DestVI model on multiple visium slides

I have been following the tutorial on deconvoluting the visium data here: Multi-resolution deconvolution of spatial transcriptomics — scvi-tools .

The tutorial only fits the model to one slide. Is it the best approach to process each slide separately, or is it ok to merge multiple visium slides into one anndata object?

Thank you!

Thanks for trying out destvi! Actually, the model here is fitted on two merged slides. We had two lymph nodes on each slide, and I merged them. However, one lymph node out of four has low QC metrics, so I took it out.

Merging should help learning a better model (especially for the cell-type specific signal), so I highly advise doing that!

Let me know how it goes, and whether I can help with interpreting the results or other.


How did you merge them? Just by concatenating the anndata objects?

Am I correct in thinking that DestVI does not utilise the spatial coordinates of the spots in the model? If iI am wrong, and it does use the coordinates then how did you account for that in the merging?

Best wishes