Error in eigs_real_sym; k must satisfy 0 < k < rnow(A)

[VitorAguiar]

Hi, thanks for developing scSHC!

When running the following code:

clusters <- scSHC(SeuratObj@assays$RNA@data, batch = NULL, alpha = 0.05, num_features = 2000, num_PCs = 30, parallel = TRUE, cores = 4)

I get the error:

Error in eigs_real_sym(A, nrow(A), k, which, sigma, opts, mattype  = "sym_matrix", :
    'k' must satisfy 0 < k < nrow(A)
In addition: Warning message:
In eigs_real_sym(A, nrow(A), k, which, sigma, opts, mattype  = "sym_matrix", :
    all eigenvalues are requested, eigen() is used instead

Please, any ideas on how to deal with that?

Thank you!

[igrabski]

Hi Vitor, I haven't seen that error before but it seems like at some point, for some reason, num_PCs is larger than the dimension of the data. Could you please confirm/clarify: (1) is the input data raw counts and of the class matrix or Matrix, (2) what is the dimension of the input data, and (3) how long does the code run before the error appears? Thanks!

[VitorAguiar]

Hi Isabella,

I see now that I need to use raw counts. I was using Seurat's log normalized counts.

It works!

Thank you, Vitor

[Pacomito]

Hello, Than you for this tool, sorry to re-open this issue, the same error happened to me running on scRNA public dataset:

clusters <- scSHC(data, cores = 3) Erreur dans eigs_real_sym(A, nrow(A), k, which, sigma, opts, mattype = "sym_matrix", : 'k' must satisfy 0 < k < nrow(A)

1) Dimension dim(data) [1] ´ 13098 35312´

2) Class

class(data)
[1] "dgCMatrix"
attr(,"package")
[1] "Matrix"

3) Raw counts:

13098 x 35312 sparse Matrix of class "dgCMatrix"
  [[ suppressing 85 column names ‘01:02:16:08’, ‘01:02:21:09’, ‘01:02:25:03’ ... ]]
  [[ suppressing 85 column names ‘01:02:16:08’, ‘01:02:21:09’, ‘01:02:25:03’ ... ]]

Gnai3 1 . . . . . . . . 1 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Pbsn  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Scml2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Apoh  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Cav2  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Wnt9a 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 . . . . 1 . . . . 1 . . . . . . . 1 . . . . . . . . . . . . . . ......

 ..............................
 ........suppression de 35227 colonnes et 13087 lignes dans show() ; ajustez peut-être 'options(max.print= *, width = *)'
 ..............................
  [[ suppressing 85 column names ‘01:02:16:08’, ‘01:02:21:09’, ‘01:02:25:03’ ... ]]

Gm49248 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Gm48169 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Gm47603 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . 1 . . . . . . . ......
Gm47853 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......
Gm47451 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......

4) Time after error =~20-30 min

Thank you in advance, Pacôme

[igrabski]

Hi Pacôme, you mentioned the dataset is public -- would it be possible to share the data so I can take a look at this issue on my end? Thanks!

[Pacomito]

Of course, here is a link: https://we.tl/t-pn25XhCSE6, Thanks

[Pacomito]

Just to let you know, I tried removing empty cells / columns first, which might cause PCA to fail, but it produced the same error indicating it might not be that.

data = data[which(rowSums(data) > 0), ]
data = data[,which(colSums(data) > 0)]
clusters <- scSHC(as.matrix(data), parallel = TRUE, cores = 8)
Erreur dans eigs_real_sym(A, nrow(A), k, which, sigma, opts, mattype = "sym_matrix",  : 
  'k' must satisfy 0 < k < nrow(A)

Best, Pacôme

[igrabski]

Thanks! I think I figured out the issue -- at some point, the number of relevant genes considered in the comparison of two tiny clusters is smaller than the default number of PCs, which leads to an error in the eigendecomposition. I just pushed a simple fix and now your data runs without issues on my end. Please re-install the package and let me know if the issue persists!

[Pacomito]

Thank you, I ran it and it works now.