Inconsistence between N_cells_per_location and the deconvolution result

[thestarocean]

Thank you for providing such an amazing tool for spatial transcriptome analysis. However, when running it, I am somewhat confused by the inconsistence between N_cells_per_location and the deconvolution result.

I was running an analysis on sample collected from mouse liver. I gave the hyperparameter as N_cells_per_location equal to 10, but the deconvolution results showed that some spots had as many as ~20 hepatocytes.

Is this result normal? Or should I adjust the parameters somehow?

[vitkl]

This is normal and depends on the amount of technical variability in RNA detection across the slides. You can increase detection_alpha or use segmented nuclei counts with our new experimental branch #344 if getting absolute values is very important for project goals (such as comparing many samples).