aet21 / ELVAR

Extended Louvain Clustering algorithm for applications in scRNA-Seq data analysis
GNU General Public License v3.0
5 stars 0 forks source link

Speed considerations #2

Open LinearParadox opened 1 year ago

LinearParadox commented 1 year ago

A single run of eva.o takes around 15-20 mins. To run it 100 times (as suggested in the manual) would obviously take a pretty long time. I was wondering if there are any parallelization or speedup options available?

aet21 commented 1 year ago

Yes, you are right, if you have lots of cells, one run can take this long, but nowadays we do normally have access to servers with 100 nodes or so, and so 100 runs are achievable within 30 mins. To run in parallel mode should be quite easy with the parallel R-library, but I agree for those unfamiliar with this, we should make the parallel option available within the package. We will work on it and update package fairly soon. If you can't wait, a quick fix would be to refine Eva_partitions function (call this new function myEva_partitions) so that the 1st argument is just a dummy integer index, then you could just do library(parallel); nR <- 100; ## number of runs desired idx.l <- as.list(1:nR); mcl.o <- mclapply(idx.l,myEva_partitions,all other arguments here,mc.cores=100); each entry of mcl.lo contain the output of one Eva run

kr A


Professor Andrew E Teschendorff Computational Systems Epigenomics Lab CAS Key Lab of Computational Biology Shanghai Institute for Nutrition and Health Chinese Academy of Sciences 320 Yueyang Road Shanghai 200031 Tel: +86 183-1704-7442 email: @.*** web: https://aeteschendorff-lab.github.io



From: @. @.> on behalf of Alexei Martsinkovskiy @.> Sent: Saturday, July 15, 2023 7:11 AM To: aet21/ELVAR @.> Cc: Subscribed @.***> Subject: [aet21/ELVAR] Speed considerations (Issue #2)

A single run of eva.o takes around 15-20 mins. To run it 100 times (as suggested in the manual) would obviously take a pretty long time. I was wondering if there are any parallelization or speedup options available?

— Reply to this email directly, view it on GitHubhttps://github.com/aet21/ELVAR/issues/2, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AE6NRFRFIVSGF3NQFH63Q2TXQHGZVANCNFSM6AAAAAA2K4NFUY. You are receiving this because you are subscribed to this thread.Message ID: @.***>