Change weights in favor of LR expression

[CDSchuster]

Hello!

I've been trying the tool for some time and it's wonderful. However, I noticed that the ligands and receptors that it returns as significant by prioritization score are not highly expressed in the correspondent experimental condition. Is it possible to change weights in such a way that high ligand and receptor expression is more relevant than the other criteria for prioritization score? If not, I was thinking of selecting via Seurat those genes that are highly variable and use only those for the ccc analysis. That way I can make sure that the ligands and receptors are highly differentially expressed in the correspondent experimental condition.
Thanks a lot in advance!

Claudio

[browaeysrobin]

Hi @CDSchuster

Is the problem mainly that your top interactions are not highly expressed, or that they are not very strongly differentially expressed?

If the issue is weak general expression:
Have you tried out tweaking the parameters related to expression of ligands/receptors (mainlyfraction_cutoff, less influential min_sample_prop)? In this file: https://github.com/saeyslab/multinichenetr/blob/main/parameter_interpretation.pdf, you can find some more information.

In any case:I strongly suggest to not use only Seurat's HVGs.

[CDSchuster]

Hi,

The problem is that there is no strong differential expression. When plotting violin plots for the receptors or ligand for a given celltype between experimental conditions, there is not a very noticeable difference.

I will try with the parameters that you mentioned

Thanks a lot!

[browaeysrobin]

Hi @CDSchuster

If that's the case, changing the parameters above won't help much. Within the top n predictions: are there some LR pairs with low ligand activity that do show stronger differential expression?

We may consider creating a new scenario where we increase the weight of the DE-criteria, so that DE has more influence on the final prioritization. But it could be more likely that the biological DE signal in your data is relativey weak is all of the prioritized LR pairs are now not strongly DE.