This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Speed up Seurat FindClusters execution?

Hi all,

I am processing a scRNA-seq dataset of 200k cells, and am at the stage of finding clusters with a resolution of 2. The FindClusters function from Seurat seems to take a long time to run. Wondering if anyone have any ideas how to speed up the execution, and which method is most efficient (with less tradeoff in the clustering quality). Currently, I have three alternatives in mind:

1) Decreasing the resolution from 2 to 0.8 or even lower

2) Use sparse matrix with do.sparse = TRUE (but this is no longer available in Seurat V5, is there a workaround?)

3) Process it on HPC (my computer is quite powerful (64Gb RAM) and the HPC queueing usually takes a long time if a lot of resources are demanded, also prone to error hence re-queueing).

Thanks for the help.

seurat
• 3,173 views • link
updated by E. 0 • | written by CTLong 14

any solutions ?

• 0 views • link
written by Picasa 71

I overcame this by running it on HPC

• 0 views • link
written by CTLong 14

2 answers

Unfortunately, FindClusters works in parallel (future) only when multiple resolution are passed ( I assume 1 cpu x resolution). That was true for seurat v4, and I am not sure if it is still true with seurat 5, as they remvoed that vignette

• 0 views • link
written by fracarb8 176

use presto package on the HPC https://github.com/immunogenomics/presto

• 0 views • link
written by E. 0

Please edit the post and include links for the software when recommending packages. There can be multiple packages with same/similar names that may be completely unrelated..

• 0 views • link
written by GenoMax 16K

Log in to answer this question.

More posts like this