ncats / cellranger-snakemake Goto Github PK

This has been a request for a while now.

foreach ( ) %dopar% is failing sporadically. Determine the cause of these failures.

Remove a bunch of the 'intermediate' files that are of no interest after the pipeline is finished.
Be sure to document which files are removed.

FindAllMarkers has been added, and code is in place that should create the heatmap for Seurat objects containing merged 'groups':

if ( str_detect( newname, 'group_') )  {
  
    message( 'Running FindAllMarkers')
    DE_markers <- FindAllMarkers(x)
  
    clusters_up   <- setorder( setDT( DE_markers ), -avg_logFC )[ , head( .SD, 100 ), keyby='cluster' ]
    clusters_down <- setorder( setDT( DE_markers ),  avg_logFC )[ , head( .SD, 100 ), keyby='cluster' ]
  
    pdffile = str_replace( seurat_filepath, "seurat\\.Rdata", "heatmap.pdf" )
    pdfs[[pdffile]] <- DoHeatmap( x, features = c( clusters_up$V1, clusters_down$V1 ), raster=F ) +
      scale_fill_gradientn( colors = rev( RColorBrewer::brewer.pal( n = 10, name = "RdBu" )) ) +
      guides( color = FALSE)

  }

... but it is unclear whether this is working or not, as the .pdf files are not being generated.

NOTE: This issue may be a non-issue if the .pdf file-making issues are resolved and they appear to be ok.

There is a list of house-keeping genes, pseudo-genes, etc. that should be removed from these kinds of analysis. Import that list and remove them from the analysis, probably after the conversion to gene symbols from ensg.genes.

An attempt has been made at having the pipeline produce pdf files of the following plots:

• pca
• tSNE
• UMAP
• DE Heatmap

However, having the pipeline create and output the plots in parallel is presenting challenges. As of now, the code tries to store the plots in a list, and outside of the foreach loop print them to disk like so:

for ( pdffile in pdfs ) {
 
  pdf(pdffile)
  pdfs[[pdffile]]
  dev.off()

}

This produces no image files.
Investigate R best-practices for this and implement them.

Documentation needs to be provided that explains the setup and execution of the pipeline, including:

• Required inputs
• Pre-requisites for execution (biowulf acct, etc.)
• Steps for typical execution
• Description of output.

Ideally, this documentation will exist as perldoc within run_cellranger.pl and in README.md of this repository.

Currently the pipeline is using ensg.genes if it exists in the working directory. It should default to looking for it in the ./bin dir. There should be a way to have it copy the file into the workdir; and retain the ability to have it newly created, if desired.