Error in CalcPerturbSig

[Saharprh]

Hello,

I have an issue with running the CalcPerturbSig on my own data.
seurat_objcompass_perturb <- CalcPerturbSig(
object = seurat_objcompass,
assay = "RNA",
slot = "data",
gd.class ="gene",
nt.cell.class = "neg",
reduction = "pca",
ndims = 40,
num.neighbors = 20,
new.assay.name = "PRTB_compass",
split.by = NULL) # to specify the metadata column if multiple biological contexts (e.g., cell lines) exist

I received this error message:
Error in nn2(data = numeric(0), query = c(-9.89631522133093, -15.278225783592, :
Cannot find more nearest neighbours than there are points.

I appreciate if you can help me solve this issue.
Thank you!
Sahar

[longmanz]

Hello Sahar,
Please consider reducing the "num.neighbors" to a smaller number (e.g., 10). May I ask, how many non-targeting control cells are there in your dataset?

[Saharprh]

Thank you for your response. I changed the "num.neighbors" to different values, but it did not solve the issue. We have four non-targeting controls. How many non-target did you have?

[longmanz]

Hi, do you mean four non-target cells or four non-target gRNAs? In our data, we had 14 non-target gRNAs and ~a few thousands non-target cells.

[Saharprh]

We had: Target gRNAs: 18
Non-targeting gRNAs: 5
And the same, a few thousands non-target cells

[azbukinanadezda]

Hi,
Is it possible to add the feature, that it does not output the error, but warning, the n of neighbours is more than n cells, so for this cell type it is not calculated? I am running it for a dataset with many cell types, some of them are small and have few perturbed cells and it fails the run completely. One can for sure exclude this cell type in advance, but would be good to have warning instead of an error message an abortion of the run... thank you!!!