Accepted File Formats

Cellar accepts .h5ad, .csv, .h5, and .tar.gz files.

Internally, Cellar uses the AnnData data structure. These objects can be uploaded to Cellar as .h5ad files. We also accept .csv or .h5 files or feature barcode sequences in cellranger or spaceranger format as .tar.gz files. The archive must contain barcodes.tsv, genes.tsv, and matrix.mtx files (possibly gzipped). In the case of .h5ad or .csv files, please make sure the data follows the cell x feature format, i.e., rows of the matrix correspond to cells and columns to features (genes, proteins). Additionally, row and columns names must be provided with the file.

• Example cell barcode: TTACCATTCCAGCACG.
  Example gene name (ensembl format or HGNC symbol): ENSG00000211699 or TRGV3.

We recommend uploading the data as an .h5ad file. An AnnData object stores expression matrices and annotations efficiently while also providing convenient methods for accessing and storing any additional unstructured objects. Furthermore, an AnnData object can be highly compressed, thus saving precious upload (download) time.

Here we provide a Python snippet that can convert a .csv file into a maximally compressed AnnData object that can be uploaded to Cellar. All you need is the anndata Python package that can be installed via pip install anndata. The following code reads a csv file path/to/your/csv/file.csv and writes it to another file new/file/name.h5ad.

import anndata
adata = anndata.read_csv('path/to/your/csv/file.csv')
adata.write('new/file/name.h5ad', compression=9)

The file can now be uploaded to Cellar. This procedure may reduce a dataset’s size from several GBs to just a few MBs.