satijalab/azimuth-references

Understanding reference building approach

LuckyMD opened this issue · 0 comments

LuckyMD commented

Hi!

Really cool tool and web portal! I was trying to understand how you have been building references for Azimuth. A lot of what you did makes sense to me, but I stumbled across a line in the human BMMC reference script indicating an additional normalization step after individual batches had already been scTransformed.

See:

azimuth-references/human_bonemarrow/scripts/integrate.R

Line 58 in 536fcbb

integrated <- NormalizeData(integrated)

I was wondering why the second normalization step is needed here?

Thanks a lot!