velocyto on 6 samples with integrated seurat obj.

[dmsalsgh97]

Hi, first of all, thanks for developing Seurat and wrappers! It is awesome.

I'm analyzing 6 samples with Seurat's SCTransform integration method.
Now, I'm trying to project velocyto's output on the umap embedding of the SCTransform integrated Seurat obj.

First, I tried to merge 6 individual .loom files generated by scvelo.py CLI on 10x output.
BUT, When I tried it following @velocyto-team/velocyto.R#54, it needs huge memory and causes MEM allocation error.
(exceeds 100Gb, and I think it is strange...)

sample1 <- ReadVelocity(file = ".../1-sample.loom")
sample2 <- ReadVelocity(file = ".../2-sample.loom")
bm <- merge(x = sample1, y = sample2, merge.data = TRUE)
Error in asMethod(object) :
Cholmod error 'out of memory' at file ../Core/cholmod_memory.c, line 146

My second thought was to aggregate 6 10x outputs using cellranger-aggr and run scvelo.py CLI on the aggregated 10x output and project it on seurat integrate object's umap.

Thanks,

[victorwang123]

How would you solve this problem？