How to use splice/unsplice proportion and phase plot to judge the applicability of scVelo to a dataset

[gundalav]

Hi,

I have the following data set that gives this proportion:

The phase plot looks like this:

Currently the final velocity embedding stream looks like this:

My question is based on the near 50/50 splice/unspliced proportion and phase plot.
Can we deem whether or not scVelo can be used in our dataset?

[WeilerP]

There is no clear football shape (or part thereof) in your phase portraits, i.e. the fundamental assumption of parameter inference with scVelo is violated. This has already been extensively discussed in numerous other issues and discussions. See e.g. #216, #462, #472.

[gundalav]

@WeilerP Thanks for your reply. Can you give example of the ideal "football/almond" shape?

[WeilerP]

You can see them in the paper or it's supplement, for example.