anndatar

anndataR: H5AD to R Data Conversion

This skill enables seamless conversion between .h5ad files (AnnData format from Python/scanpy) and R single-cell objects (Seurat, SingleCellExperiment) using the anndataR package.

Installation

Install anndataR using BiocManager:

if (!requireNamespace("BiocManager", quietly = TRUE)) {
  install.packages("BiocManager")
}
BiocManager::install("anndataR")

Core Conversion Functions

Reading .h5ad Files

Read as AnnData object (default - in-memory):

library(anndataR)
h5ad_path <- "path/to/your/file.h5ad"
adata <- read_h5ad(h5ad_path)

Read as SingleCellExperiment:

sce <- read_h5ad(h5ad_path, as = "SingleCellExperiment")

Read as Seurat:

obj <- read_h5ad(h5ad_path, as = "Seurat")

Read as HDF5-backed AnnData (memory-efficient for large files):

adata <- read_h5ad(h5ad_path, as = "HDF5AnnData")

Converting Between Formats

AnnData → SingleCellExperiment:

sce <- adata$as_SingleCellExperiment()

AnnData → Seurat:

obj <- adata$as_Seurat()

SingleCellExperiment → AnnData:

adata <- as_AnnData(sce)

Seurat → AnnData:

adata <- as_AnnData(obj)

Writing .h5ad Files

Write AnnData to disk:

tmpfile <- tempfile(fileext = ".h5ad")
adata$write_h5ad(tmpfile)
# Or: write_h5ad(adata, tmpfile)

Write SingleCellExperiment to disk:

write_h5ad(sce, tmpfile)

Write Seurat to disk:

write_h5ad(obj, tmpfile)

Working with AnnData Objects

Accessing Slots

# Dimensions
dim(adata)
nrow(adata)
ncol(adata)

# Observation metadata (cells)
adata$obs        # Returns data.frame
adata$obs[1:5, ]  # First 5 cells

# Variable metadata (genes)
adata$var         # Returns data.frame
adata$var[1:5, ]  # First 5 genes

# Main expression matrix
adata$X

# Additional matrices (layers)
adata$layers$counts
adata$layers$dense_X

# Embeddings (obsm)
adata$obsm$X_pca
adata$obsm$X_umap

# Gene loadings (varm)
adata$varm$PCsstructured metadata

# Un (uns)
adata$uns$leiden
adata$uns$pca