Show that cell type enrichments become purer as you move towards leaf nodes - Githubissues

neurogenomics / rare_disease_celltyping

Code, data and results associated with the "Rare diseases cell-typing" project.

6 stars 0 forks source link

Show that cell type enrichments become purer as you move towards leaf nodes #27

Closed NathanSkene closed 1 year ago

NathanSkene commented 2 years ago

Currently we hint at this but don't actually show it

Could you add to this row a plot with the distribution (violin plot) for number of associated celltypes at each ontology level?

NathanSkene commented 2 years ago

Would also be good to have a graph of:

Number of phenotypes/terms found at each ontology level

bobGSmith commented 2 years ago

phenos_per_level

bobGSmith commented 2 years ago

associated_cells_per_pheno_by_level_jitter

bobGSmith commented 2 years ago

associated_cells_per_pheno_by_level_violin

bobGSmith commented 2 years ago

@NathanSkene This kind of thing?

NathanSkene commented 2 years ago

Yep! Could you try merging the last two plots? E.g. violin + all points + smoothed line of best fit?

bobGSmith commented 2 years ago

associated_cells_per_pheno_by_level_jitter

@NathanSkene Can mess about with colors and stuff or color by another factor? I will add it to the figures_demo.Rmd script rather than try to put it into the report.Rmd for now.

bobGSmith commented 2 years ago

@NathanSkene added this to the dev_descartes branch figures_demo.Rmd

NathanSkene commented 2 years ago

That graph looks good. We'll need to remake the graphs above (in first comment here) so they are also violin plots with scatter behind

bobGSmith commented 2 years ago

ont_fold_plt

@NathanSkene in the original we used mean fold change by ont level and it gave that nice curve. but i guess seeing as its not normally distributed when i plot it like this the curve is much lower. This plot is showing all results with q < 0.05

bobGSmith commented 2 years ago

Actually, both are showing an average fold change of around 8 at ont level 0. Its just the outliers throwing off the scale in the scatter plot.

bobGSmith commented 2 years ago

ont_ngenes_plt

bobGSmith commented 2 years ago

ontlvl_facet