only some clusters are annotated

[feanaros]

Hello, I have a singlecell dataset composed by 8 clusters in a Seurat object that I converted into a Singlecellexperiment following the indications in the SingleR book. I annotated it on a reference dataset with the following parameters:

table(subset_c2_sce$RNA_snn_res.0.5)

  0   1   2   3   4   5   6   7 
483 453 382 366 335 243 228 122 


pred.subset <- SingleR(test = Counts, ref = subset_c2_sce, labels = subset_c2_sce$RNA_snn_res.0.5,
                       de.method = "wilcox")

table(pred.subset$pruned.labels)

   0    1    3    7 
  59  157 1529  188 

to.remove<- pruneScores(pred.subset, min.diff.med = 0.05)

# Create pruned labels
pruned_labels <- pred.subset$labels
pruned_labels[to.remove] <- "NI"

# Assign pruned labels to the metadata
tbx1$singleR <- as.factor(pruned_labels)
table(tbx1$singleR)

   0    1    3    7   NI 
   5  107 1147   14  660 

My question is... why the clusters 2-4-5-6 have not been annotated?

I tried to annotate them on another dataset, and again, there are only the labels of clusters 0-1-3-7

Thank you in advance

[j-andrews7]

Presumably because your test set matches the cells in your reference from those clusters more closely than the others.

…

On Fri, Mar 21, 2025, 1:36 PM olga ***@***.***> wrote: Hello, I have a singlecell dataset composed by 8 clusters in a Seurat object that I converted into a Singlecellexperiment following the indications in the SingleR book. I annotated it on a reference dataset with the following parameters: table(subset_c2_sce$RNA_snn_res.0.5) 0 1 2 3 4 5 6 7 483 453 382 366 335 243 228 122 pred.subset <- SingleR(test = Counts, ref = subset_c2_sce, labels = subset_c2_sce$RNA_snn_res.0.5, de.method = "wilcox") table(pred.subset$pruned.labels) 0 1 3 7 59 157 1529 188 to.remove<- pruneScores(pred.subset, min.diff.med = 0.05) # Create pruned labels pruned_labels <- pred.subset$labels pruned_labels[to.remove] <- "NI" # Assign pruned labels to the metadata tbx1$singleR <- as.factor(pruned_labels) table(tbx1$singleR) 0 1 3 7 NI 5 107 1147 14 660 My question is... why the clusters 2-4-5-6 have not been annotated? I tried to annotate them on another dataset, and again, there are only the labels of clusters 0-1-3-7 Thank you in advance — Reply to this email directly, view it on GitHub <#296>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACOAQNDU2FKCTP7JR6IKOKT2VRL2VAVCNFSM6AAAAABZQU5YIKVHI2DSMVQWIX3LMV43ASLTON2WKOZSHEZTSMRRGEYTGMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> [image: feanaros]*feanaros* created an issue (SingleR-inc/SingleR#296) <#296> Hello, I have a singlecell dataset composed by 8 clusters in a Seurat object that I converted into a Singlecellexperiment following the indications in the SingleR book. I annotated it on a reference dataset with the following parameters: table(subset_c2_sce$RNA_snn_res.0.5) 0 1 2 3 4 5 6 7 483 453 382 366 335 243 228 122 pred.subset <- SingleR(test = Counts, ref = subset_c2_sce, labels = subset_c2_sce$RNA_snn_res.0.5, de.method = "wilcox") table(pred.subset$pruned.labels) 0 1 3 7 59 157 1529 188 to.remove<- pruneScores(pred.subset, min.diff.med = 0.05) # Create pruned labels pruned_labels <- pred.subset$labels pruned_labels[to.remove] <- "NI" # Assign pruned labels to the metadata tbx1$singleR <- as.factor(pruned_labels) table(tbx1$singleR) 0 1 3 7 NI 5 107 1147 14 660 My question is... why the clusters 2-4-5-6 have not been annotated? I tried to annotate them on another dataset, and again, there are only the labels of clusters 0-1-3-7 Thank you in advance — Reply to this email directly, view it on GitHub <#296>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACOAQNDU2FKCTP7JR6IKOKT2VRL2VAVCNFSM6AAAAABZQU5YIKVHI2DSMVQWIX3LMV43ASLTON2WKOZSHEZTSMRRGEYTGMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[feanaros]

Ok, I thought the same, but if I revert the analysis (using my dataset as test and the other as reference) my clusters are matching with different clusters.

[j-andrews7]

You mean if you swap test and reference? I can imagine that happening for a few reasons, perhaps clustering is more granular in your actual dataset than your original reference or it has more actual cell types.

I don't know how good a match your reference dataset actually is since a large chunk of cells aren't being confidently labeled in your original post.

[feanaros]

Ok, thank you very much for your clarifications!