Skip to content

dynverse/anndata

Repository files navigation

anndata for R

CRAN CRAN Downloads R-CMD-check Codecov test coverage

anndata is a commonly used Python package for keeping track of data and learned annotations, and can be used to read from and write to the h5ad file format. It is also the main data format used in the scanpy python package (Wolf, Angerer, and Theis 2018).

However, using scanpy/anndata in R can be a major hassle. When trying to read an h5ad file, R users could approach this problem in one of two ways. A) You could read in the file manually (since it’s an H5 file), but this involves a lot of manual work and a lot of understanding on how the h5ad and H5 file formats work (also, expect major headaches from cryptic hdf5r bugs). Or B) interact with scanpy and anndata through reticulate, but run into issues converting some of the python objects into R.

We recently published anndata on CRAN, which is a reticulate wrapper for the Python package – with some syntax sprinkled on top to make R users feel more at home.

anndata for R is still under active development at github.com/dynverse/anndata. If you encounter any issues, feel free to post an issue on GitHub!

Installation

You can install anndata for R from CRAN as follows:

install.packages("anndata")

Normally, reticulate should take care of installing Miniconda and the Python anndata.

If not, try running:

reticulate::install_miniconda()
anndata::install_anndata()

Getting started

The API of anndata for R is very similar to its Python counterpart. Here is an example:

library(anndata)
## 
## Attaching package: 'anndata'

## The following object is masked from 'package:readr':
## 
##     read_csv
ad <- read_h5ad("example_formats/pbmc_1k_protein_v3_processed.h5ad")

ad
## AnnData object with n_obs × n_vars = 713 × 33538
##     var: 'gene_ids', 'feature_types', 'genome', 'highly_variable', 'means', 'dispersions', 'dispersions_norm'
##     uns: 'hvgParameters', 'normalizationParameters', 'pca', 'pcaParameters'
##     obsm: 'X_pca'
##     varm: 'PCs'
Matrix::rowMeans(ad$X[1:10,])
## AAACCCAAGTGGTCAG-1 AAAGGTATCAACTACG-1 AAAGTCCAGCGTGTCC-1 AACACACTCAAGAGTA-1 
##         0.06499579         0.06385104         0.06102355         0.06739055 
## AACACACTCGACGAGA-1 AACAGGGCAGGAGGTT-1 AACAGGGCAGTGTATC-1 AACAGGGTCAGAATAG-1 
##         0.08891241         0.08648681         0.09318970         0.09140243 
## AACCTGAAGATGGTCG-1 AACGGGATCGTTATCT-1 
##         0.06664118         0.07866523

See ?anndata for a full list of the functions provided by this package. Check out any of the other vignettes by clicking any of the links below:

Future work

In some cases, this package may still act more like a Python package rather than an R package. Some more helper functions and helper classes need to be defined in order to fully encapsulate AnnData() objects. Examples are ad$chunked_X(...), backed file modes, read_zarr() and ad$write_zarr().

References

Wolf, F Alexander, Philipp Angerer, and Fabian J Theis. 2018. “SCANPY: Large-Scale Single-Cell Gene Expression Data Analysis.” Genome Biology 19 (February): 15. https://doi.org/10.1186/s13059-017-1382-0.