Using Integrated single cell data for alignment

[abhishekmaj08]

Greetings,

I have a query about the alignment step. I have to integrate 2 single cell datasets and then use it in Tangram for the alignment portion. I use Seurat for the single cell integration and then convert the integrated Seurat object to H5AD format. Seurat creates 2 assays: 1) integrated and 2) RNA. Should I be using the integrated assay or the RNA assay for alignment of spatial data with the integrated dataset?

I tried with the integrated assay, and ran the pipeline. In the step where I print out " ad_map.uns['train_genes_df'] " I get some negative training score values. However when I manually scaled the integrated assay to [0-1] range, and ran the same pipeline, I do not get the negative training scores. I assumed that the integrated data assay contain negative values (because of the integration process) which was why I was getting negative training values down the pipeline. That is why I scaled it to positive values (0 to 1) as gene expression should be positive. I am wondering if my line of reasoning is correct or not? And also should I be using integrated assay in the first place or RNA assay in the alignment step?

Thanks