FISHscale to anndata and Marker gene plotting

[ernohanninen]

Hi,
I am working with the spatial data presented in your developing human brain paper. I would like to map RNA-seq data against your spatial data using Cell2Location tool.

1. Is there a way to convert the FISHscale data to anndata object?
2. Are there tools available that enables mapping RNA-seq data against FISHscale object?

Thanks!

[larsborm]

Hi,
Thanks for the interest in FISHscale. We do not have a direct method to make an AnnData object. But you can simply make it yourself. I would suggest you make hexagonal bins using the .hexbin_make() function, and then take that Pandas Dataframe to make the AnnData object for Cell2Location. To map single cell to a FISHscale object, we have Bonefight, which is a generalized version on Tangram (https://www.nature.com/articles/s41592-021-01264-7 please cite them if you use Bonefight). Please see the tutorials to see how to map single cell data spatially.
Good luck!

[ernohanninen]

Thanks for the help. I'll try this conversion approach.

I used bonefight to map our RNA-seq data against the developing brain spatial data and to validate the results I would like to plot few marker genes on a hexbin plot. For the marker gene plotting I tried to follow the FISHscale tutorials.

Here is a description of my code:

1. I read and process the spatial data in the way that was done here: https://github.com/linnarsson-lab/developing-human-brain/blob/main/notebooks/Figure2_Spatial_mapping_of_clusters.ipynb
2. I read my data in adata format and create a cluster expression (mean expression of a gene) matrix from it, and run bonefight:

From the tutorials, I understand that for bonefight I should provide a vector containing the number of cells in each cluster. I assume that this should be ordered like the cluster columns in the cluster expression matrix, am I right?

3. Then I run the smoothing like in the notebook I attached
4. Finally I add the colors and plot the mapping result:

After this, I would like to color this hexbin_plot with marker genes. What is the easiest way to do this in my case?
Additionally is it possible to visualize these mappings using the interactive visualization approach presented on this GitHub page? Running md.visualize() would visualize the spatial data before mappings, right?

[larsborm]

Hi, For Bonefight you indeed need to provide the number of cells in the same order. This parameter is the size and tells the program what proportion of each identity it can expect. That it also why the hexbins all get a size of one, because these are all the same size. (It might be more correct to count the number of nuclei in each tile though.)

The hexbin_plot() function accepts a number of different ways to color the tiles. Looking at your code md.hexbin_plot(c = df_eel.loc['SOX2'] ) would do it for the gene SOX2. You can change the colormap using the cm parameter if you want.

The .visualize() function is only meant for the raw RNA data. For all other plotting we rely on Matplotlib.

Hope this helps.