Making a cut off for ligand receptor interaction potential about nichenetr

[browaeysrobin]

Dear @saeedfc

The ligand-receptor interaction potential score is calculated as the weighted sum of ligand-receptor-signaling data sources that document the ligand-receptor interaction; with weights of each data source defined via parameter optimization. For more information about, see the Online Methods section of the paper (and this FAQ note: https://github.com/saeyslab/nichenetr/blob/master/vignettes/faq.md#if-i-look-at-the-ligand-receptor-network-i-see-some-low-interaction-scores-between-highly-expressed-ligand-receptor-interactions-or-some-high-scores-between-lowly-expressed-pairs-what-is-the-reason-for-this).

You could apply a cutoff on this LR interaction potential score, but I have found that you might then lose LR interactions that are still interesting. I would rather filter ligand-receptor pairs based on whether they are 'bona fide' or not, than based on the interaction potential score. But to narrow down the list of potential interactions, I would rather suggest checking expression of ligand and receptor. See also this FAQ section: https://github.com/saeyslab/nichenetr/blob/master/vignettes/faq.md#although-nichenet-already-prioritizes-ligand-receptor-pairs-strongly-i-still-find-there-are-too-many-possible-pairs-to-experimentally-validate-what-types-of-information-would-you-recommend-to-consider-for-even-further-prioritization

[saeedfc]

Thanks for the response! Shall follow your suggestion.