Additional metadata in results dataframe

[dnjst]

Hi, I would really like a way to print the percent expression and CPM values for the (minimally expressed) ligand and receptor in the results dataframe for li.mt.rank_aggregate() or any of the individual methods (which I know also have the supp_cols argument - maybe it is possible there? but rank_aggregate() seems to lose supp_cols)

The ones of interest would be something like:

ligand_percent
receptor_percent
ligand_cpm
receptor_cpm

It'd improve the interpretability of the results a lot - for instance, when return_all_lrs=True and an interaction fails the tests, it gives insight into why it fails, because of the ligand or receptor. Also, it allows for simple TPM threshold-based interaction calling - the most rudimentary method discussed by Armingol et al. 2021 (https://doi.org/10.1038/s41576-020-00292-x), but still a useful one. I don't know if you think that CPM thresholding would work as a "method" to liana, but either way least returning this info in the dataframe would be nice.

Thanks for any input!

[dbdimitrov]

Hey @dnjst,

Indeed, in a past version I decided not to include supp columns in rank_aggregate due to (long-story-short) the way that I used to deal with complexes. I can have a go at implementing supp cols for it in next versions.

In the meantime, you should be able to get those columns for the individual methods. Percentage if understood correctly should be the *_prop columns, while CPM is just *_means if your data is CPM-normalized.

[dbdimitrov]

Thresholding is already done implicitly via expr_prop but I agree having the option to include these columns would make it more interpretable 🙂