Loading data

To view a dataset in Portable-CELLxGENE, simply create a folder containing the .h5ad files of interest and select this folder when prompted by the software. Rendered notebooks can be made accessible to the user by storing them within this folder in a subfolder named rendered_notebooks.

Creating an h5ad file from your data

If you want to distribute your own dataset for others to view in Portable-CELLxGENE, you first need to save it as an h5ad file. This is the native format of the scanpy package in Python. If your dataset is called cells, then it can be saved in this format with:

cells.write_h5ad("my_dataset.h5ad")

Converting Seurat objects to .h5ad files

If you are instead analysing your dataset in Seurat, then you can save your data as an h5ad using the SeuratDisk package. By default, the .h5ad files created by the standard Seurat to Anndata conversion process (detailed here) contain only the highly variable genes, and therefore other genes cannot be annotated in CELLxGENE. This can be fixed by removing the scale.data slot of the Seurat object (along with the count data, for memory reasons) with the DietSeurat function. This object can then be converted to a .h5ad file using the process detailed in the link above. Putting this together:

# For each Seurat object you want to include in Portable-CELLxGENE:
# Assume Seurat object is called "cells"
library(Seurat)
library(SeuratDisk)
# Remove everything except data slot and UMAP reduction
cells <- DietSeurat(cells, scale.data = FALSE, counts = FALSE,
                        dimreducs = "umap")
SaveH5Seurat(object = cells, filename = "cells.h5Seurat")
Convert(source = "cells.h5Seurat", dest = "h5ad")

If you experience problems with the above process, an alternative approach is to use the zellkonverter package. If you still cannot save your data in this format, submit an issue on GitHub.

Analysing public data

The h5ad format is often used to distribute single-cell transcriptomic datasets. All datasets in the CELLxGENE Discover database can be downloaded in this format by clicking “Download” and then selecting .h5ad under “Data format”. The downloaded files can be loaded in Portable-CELLxGENE as detailed at the top of this page.