search

neurorestore / Augur

Cell type prioritization in single-cell data
MIT License
100 stars 11 forks source link

What fraction of cells per celltype is recommended to use as subsample_size? #26

Open kaizen89 opened 1 year ago

kaizen89 commented 1 year ago

Hi! Thanks a lot for this very useful tool. I succesfully ran Augur on a large dataset ~70K cells using default parameters, I want to increase the number of celles in subsample_size to 1000-2000 cells to better capture the heterogeneity. I was wandering if you can give any recommendation about the fraction of cells per celltype and per condition that you think would be reasonnable to use. Thank you!

skinnider commented 1 year ago

I essentially never change this parameter from its default values, but doing so should not impact your results very much. In Supplementary Fig. 6 of the Augur paper, we show that prioritizations are quite robust to the value of subsample_size - the main risk is removing cell types with too few cells from consideration.

kaizen89 commented 1 year ago

I guess you are referring to Supp Fig7 (not 6), I saw this part but unfortunately it seems that the dataset tested is a bit small and it's difficult ton conclude whit subsample_size 50-100. image