How to use SingleCellSignalR on 2 different Seurat objects.

[choinc615]

Hi,

Thank you for the very interesting tool.

In the demo files, you have provided ways to load up one Seurat object and process it further for this package.

I want to determine whether there is interaction between the immune cells and the skin cells. To that end, I have performed some clustering using Seurat. To that end, I have generated two different Seurat objects and want to know study the interaction between the clusters found within the different Seurat objects.

Is there a way to integrate the two objects after the clustering analysis and use it for your SCSR?

Many thanks!

Inchul

[SCA-IRCM]

Hi, Yes I think you can do it, you must merge the data from the two Seurat objects:

cluster.1 = as.numeric(Idents(seurat.1))
data.1 = data.frame(seurat.1[["RNA"]]@data)

cluster.2 = as.numeric(Idents(seurat.2)) + max(cluster.1)
data.2 = data.frame(seurat.2[["RNA"]]@data)

cluster = c(cluster.1, cluster.2)
data = cbind(data.1, data.2)

Then you can use dataand clusterin the cell_signalingfunction of SCSR. I advise you to create a c.names vector (see documentation) to keep a good track of your different clusters.

Thank you for using SingleCellSignalR.

SCA

[choinc615]

Thanks for the fast response. Will give this a try and let you know if it would work :)

[choinc615]

Hi,

Thank you for your response.

I have attempted to follow your advice and updated my script as below:

Retreiving the results of the preprocessing from the Seurat object

cluster.treg = as.numeric(Idents(kal9Subset)) cnames.treg <- paste0("treg", cluster.treg)

data.treg = data.frame(kal9Subset[["RNA"]]@data) all.genes.treg <- rownames(kal9Subset) nrow(data.treg) # 14461

From IC_B5:

cluster.skin = as.numeric(Idents(kasp9)) cnames.skin <- paste0("skin", cluster.skin) data.skin = data.frame(kasp9[["RNA"]]@data) all.genes.skin <- rownames(kasp9) nrow(data.skin) #13834

The number of different transcripts/genes in the seurat object is different I need to make the number of rows the same.

data.treg = subset(data.treg, rownames(data.treg)%in%rownames(data.skin)) # After this, the number in data.treg is reduced to 12686, which is less than that of skin. data.skin = subset(data.skin, rownames(data.skin)%in%rownames(data.treg)) # I will therefore also shrink the skin data. After this, I am left with also 12686 genes. I can now combine the two tables.

data.combined = cbind(data.treg,data.skin) all.genes.combined = rownames(data.combined) cluster.combined = c(cluster.treg,cluster.skin) cnames.combined = c(cnames.treg,cnames.skin)

length(cluster.combined) # 8961 length(cnames.combined) #8961

Ligand/Receptor analysis using SingleCellSignalR

signal = cell_signaling(data=data.combined,genes=all.genes.combined,cluster=cluster.combined, species = "mus musculus") signal

After running all the analysis, the signal gives me the interactions between clusters 1 to 13. My understanding is that the function is unable to discriminate between cluster 1 from the Treg data and cluster 1 from Skin data.

Trying to get around the problem

I have tried to get around the problem by adding the c.names vector as you suggested. How I compiled the c.vector is written above, where I use the paste function.

Upon running the analysis with the c.names argument as below: signal = cell_signaling(data=data.combined,genes=all.genes.combined,cluster=cluster.combined, c.names=cnames, species = "mus musculus")

I get the following error message: The length of c.names must be equal to the number of clusters and must contain no duplicates. The cluster names must not include special characters

I have attempted several different ways to address this error: 1) made the length of c.names and the number of clusters the same 2) Manually defined a list of cluster names to c.names 3) removed all the special characters and spaces. But none of them worked.

I am assuming at this stage that the problem is maybe because of the cluster variable. It seems that the function recognises cluster 1 from Treg and cluster 2 from Tregs in the same way. In other words, it recognises the two different clusters as 1 cluster. Therefore, the function sees that the number of clusters are actually less than what they really are.

Is there a way to address this? It seems to me that the clusters have to be in numerics.

Thanks,

Inchul

[SCA-IRCM]

Hi, As I told you, you need to add max(cluster.1) to cluster 2: cluster.skin = as.numeric(Idents(kasp9)) + max(cluster.treg)

For the c.names vectors: cnames.treg <- paste0("treg", unique(cluster.treg)) cnames.skin <- paste0("skin", unique(cluster.skin)) should work.

SCA

[choinc615]

Thanks. This has worked :). For those interested, to make the distinction between Tregs and skin clusters a little easier, I have used: cluster.skin = as.numeric(Idents(kasp9)) + 100

instead of the suggested: cluster.skin = as.numeric(Idents(kasp9)) + max(cluster.treg)

Skin clusters then have 3 digits, which makes them easier to distinguish. But this is personal preference.