Opposing directionality of predicted ligand vs. predicted targets about nichenetr HOT 1 CLOSED

Hi @ajwilk

The ligand prioritization is currently done without taking into account to expression value of the ligand, the only condition is that the ligand is expressed. This means that the algorithm does not consider whether a ligand is more strongly expressed in the case vs control, as long as it is expressed at a sufficient level. Of course, this might be interesting information to use for further prioritization after the NicheNet ligand activity analysis.

Changing geneset = "up" will make that the possible target gene set of interest will only include genes that are upregulated in the case vs control, and not all DE genes as is default. So this has nothing to do with the ligands you select for the analysis.

Now to explain your observation, you could consider following hypothesis. Note that some ligands might downregulate many genes (instead of, or in addition to, upregulate many genes). This means that if TGFB1 is active in the control condition, it can downregulate genes there. When TGFB1 itself is repressed in the case condition, and has a lower expression, its downregulatory effect might disappear, leading to upregulation of genes that are a downregulated target gene of TGFB1.

from nichenetr.