# min_shared_counts cutoff

[TarekGharsalli]

Hello,
I'm new user of scvelo
Does any one has an idea how to choose min_shared_counts when we run scv.pp.filter_and_normalize.
I tried many values 20, 5, 1 and each times I get different direction of the arrows.
My second question : It's mandatory to filter genes with low count ?
Thanks

[WeilerP]

@TarekGharsalli, as far as I am aware, there is no recipe on how to choose the value. I would also not focus too much on the direction of arrows but rather how the phase portraits look like, i.e. if they include any info that can be leveraged for RNA velocity analysis.

Regarding you second question: Yes, you should filter genes by low counts and only keep highly variable ones for the following RNA velocity analysis. This is a standard step of preprocessing single-cell datasets.

[TarekGharsalli]

Thanks @WeilerP :)