search

Teichlab / cellphonedb

MIT License
339 stars 105 forks source link

Missing interactions #251

Open Adina-ballaban opened 3 years ago

Adina-ballaban commented 3 years ago

Hi,

I have a dataset which I have split into multiple sub-groupings of samples to run CellphoneDB on. When I run it on each subset, for 2 of the 3 subsets I get the same number of overall interactions in the means.txt output, but for one of the subsets I get less interactions. Some of these interactions that are missing from the one subset are particularly interesting to us and initially I thought the problem was with the threshold because the genes in these interactions are expressed on a lower number of cells in this subset, but after trying this it did not make a difference. Additionally, I have checked and, although some of the clusters in this subset have a very small number of cells, the percentage of cells expressing the genes for the missing interactions are actually quite high in some clusters. Is there a reason why these interactions are not being included in one subset from this dataset but are included in runs for the other 2 subsets?

Thanks

luzgaral commented 3 years ago

Can you provide an example that we can test?

Note that, if the interaction involves a complex, all the members of the complex are required to be present. Otherwise, the interaction is excluded.