Python Integration with anndataR

Introduction

anndataR works with Python AnnData objects through reticulate. You can load Python objects, apply Python functions to them, and convert them to Seurat or SingleCellExperiment objects.

message(
  "Python packages scanpy and mudata are required to run this vignette. Code chunks will not be evaluated."
)
#> Python packages scanpy and mudata are required to run this vignette. Code chunks will not be evaluated.

Prerequisites

This vignette requires Python with the scanpy and mudata packages installed. If these are not available, the code chunks will not be evaluated but the examples remain visible.

Basic Integration with scanpy

Install required Python packages if needed:

reticulate::py_require("scanpy")
library(anndataR)
library(reticulate)
sc <- import("scanpy")

Load a dataset directly from scanpy:

adata <- sc$datasets$pbmc3k_processed()
adata

Apply scanpy functions directly:

sc$pp$filter_cells(adata, min_genes = 200L)
sc$pp$normalize_total(adata, target_sum = 1e4)
sc$pp$log1p(adata)

Conversion to R objects

Convert to SingleCellExperiment (see vignette("usage_singlecellexperiment")):

sce_obj <- adata$as_SingleCellExperiment()
sce_obj

Convert to Seurat (see vignette("usage_seurat")):

seurat_obj <- adata$as_Seurat()
seurat_obj

Multi-modal data with mudata

Install required Python packages if needed:

reticulate::py_install("mudata")
md <- import("mudata")

Load a MuData object from file:

cache <- BiocFileCache::BiocFileCache(ask = FALSE)
h5mu_file <- BiocFileCache::bfcrpath(
  cache,
  "https://github.com/gtca/h5xx-datasets/raw/b1177ac8877c89d8bb355b072164384b4e9cc81d/datasets/minipbcite.h5mu"
)

mdata <- md$read_h5mu(h5mu_file)

Access individual modalities and convert them:

rna_mod <- mdata$mod[["rna"]]

rna_seurat <- rna_mod$as_Seurat()
print(rna_seurat)

rna_sce <- rna_mod$as_SingleCellExperiment()
print(rna_sce)

Session info

R

sessionInfo()

Python

reticulate::py_config()

reticulate::py_list_packages()