-back on the original `plot_cpdb` function and generates the results from cellphonedb that are run separately on replicates. The goal of this function is to able to compare interactions between groups.

I've provided example datasets from randomnly selected `Healthy` and `Severe` samples from [Stephenson et al. COVID-19 single cell data set published in Nature Medicine 2021](https://www.nature.com/articles/s41591-021-01329-2). This should get it running and for you to see how the input data folder should be organised:
```R
data(covid_sample_metadata)
data(covid_cpdb_meta)
file <- system.file("extdata", "covid_cpdb.tar.gz", package = "ktplots")
# copy and unpack wherever you want this to end up
file.copy(file, ".")
system("tar -xzf covid_cpdb.tar.gz")
```

It requires 1) an input table like so:
```R
> covid_cpdb_meta
sample_id cellphonedb_folder sce_file
1 MH9143325 covid_cpdb/MH9143325/out covid_cpdb/MH9143325/sce.rds
2 MH9143320 covid_cpdb/MH9143320/out covid_cpdb/MH9143320/sce.rds
3 MH9143274 covid_cpdb/MH9143274/out covid_cpdb/MH9143274/sce.rds
4 MH8919226 covid_cpdb/MH8919226/out covid_cpdb/MH8919226/sce.rds
5 MH8919227 covid_cpdb/MH8919227/out covid_cpdb/MH8919227/sce.rds
6 newcastle49 covid_cpdb/newcastle49/out covid_cpdb/newcastle49/sce.rds
7 MH9179822 covid_cpdb/MH9179822/out covid_cpdb/MH9179822/sce.rds
8 MH8919178 covid_cpdb/MH8919178/out covid_cpdb/MH8919178/sce.rds
9 MH8919177 covid_cpdb/MH8919177/out covid_cpdb/MH8919177/sce.rds
10 MH8919176 covid_cpdb/MH8919176/out covid_cpdb/MH8919176/sce.rds
11 MH8919179 covid_cpdb/MH8919179/out covid_cpdb/MH8919179/sce.rds
12 MH9179826 covid_cpdb/MH9179826/out covid_cpdb/MH9179826/sce.rds
```
where each row is a sample, the location of the `out` folder generated by cellphonedb and a single-cell object used to generate the cellphonedb result. If cellphonedb was ran with a `.h5ad` the sce_file would be the path to the `.h5ad` file.

and 2) a sample metadata data frame for the tests to run:
```R
> covid_sample_metadata
sample_id Status_on_day_collection_summary
MH9143325 MH9143325 Severe
MH9143320 MH9143320 Severe
MH9143274 MH9143274 Severe
MH8919226 MH8919226 Healthy
MH8919227 MH8919227 Healthy
newcastle49 newcastle49 Severe
MH9179822 MH9179822 Severe
MH8919178 MH8919178 Healthy
MH8919177 MH8919177 Healthy
MH8919176 MH8919176 Healthy
MH8919179 MH8919179 Healthy
MH9179826 MH9179826 Severe
```

So if I want to compare between Severe vs Healthy, I would specify the function as follows:
```R
## set up the levels
covid_sample_metadata$Status_on_day_collection_summary <- factor(covid_sample_metadata$Status_on_day_collection_summary, levels = c('Healthy', 'Severe'))

out <- compare_cpdb(cpdb_meta = covid_cpdb_meta,
sample_metadata = covid_sample_metadata,
celltypes = c("B_cell", "CD14", "CD16", "CD4", "CD8", "DCs", "MAIT", "NK_16hi", "NK_56hi", "Plasmablast", "Platelets", "Treg", "gdT", "pDC"), # the actual celltypes you want to test
celltype_col = "initial_clustering", # the column that holds the cell type annotation in the sce object
groupby = "Status_on_day_collection_summary") # the column in the sample_metadata that holds the column where you want to do the comparison. In this example, it's Severe vs Healthy
```
This returns a list of dataframes (for each contrast found) with which you can use to plot the results.

`plot_compare_cpdb` is a simple function to achieve that but you can always just make a custom plotting function based on what you want.

```R
plot_compare_cpdb(out) # red is significantly increased in Severe compared to Healthy.
```
![plot_compare_cpdb](exampleImages/plot_compare_cpdb_example.png)

If there are multiple contrasts and groups, you can facet the plot by specifying `groups = c('group1', 'group2')`.
### plot_cpdb3
Generates a chord diagram inspired from [CellChat](https://github.com/sqjin/CellChat)'s way of showing the data!

Usage is similar to `plot_cpdb2` but with reduced options. Additional kwargs are passed to `plot_cpdb`.
```R
# let's mock up a new contrast like this
covid_sample_metadata$Status_on_day_collection_summary <- c(rep('Severe', 3), rep('Healthy', 2), rep('notSevere', 2), rep('Healthy', 4), 'notSevere')
out <- compare_cpdb(cpdb_meta = covid_cpdb_meta,
sample_metadata = covid_sample_metadata,
celltypes = c("B_cell", "CD14", "CD16", "CD4", "CD8", "DCs", "MAIT", "NK_16hi", "NK_56hi", "Plasmablast", "Platelets", "Treg", "gdT", "pDC"), # the actual celltypes you want to test
celltype_col = "initial_clustering", # the column that holds the cell type annotation in the sce object
groupby = "Status_on_day_collection_summary")
plot_compare_cpdb(out, alpha = .1, groups = names(out)) # there's no significant hit at 0.05 in this dummy example
```
![plot_compare_cpdb4](exampleImages/plot_compare_cpdb_example4.png)

The default method uses a pairwise `wilcox.test`. Alternatives are pairwise Welch's `t.test` or a linear mixed model with `lmer`.

To run the linear mixed effect model, it expects that the input data is paired (i.e an individual with multiple samples corresponding to multiple timepoints):
library(ktplots)
data(kidneyimmune)
data(cpdb_output2)

```R
# just as a dummy example, lets say the samples are matched where there are two samples per individual
covid_sample_metadata$individual <- rep(c("A", "B", "C", "D", "E", "F"), 2)

# actually run it
out <- compare_cpdb(cpdb_meta = covid_cpdb_meta,
sample_metadata = covid_sample_metadata,
celltypes = c("B_cell", "CD14", "CD16", "CD4", "CD8", "DCs", "MAIT", "NK_16hi",
"NK_56hi", "Plasmablast", "Platelets", "Treg", "gdT", "pDC"),
celltype_col = "initial_clustering",
groupby = "Status_on_day_collection_summary",
formula = "~ Status_on_day_collection_summary + (1|individual)", # formula passed to lmer
method = "lmer")

plot_compare_cpdb(out, contrast = 'Status_on_day_collection_summarySevere') # use the colnames(out) to pick the right column.
p <- plot_cpdb3(cell_type1 = 'B cell', cell_type2 = 'CD4T cell|MNPd',
scdata = kidneyimmune,
idents = 'celltype', # column name where the cell ids are located in the metadata
means = means2,
pvals = pvals2,
deconvoluted = decon2, # new options from here on specific to plot_cpdb3
keep_significant_only = TRUE,
standard_scale = TRUE,
remove_self = TRUE
)
p
```
![plot_compare_cpdb2](exampleImages/plot_compare_cpdb_example2.png)

Specifying `cluster = TRUE` will move the rows and columns to make it look a bit more ordered.
```R
plot_compare_cpdb(out, contrast = 'Status_on_day_collection_summarySevere', cluster = TRUE)
```
![plot_compare_cpdb3](exampleImages/plot_compare_cpdb_example3.png)
![plot_cpdb3](exampleImages/plot_cpdb3.png)


## Other useful functions
Expand Down

0 comments on commit 7240a53

Please sign in to comment.