Closed DelongZHOU closed 2 months ago
Hi,
This is a good question, I have come across this problem in my work as well. To overcome this issue, I recommend changing your group.by
parameters in MetacellsByGroups
. For example, in the tutorial we run this code:
# construct metacells in each group
seurat_obj <- MetacellsByGroups(
seurat_obj = seurat_obj,
group.by = c("cell_type", "Sample"), # specify the columns in seurat_obj@meta.data to group by
reduction = 'harmony', # select the dimensionality reduction to perform KNN on
k = 25, # nearest-neighbors parameter
max_shared = 10, # maximum number of shared cells between two metacells
ident.group = 'cell_type' # set the Idents of the metacell seurat object
)
You can allow hdWGCNA to group together cells from different samples if you remove `"Sample":
seurat_obj <- MetacellsByGroups(
seurat_obj = seurat_obj,
group.by = c("cell_type"), # only using cell_type, not Sample
reduction = 'harmony',
k = 25,
max_shared = 10,
ident.group = 'cell_type'
)
It sounds like you want to make sure that you don't merge cells from control + treatment into one meta-cell. For example, let's say your dataset has a meta-data column called "Disease_Group" which tells us which cells are from control or treatment samples. Since you probably don't want those to mix, you can provide this grouping variable to MetacellsByGroups
:
# construct metacells in each group
seurat_obj <- MetacellsByGroups(
seurat_obj = seurat_obj,
group.by = c("cell_type", "Disease_Group"), # this will form separate metacells for each cell type and each treatment group!
reduction = 'harmony',
k = 25,
max_shared = 10,
ident.group = 'cell_type'
)
While this should work in general, you also mentioned that this group has a total of 400 cells across all samples. I worry that this may be too few cells for obtaining meaningful results with hdWGCNA, so proceed with caution.
Hi Sam,
Thanks for your response.
My reasoning was based on this paper on DEG https://www.nature.com/articles/s41467-021-25960-2 which states that pseudobulk methods have fewer false positive by considering sample / replicate variations. So I want to keep the replicate information as much as possible, which is why I used the group.by with sample. Given the differential ME value is calculated using single cell method (and I don't see performing pseudobulk WGCNA with 3 samples / condition feasible), I might try to group within conditions.
Do you have suggestions for quality control for hdWGCNA with low cell number? For example is there some metrics that I can use to compare them to modules identified in the same dataset but higher cell number? Thank you!
Do you have suggestions for quality control for hdWGCNA with low cell number?
Unfortunately I do not recommend running hdWGCNA with extremely low cell numbers.
Do you have suggestions for quality control for hdWGCNA with low cell number?
Unfortunately I do not recommend running hdWGCNA with extremely low cell numbers.
That's fair. Thank you!
Hi Sam,
What's the best approach for cell groups with few cell counts?
In my case I have two groups that I'm interested in with 10~100 cells per sample totalling to ~400 cells. Would it be alright to group some samples into a "meta sample"? What's the minimum threshold for min_cell? For example, the cell count for one group is: Control: 16,31,43,76 Treatment: 16,54,60,103 Would it be acceptable to group the samples in bold, and set the min_cell at 40?
Thank you!