search

ebi-gene-expression-group / container-galaxy-sc-tertiary

Galaxy container for single cell RNA-Seq tertiary analysis tools
MIT License
16 stars 12 forks source link

SCEasy - AnnData-->Seurat #225

Open nomadscientist opened 3 years ago

nomadscientist commented 3 years ago

AnnData --> Seurat, particularly now there is Seurat v4 trajectory stuff/automated annotation stuff/Monocl can import Seurat objects... this is important!

nh3 commented 3 years ago

anndata -> seurat is in the github version of sceasy, but not released yet. there is also anndata -> monocle cds in it.

I was hoping to also get the dependency issues fixed in the new release to support anndata >= 0.7 (with the cost of dropping support for exchangeable loom but still supports standard loom). Full automatic interoperability at intermediate steps without user input is quite unrealistic and perhaps not of high usage. What do people think?

nomadscientist commented 3 years ago

That's 9 kinds of awesome! Horray! "Full automatic interoperability at intermediate steps without user input is quite unrealistic and perhaps not of high usage." --> So a user example: "I've analysed all my stuff in scanpy, I found cell types and marker genes and annotated them. Now I want to do a trajectories analysis, which is superior in monocle as compared to scanpy. I now want to convert my object into Monocl object". This would be a common use. Second example: "I've input all my metadata (batch, sex, gender, etc.) but after analysing it in Scanpy I've decided Seurat is a better option. I want to take my annotated AnnData object and analyse it in Seurat instead" or, on that line, "I want to now compare the clusters found in Scanpy to those in Seurat, so keeping the cluster-calls from Scanpy as I convert to Seurat would be good"

nh3 commented 3 years ago

those use cases can be covered without the "exchangeable loom". It was for things in uns of anndata and non-standard information in @metadata of sce. Information you mentioned are in obs and colData and will be passed on.

pcm32 commented 3 years ago

I have been trying the new SeuratDisk for Seurat v4, but at least with the examples I have tried it seems broken, and a lot of people are complaining for issues with the HDF5 capability of SeuratDisk (for Seurat v4). On Seurat v3 we have been able to read annData for sometime now (I haven't tested it with AnnData 0.7.x though).

Is SeuratDisk working for your @nh3 ? or how are you doing AnnData -> Seurat? Or you are doing it on Seurat v3?

What I was thinking was to add a Seurat v3 converter with all the formats that we support already and release that before moving Seurat scripts to v4. Then one could do AnnData -> Seurat v3 -> read it on a v4 tool (that should work I think).

nh3 commented 3 years ago

We are still working with Seurat v3. Never tried v4. Sorry that I didn't realise this is v4 specific.

nomadscientist commented 2 years ago

@pcm32 Did your fix for this work? AnnData --> Seurat would be great!