search

YMa-lab / CARD

GNU General Public License v3.0
80 stars 20 forks source link

Multiple slices #74

Open GouQiao opened 2 months ago

GouQiao commented 2 months ago

Hi,

I have multiple slices from 6 patients. For ST data, I divided them into 3 groups (2 sample for each group). And I also integrate scRNA-seq by groups. For CARD deconvolution, can we process multiple slices and integrated scRNAseq data together? If not, which way is better? 1. Use ST integration package such as PRECAST to integrate 2 slices for each group, and split scRNAseq by groups as well. 2. Just split ST and scRNAseq data into 6 samples, and do CARD analysis individually.

YMalab commented 2 months ago

Hi @GouQiao,

Thank you for your interest in our package!

Currently, CARD does not process multiple slices as the spatial structure is built by each slice, and the spatial correlation structure depends on the coordinates. For scRNA-seq, you can use integrated scRNA-seq reference data (which must be non-negative) as this will not affect the spatial correlation structure.

Regarding your proposed solutions:

  1. I haven't tried PRECAST before, but PRECAST models are based on centered normalized expression, which is not necessarily non-negative. Therefore, if the gene expression data with batch effects removed that PRECAST constructed is not non-negative, CARD will not be able to run.
  2. You can perform CARD analysis individually, and the cell type proportions will provide information about both homogeneity and heterogeneity across slices and patients.

I hope this helps.

Best, Ying