How should I download the entire dataset where these single-cell data have already been mapped to brain slices?

[seasonzhao]

Dear Author

How should I download the entire dataset where these single-cell data have already been mapped to brain slices?

[danielsf]

I suspect you will get a more helpful answer posting to the Allen Brain Map Community Forum. Questions there are more likely to get to the science teams that actually did the analysis. I'm not sure that the single cell transcriptomics data was mapped to brain slices (I know that the spatial transcriptomics data was mapped to cell types defined by the single cell transcriptomics data). The science teams will be able to correct me if I am wrong, though.

[danielsf]

For instructions on how to do a bulk download of ABC Atlas data, see this issue. To download all the spatial transriptomics data, you will likely want to call get_directory_data() on the directory 'MERFISH-C57BL6J-638850'.

[seasonzhao]

Thank you very much for your response! What I meant was whether there is one or multiple .h5ad files that contain the spatial location information of all cells across the brain sections shown in the figure above, as well as the expression of each gene (the mapped 30,000+ genes). I would like to use this unified dataset for analysis. Thank you! Looking forward to your reply.

[danielsf]

The single cell RNA seq data was not mapped to fine-grained anatomical structures. The best we have is a course-grained region of interest label associated with each cell, which can be accessed from the cell_metadata.csv file as discussed in this Jupyter notebook.

If that is too imprecise for you and you do not mind losing some precision in the gene expression profile, there is a dataset available which uses the original 500 gene MERFISH spatial transcriptomic gene expression profiles of our spatial transcriptomic dataset to impute the gene expression profile in a broader 8,000 gene panel dataset (see this Jupyter notebook). While the gene expression profiles in this dataset are a theoretical prediction, the spatial locations of these cells are as precise as our spatial transcriptomics pipelines can make them.

A third option is to consult this community post, which discusses how one might use cell type assignments to infer anatomical regions (though this is a probabilistic process as there is not a 1:1 mapping between transcriptomic cell types and anatomical region).