How to identify cell types using addModuleScore function?
1
1
Entering edit mode
20 months ago
a511512345 ▴ 160

Hello, there,

I am learning single-cell RNA-seq analysis using Seurat package. I have clustered cells into 12 clusters. Next, I want to identify the cell types of these cell clusters.

I read some paper using addModuleScore function based on the known cell markers. I did not find any example.

however, I do not know how to do identify cell types based on the scores as one cell cluster has several cell type scores. For example, cell cluster one got an average of 0.5 in stem cell score and 0.6 in bronchial epithelial cell score，should the cell cluster one be considered as bronchial epithelial cell？

Thank you very much

single cell RNA-seq • 11k views
0
Entering edit mode

I have a recent answer on another question that may be helpful for you.

0
Entering edit mode

Thank you for this answer. However， I feel very hard to follow the answer as it was not an example.

1
Entering edit mode

There is not a magical answer with your current method - there is no way for us to tell you the appropriate score thresholds to use for your annotations. If you want an automated approach, you can try SingleR, garnet, or any of the various other automated methods.

1
Entering edit mode
13 months ago
BenBoErBa ▴ 10

If a simple example is what your need, then this may be helpful.

data("pbmc_small")
cd_features <- list(c(
'CD79B',
'CD79A',
'CD19',
'CD180',
'CD200',
'CD3D',
'CD2',
'CD3E',
'CD7',
'CD8A',
'CD14',
'CD1C',
'CD68',
'CD9',
'CD247'
))
object = pbmc_small,
features = cd_features,
ctrl = 5,
name = 'CD_Features'
)

0
Entering edit mode

Could you please comment, how changing the "ctrl = 5" flag in the function can affect the module score?

1
Entering edit mode

I imagine using only 5 control genes from each bin would add additional variability. I'd leave it at the default of 100. The differences may be minor in the resulting scores, but using only 5 would make them more susceptible to extremes within each bin.

The example here uses 5 because the pbmc_small dataset has been stripped down to only a handful of genes so that is could be shipped with the package.

0
Entering edit mode

Thank you!