migrate to Seurat version 3

[TomKellyGenetics]

See #93 for previous discussions

[seb-mueller]

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]

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]

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]

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).

[Hoohm]

What does this do?

I changed the snakemake object to a list since that's what the seurat object was asking but I haven't doublechecked how it looked like.

[TomKellyGenetics]

Seurat version 3 uses “Assay” slots so I think it needs to be specified differently. Their function seems to be recommended for this so they can update it if they change the object in future versions. The test environment on Travis threw an error on this and this fixes it.

[Hoohm]

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.
TomKellyGenetics#1

[Hoohm]

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

[seb-mueller]

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!