Identifying bonafide ligand-receptor pairs in the updated ligand-receptor database (V2)

[alecorr]

Hi,

Thanks for providing such a useful tool!

I was wondering if there is a way to filter the NicheNet-v2 LR network to remove ligand-receptor pairs inferred through PPI databases.

In my analyses, I have noticed that some of the top predicted interactions comprise a few spurious ligand-receptor pairs (for example, CSF1-CSF3R), and as a result I have been looking for ways to filter the LR network to remove low-confidence LR pairs. In previous versions of the package, there used to be an option to filter to 'bonafide' ligand-receptor pairs by removing pairs annotated with the 'ppi_prediction' and 'ppi_prediction_go' values in the 'database' column of the ligand-receptor network - but as far as I can see, this functionality has been removed since the upgrade of the ligand-receptor database + prior model in nichenet V2.

Whilst it's possible to manually remove problematic LR pairs one by one, I was wondering if there was a more systematic approach I could take. I have tried following the script/functions used to generate the NicheNet-v2 LR network (provided on this zenodo page), but I have been unable to find a way to identify LR pairs inferred through PPI resources using the tools provided by omnipath (although I'm not very experienced with omnipath so maybe I'm missing something obvious).

Thanks!
Alex