snRNAseq subclass annotation of thalamus

Technical
1

Dear community,

I’m a beginner in snRNA-seq analysis and currently working on cell type annotation for mouse thalamus datasets. While exploring the “Cell Type Knowledge Explorer” module, I noticed it primarily contains subclass annotations for the primary motor cortex (MOp).

Could anyone share insights on whether thalamic neuronal subtypes follow similar classification schemes as cortical/hippocampal regions? Specifically:

    Are excitatory neurons still categorized by layer-specific markers (e.g., L6 CT)?

    Do inhibitory neurons maintain the canonical cortical subtypes (e.g., Pvalb+/Sst+/Vip+/Lamp5+ populations)?

Alternatively, does the thalamus require distinct classification frameworks given its unique nuclear organization and glutamatergic/GABAergic neuron distribution? Any references to thalamus-specific marker genes or published annotation pipelines would be immensely helpful.

Thank you in advance for your expertise!

2

I checked with Zizhen Yao (first author the Allen Institute’s most recent whole mouse brain paper https://www.nature.com/articles/s41586-023-06812-z and several others), and here was her response: “There is no thalamus specific taxonomy, but it is pretty clear where cell types in thalamus are present in the whole brain atlas. The TH class include most of the Glut cells, and GABA cells are grouped together with HY cell types. The users can use MapMyCells to map their dataset to the whole brain atlas and identify corresponding cell types. You don’t need TH specific taxonomy to map TH cells.”

In theory, you could subset the whole mouse brain taxonomy to TH-specific types, that’s probably not the best path forward in this case. If you want to do it anyway, data for this taxonomy is accessible here.

There are two easier ways for mapping your data to these cell types:

  1. Use the MapMyCells GUI: The GUI and instructions can be accessed via the MapMyCells website, but essentially you just need to save your cell by gene matrix as a csv or h5ad, upload it to the GUI, and download your results. This is the approach I would start with and in most cases works extremely well with reasonable quality data.

  2. Download the data and run cell_type_mapper: This can be done by following this python Jupyter Notebook. This is the approach I would take if # 1 gives you mapping results that don’t make sense (e.g., lots of thalamus cells mapping to cell types out of thalamus). The code-based approach allows you to restrict the cell types you map against, meaning you could create a sub-taxonomy that only includes the thalamus types you expect to find, if needed.

Note that both approaches use the same underlying mapping algorithm, so you’ll get the same answer with both approaches if you map against the full taxonomy (+/- differences in random number generators with subsampling, that will have some minor impact).

3

Dear Jeremy,

Your expertise made a real difference, and I’m grateful for your time and effort. Looking forward to applying your advice!

Best regards,

Sivan