search

SydneyBioX / scMerge

Statistical approach for removing unwanted variation from multiple single-cell datasets
https://sydneybiox.github.io/scMerge/
66 stars 13 forks source link

Unexpected error in specifying kmeansK parameters #13

Closed makimenkova closed 4 years ago

makimenkova commented 5 years ago

Hi,

I'm interested in integrating sc-RNA data sets and your package seems very suitable for this. However I get this error while running scMerge.

sce_mESC_mg <- scMerge::scMerge(sce_combine = sce_combine, ctl = ctl, kmeansK = c(1, 2), assay_name = 'scMerge')

Step 1: Computation will run in serial
Error in igraph::fastgreedy.community(g) : 
  At fast_community.c:639 : fast-greedy community finding works only on graphs without multiple edges, Invalid value

For any other value of the kmeansK parameter everything works, but I need exactly kmeansK = c(1, 2). I tried to change data sets order and specify kmeansK = c(2, 1), but it didn't work

Step 1: Computation will run in serial
Error in utils::combn(c(seq_len(batch_num))[-batch_oneType], 2) : n < m

Thanks in advance, Maria

kevinwang09 commented 5 years ago

Hi Maria, We have some ideas as to why this occurred. However, in order to help us to diagnose the problem, could you run devtools::session_info() and post the relevant lines for scMerge? Many thanks, Kevin

makimenkova commented 5 years ago

Hello! Here is my session_info:

─ setting  value                       
 version  R version 3.6.1 (2019-07-05)
 os       Ubuntu 18.04.2 LTS          
 system   x86_64, linux-gnu           
 ui       RStudio                     
 language (EN)                        
 collate  en_US.UTF-8                 
 ctype    en_US.UTF-8                          

─ Packages 
 scMerge                1.0.0      2019-05-02 [1] Bioconductor

Thanks

kevinwang09 commented 5 years ago

Hi Maria, That is very helpful, thank you. Could I ask you to try to install the latest version using devtools::install_github("SydneyBioX/scMerge, force = TRUE)? This should install Version 1.1.3 which might already provided a fix. Thanks, Kevin

Get Outlook for Androidhttps://aka.ms/ghei36

From: makimenkova notifications@github.com Sent: Thursday, July 25, 2019 10:13:36 PM To: SydneyBioX/scMerge scMerge@noreply.github.com Cc: Kevin Wang kevin.wang@sydney.edu.au; Comment comment@noreply.github.com Subject: Re: [SydneyBioX/scMerge] Unexpected error in specifying kmeansK parameters (#13)

Hello! Here is my session_info:

─ setting value

version R version 3.6.1 (2019-07-05)

os Ubuntu 18.04.2 LTS

system x86_64, linux-gnu

ui RStudio

language (EN)

collate en_US.UTF-8

ctype en_US.UTF-8

─ Packages

scMerge 1.0.0 2019-05-02 [1] Bioconductor

Thanks

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://protect-au.mimecast.com/s/xhOdCNLwM9iQ1V85u4PDBm?domain=github.com, or mute the threadhttps://protect-au.mimecast.com/s/bCP0COMxNytgo5jrFkrPdC?domain=github.com.

makimenkova commented 5 years ago

Thanks! I did it, but it didn't help unfortunately V scMerge 1.0.0 2019-07-25 [1] Github (SydneyBioX/scMerge@1936764)


Error in igraph::fastgreedy.community(g) : 
  At fast_community.c:639 : fast-greedy community finding works only on graphs without multiple edges, Invalid value```
kevinwang09 commented 5 years ago

Hi, It seems like you still have scMerge V1.0.0 installed. This is an outdated version, so you should try to run (edit) devtools::install_github("SydneyBioX/scMerge, force = TRUE) BiocManager::install("scMerge", version = "3.10) , restart R and try to load scMerge again. You should have (edit) V1.1.4 V1.1.3 installed after this.

As for the error itself, I wonder if you see a plot that looks somewhat like this one before the error occur. If so, could you share that with us?

image

Thanks, Kevin

DarioS commented 5 years ago

... 1.0.0 ... is an outdated version ...

Not really. Almost all end-users should be using the release branch, which is at 1.0.0. The development branch is for the package's developer and early adopters / testers. Are there changes to the developer branch which ought to be ported to the release branch? Installing using devtools::install_github just annoys Martin Morgan on the mailing list when users end up with a mix of incompatible R and R package versions, which wouldn't happen if BiocManager is used.

kevinwang09 commented 5 years ago

@DarioS, scMerge V1.0.0 is fine as is. But after the BioC release this year, we went on to expand scMerge's capabilities in V1.1.z, including resolving this issue.

Maybe for the sake of consistency, @makimenkova, could you try BiocManager::install("scMerge", version = "3.10) instead? Once we resolve this issue, I will push the patched version to BioC.