Description
I was wondering if my understanding of intraviews, juxtaviews, and paraviews work for the specific questions I wanted to answer in my single-cell spatial transcriptomics data. I have been playing around with using different juxta and paraviews, including using the bypass.intra option, and changing up the family methods between 'constant' and 'gaussian', and it has been difficult to see consistent results.
With my dataset (CosMx), every cell is segmented already, so from my understanding intraview should not be important (bypass.intra = TRUE) when considering cell-cell interactions. My samples are largely composed of malignant cells.
-
If I wanted to understand how different cell types in the proximity of my malignant cells contribute to the activation of different pathways and transcription factors, I would create an initial view based on my different cell compositions. Then I would create views according to results from decoupleR for pathways and TFs and set up paraviews for them using the nearest 30 cells (family = 'constant', l = 30). I would do this for each sample and then collect all the results downstream. If I don't see strong signals in any pathways, I would conclude that pathways that I observed to be enriched in a given cell type are not being activated by nearby cells?
-
If I want to figure out if there are any spatial patterns between cells in terms of ligand-receptor expression, I would set up an initial view with a paraview (l = 30, family = 'constant'). Then I would set up additional paraviews (one for ligands, one for receptors, also l = 30, family = 'constant'), and then run mistyR (bypass.intra) using these views?
Thanks for your help in advance!