Scanorama embeddings on CellRank analysis

I’m using scVelo and CellRank for pseudotime and trajectory analysis. I have a question about using data integrated with Scanorama. Is it correct to calculate the neighbors based on scanorama embeddings rather than pca or umap for scVelo analysis? Can this then be used to proceed with CellRank?
Does anyone have any input, suggestions or feedback on this?