Hoohm / dropSeqPipe

A SingleCell RNASeq pre-processing snakemake workflow
Creative Commons Attribution Share Alike 4.0 International
147 stars 47 forks source link

migrate to Seurat version 3 #94

Closed TomKellyGenetics closed 4 years ago

TomKellyGenetics commented 4 years ago

See #93 for previous discussions

seb-mueller commented 4 years ago

Thanks for migrating the PR! Following the discussion on #93, this r.yaml content seem to work for me: It's basically removing all version requirements:

channels:
  - conda-forge
  - bioconda
dependencies:
  - r
  - r-ggplot2
  - r-gridextra
  - r-viridis
  - r-stringdist
  - r-dplyr
  - r-mvtnorm
  - r-seurat=3
  - r-hmisc
  - r-tidyverse
  - r-devtools
  - r-rcolorbrewer
  - font-ttf-dejavu-sans-mono=2.37
  - fontconfig=2.13.1
seb-mueller commented 4 years ago

This seemed to have done the trick. The workflow went almost completly through, but raised an error in the `plot_violine.R':

Rscript --vanilla /home/travis/build/Hoohm/dropSeqPipe/.test/.snakemake/scripts/tmpeo8j9uf6.plot_violine.R
Activating conda environment: /home/travis/build/Hoohm/dropSeqPipe/.test/.snakemake/conda/f4398d85
In debug mode: saving R objects to inspect later
Error in `[.data.frame`(feature.names, , gene.column) : 
  undefined columns selected
Calls: Read10X -> [ -> [.data.frame

Not sure what happened there since I haven't tested your changes yet, do you have a way to test what's going on (it seems the read10x call is the culprit)?

TomKellyGenetics commented 4 years ago

I ran into the same issue. It seems that Read10X expects either features.tsv (and all files as .gz) or genes.tsv. Be default gene names are expected in column 2 but it is possible to change this.

Update: I think it may be easier to keep "genes" instead of "features" rather than compressing output files (as cellranger version 3 does). I have updated to accept "genes.tsv" with one column.

TomKellyGenetics commented 4 years ago

Sorry to spam Travis with test jobs. I've narrowed down the issue, it seems that compress_mtx_summary only compresses the umi directory (not the reads).

Screen Shot 2019-11-25 at 15 44 47
Hoohm commented 4 years ago

Not sure you saw it, but I sent you a PR on your branch. Could you merge it and PR it again? Sorry for the hassle, still struggling to keep PR authors properly in the history when merging. https://github.com/TomKellyGenetics/dropSeqPipe/pull/1

Hoohm commented 4 years ago

This looks nice. @seb-mueller if you don't have any other requests, I'll merge it

seb-mueller commented 4 years ago

I've crudely tested it and it seems to work fine, so fine with me to merge it in. @TomKellyGenetics, thanks so a lot for this, this is really helpful!