search

satijalab / seurat

R toolkit for single cell genomics
http://www.satijalab.org/seurat
Other
2.24k stars 901 forks source link

Lean Seurat object #996

Closed alextamburino closed 5 years ago

alextamburino commented 5 years ago

Hi, I have a fully processed Seurat object and I want to create a lean version (i.e. minimal size) for a Shiny app that enables plotting (@data and @scale.data) and has 1-2 @meta.data variables. What's the easiest way to do this? Thanks in advance for your time and thoughts. Alex

sansense commented 5 years ago

You can save only the required data as a df

df = FetchData(Obj, c(rownames(Obj@data), "meta.data.var1", "meta.data.var2"), use.scaled = T)
alextamburino commented 5 years ago

Thanks @sansense I built the Shiny app to plot from a Seurat object. Is there a way to save only the required data as a Seurat object for use with VlnPlot, DotPlot, etc.? The app works as is, just slow and I'm trying to avoid redeveloping it.

maximilianh commented 5 years ago

Sorry, shameless plug for my software, but you could export your data from the Seurat object to a cellbrowser, which does not use shiny so is probably a lot faster. See cells.ucsc.edu or https://github.com/maximilianh/cellBrowser (needs more docs) and contact me. It can only color by gene or cell annotation, so it may not do what you want here.

sansense commented 5 years ago

@alextamburino,

The minimum Seurat object will contain at least the @raw.data slot. As soon as you enter the raw.data in Seurat, the @ data slot is also populated with the same data (otherwise the normalized data goes here). So, create a nea lean Seurat Obj from the normalized data of bigObj

leanObj = CreateSeuratObject(bigObj@data)

## Almost all of the analyses work on scaled.data, so, it makes sense to populate also the scale data slot
leanObj@scale.data = bigObj@scale.data

## And then copy the required meta.data to your lean obj
leanObj@meta.data = bigObj@meta.data ## or only a part of the meta data that you need
mojaveazure commented 5 years ago

Seurat v3 introduces a DietSeurat function. This function allows you to select which slots (counts/raw.data, data, and/or scale.data) to keep in an object, and which features to keep (eg. only highly variable features). Currently, the function doesn't support bringing over meta data, but that's relatively easy to transfer over.

Details on installing the Seurat v3 prerelease can be found here