Calculate gene signature scores from cell2location-deconvolved spatial transcriptomics

Help spatialdata
1

Hi everyone,

I’m working with Visium spatial transcriptomics data and have performed deconvolution using cell2location. Now I want to calculate a gene signature score (e.g., an acidosis signature) and I’m unsure about the best approach given that cell2location stores deconvolved expression estimates for each cell type in separate layers.

My setup:

  • Spatial transcriptomics data (adata_spatial)

  • Cell2location deconvolution completed

  • Deconvolved expression for each cell type stored in adata_spatial.layers[‘Cell_type_name’]

  • Different cell type layers have very different expression scales:

  1. layers[‘Ductal cell type 2’]: min=0, max=37.65

  2. layers[‘Malignant cell’]: min=0, max=228.12

My question:

What’s the best practice for calculating gene signature scores from cell2location deconvolved data?

Option 1: Global score using total expression: Sum expression across all cell type layers and calculate the score globally normalizing expression
Option 2: Cell-type-specific scores: Calculate separate scores for each cell type layer. Not sure how to do this or if results are comparable.
Option 3: Weighted by cell abundance. Weight each cell type’s contribution by its abundance from cell2location.

Thank you!

Victor