diffusion maps "problem too large"

[talashuach]

I'm following the tutorial: https://github.com/r3fang/SnapATAC/blob/master/examples/10X_brain_5k/README.md Using a dataset with ~50K cells.

The diffusion maps stage crashes after a short while with this error:

x.sp = runDiffusionMaps(obj=x.sp, input.mat="bmat", num.eigs=50) Epoch: checking the inputs ... Epoch: computing jaccard similarity matrix ... Error in .local(x, y, ...) : Cholmod error 'problem too large' at file ../Core/cholmod_sparse.c, line 92

traceback:

7: .Call(Csparse_Csparse_crossprod, x, y, trans = TRUE, boolArith = boolArith) 6: .local(x, y, ...) 5: Matrix::tcrossprod(X_i, X_j) 4: Matrix::tcrossprod(X_i, X_j) 3: calJaccard(data.use1, data.use2) 2: runJaccard2(obj, obj, input.mat = input.mat) 1: runDiffusionMaps(obj = x.sp, input.mat = "bmat", num.eigs = 50)

Any advice on how to get through / around this? Thanks!

[znavidi]

Hi, I have the same problem with runDiffusionMaps. Have you solved this issue? if yes, could you please share it with me?

Many thanks,

[znavidi]

I would really appreciate it if someone could get a hint about it if possible. @r3fang @talashuach

Best,

[r3fang]

hello,

Please check the second example: https://github.com/r3fang/SnapATAC/blob/master/examples/10X_PBMC_15K/README.md

in which a subset of cells (10K) are selected as landmarks for dimensionality reduction and project the remaining cells (~5K) to the embedding. Similarly, you can first randomly sample 10K cells as landmarks and project the remaining cells ~40k to the embedding.

Best -Rongxin