Can't write loom file from seurat object about loomr HOT 2 OPEN

Hey @christiankroy, Just curious... were you able to solve this issue? I am running the same issue right now when trying to convert an existing seurat object to loom.

from loomr.

Hi @kathirij - I did figure out a work around. I ended up converting to anndata, importing into python, and writing a loom file that way. Here's some example code.

First Seurat::Convert() the Seurat object (so) to anndata format and write a .h5ad file to disk.

NB - you'll want to add as much as possible to the [email protected] slot (dimension reduction coordinates/per cell values/ etc) before exporting because it comes along with the anndata export and is automatically included in the loom file. Otherwise you'll have to add it within the python chunk.

Seurat::Convert(
    from = so,
    to = "anndata",
    filename = "./Seurat_analysis.h5ad", 
    overwrite = TRUE,
    display.progress = TRUE)

Then load the reticulate Rstudio package (I work in Rstudio).

library(reticulate)

Then load a python3 conda environment (I called mine loom below) with scanpy and loompy installed.

reticulate::use_condaenv("loom")

And then within a reticulate/python code block.

import numpy as np
import pandas as pd
import scanpy.api as sc
import scipy.sparse as sparse
import loompy as lp
adata = sc.read_h5ad("./Seurat_analysis.h5ad")
adata.write_loom("./Seurat-based.loom")
ds = lp.connect("./Seurat-based.loom")
ds.ra.GENENAME = list(adata.var_names) # Adds GENENAMES instead of index locations

And you're done.

I've been able to use loom-viewer from the Linnarsson lab to interact with these loom files. Works great and major hat-tip to Linnarsson lab.

from loomr.