search

navinlabcode / CellTrek

108 stars 19 forks source link

selection of Edge cut-off threshold for spatial colocalisation of cells #26

Open pratapstat opened 1 year ago

pratapstat commented 1 year ago

Hi CellTrek team,

Thanks for this tool which is really useful to test the hypothesis on my spatial data. I would like to know, how to decide which cut-off makes sense for me to look at the spatial colocalization of cells. Please elaborate on this?

For instance, if cell A is interacting with cell B at edge cut-off 0 but not at edge cut-off 0.2? then how to interpret that?

Best Regards, Pratap Senior Bioinformatics Specialist, NCCS, Singapore

Puriney commented 1 year ago

To me, the cutoff parameter is to help users to get a comfortable level of trimming noise vs keeping results. As long as the edge width is not zero, it implied a co-localization. But then it depends on how much noise/results the users want to show.

Puriney commented 1 year ago

Those 'thin' edge connections are probably what you are most interested in. I think you probably need some orthogonal evidence. Ligand-receptor inference analysis is a very good choice.