quadbio / simspec

Calculation of Reference/Cluster Similarity Spectrum (RSS/CSS)
30 stars 3 forks source link

Error in checkna(X) : Missing values found in 'X' when running UMAP after #10

Open Kenneth-new-user opened 1 year ago

Kenneth-new-user commented 1 year ago
org2m <- cluster_sim_spectrum(object = org2m, label_tag = "BATCH", cluster_resolution = 0.4, corr_method = "pearson", spectrum_type = "corr_kernel")

The CSS part runs fine to completion and the CSS matrix is populated

Start to do clustering for each sample...
  Done clustering of sample embryo.
  Done clustering of sample P0.
  Done clustering of sample P7.
  Done clustering of sample adult.
Finished clustering.
Calculating average profiles of clusters...
Calculating standardized similarities to clusters...
'as(<dgeMatrix>, "dgCMatrix")' is deprecated.
Use 'as(., "CsparseMatrix")' instead.
See help("Deprecated") and help("Matrix-deprecated").
Doing kernel transformation...
Done. Returning results...
Warning: Number of dimensions changing from 82 to 70

Then when I try to plot the results in a UMAP by running the UMAP function,

org2m <- RunUMAP(org2m, reduction = "css", dims = 1:ncol(Embeddings(org2m, "css")))

The following error pops up.

17:26:25 UMAP embedding parameters a = 0.9922 b = 1.112
Found more than one class "dist" in cache; using the first, from namespace 'spam'
Also defined by ‘BiocGenerics’
Error in checkna(X) : Missing values found in 'X'
zhisonghe commented 1 year ago

Do you have the same issue if you use the default spectrum type "corr_ztransform" instead of the current setting "corr_kernel"?

ChristopherZouNMU commented 4 weeks ago

hi do you have solved this problem? it did not pop up initially, but I encountered it just now

CC-DrDU commented 3 days ago

Do you have the same issue if you use the default spectrum type "corr_ztransform" instead of the current setting "corr_kernel"?

I have the same issue using "corr_ztransform". I am using Seurat V5 and I remember that I have run this step without error when using Seurat V4. Is it the possible reason?

CC-DrDU commented 3 days ago

Do you have the same issue if you use the default spectrum type "corr_ztransform" instead of the current setting "corr_kernel"?

I have the same issue using "corr_ztransform". I am using Seurat V5 and I remember that I have run this step without error when using Seurat V4. Is it the possible reason?

I've check the cell.embedding of the css reduction. Turns out that there are NAs in the cell.embedding. I could continue by converting all the NAs into 0, but is there any other solutions that can solve this issue in a more elegant way? Thank you for your help!