Error in make_cell_attr(umi, cell_attr, latent_var, batch_var, latent_var_nonreg, : cell attributes cannot contain any NA, NaN, or infinite values

[connorhknight]

Hi, I am analysing the 4 pancreas samples after applying decontX from the celda package. Some of the samples are processing correctly. However, randomly and inconsistently i receive this error:

Error in make_cell_attr(umi, cell_attr, latent_var, batch_var, latent_var_nonreg, :
cell attributes cannot contain any NA, NaN, or infinite values

Why could this be happening?

[ChristophH]

Do all input cells have at least one count?
Please provide the exact steps to reproduce the problem.

[RajneeshSrivastava]

Any luck resolving this issue?

[ChristophH]

sctransform::vst makes some basic assumptions about the input data. The error message mentioned above suggests that either the input umi matrix was not made up of non-negative counts, or there were cells (columns) that did not sum up to a value > 0 (total number of counts <= 0). Please check your input and make sure that those assumptions are met.

[smaniatis]

I'm getting this on Visium array .h5 output from spaceranger and loaded into Seurat via Load10X_Spatial(). as far as I can see, the counts matrices do not violate the assumptions above. Any advice?

[ChristophH]

@smaniatis I would need the data to figure out what exactly the problem is. If you share the exact commands you used and the input (.h5 file or counts matrix) I will have a look.

[zhiiiyang]

I encountered the same error too but solved it by filtering those genes with all zero counts.

[rocketeer1998]

I encountered the same error too but solved it by filtering those genes with all zero counts.

I agree. With sample<- subset(sample, subset = nCount_Spatial > 0) command.

[ATpoint]

A reason can be that QC was not done properly and you try to run the function on cells without any counts, meaning all gene in that cell are zero. You can check that with colSums() on your count matrix, removing those with all-zeros.

[troysgit]

Would it be worthwhile amending the error message to include, "a high % of zeroes in some columns" or similar?

[lucygarner]

I got this error as well, but I think it was caused by genes that had zero counts rather than cells. Is this possible?

[IoannisKonstantinidis]

min(obj$nFeature_RNA)
[1] 739
min(obj$nCount_RNA)
[1] 335
max(obj$nFeature_RNA)
[1] 34139
max(obj$nCount_RNA)
[1] 95475
obj <- SCTransform(obj, verbose = FALSE)
Error in make_cell_attr(umi, cell_attr, latent_var, batch_var, latent_var_nonreg, :
cell attribute "log_umi" contains NA, NaN, or infinite value

This seurat object was created from the output of cellranger-arc aggr. It contains both snATAC and snRNA data from 4 biological samples. Any idea why I am getting this error?