Velocity gives different output according to tsne and flag "scale"

[vkorobeynyk]

Dear community,

I have an issue which I cant understand. I load my data into pagoda2 where I do my clustering and everything and then I try to run RNA velocity.

I realize that with different tsne space and the flag "scale=" the RNA velocity change drastically. Even the cells that were going in one direction in one tsne started to go in another direction. How is it possible? Maybe I am doing something wrong? What does the flag does? and Why the dim.reduction changes so much the outcome of RNA velocity?.

2. Also what If I have a tsne already made and I dont want to run pagoda2? Can I just manually change the coordinates or is there a proper way?

Thanks in advance for the answers Vlad

[grimwoo]

i encounter the same problem. the direction is "correct" according to our current research in 2 samples and nearly "correct" in the other 2 samples. however, when integrated together, it is not that good. Hence, I wonder whether could we change part of arrows in UMAP?