Python adaptation - cross advertising?

[Zethson]

Hi,

we will finish a Python implementation of Augur for the Python & Scanpy ecosystem in January. I was wondering whether you would be up to cross advertising? We will of course link to the original R implementation and you could link to the Python implementation in your README or somewhere else.

Details from my side would follow.

[skinnider]

Hi @Zethson, very cool to hear, we would definitely be happy to point python users to your implementation if you can confirm the two implementations produce concordant results. There are some pre-processed datasets in our recent Nature Protocols paper and associated Zenodo repository that might be useful for this purpose.

[Zethson]

Glad to hear! CC @antschum.

I'll keep you posted.

[Zethson]

All right, it is done.

https://pertpy.readthedocs.io/en/latest/tutorials/notebooks/augurpy.html

https://pertpy.readthedocs.io/en/latest/usage/usage.html

Augurpy allows for two feature selection methods, the select_variance method that you are using as well and another one based on Scanpy's highly_variable. The results can never be perfectly the same due to different random seeds, but I think that we're close enough.

We won't able to make it any closer, but I think that the select_variance implementation of Augurpy is close enough to the actual Augur. What do you think?

[Zethson]

@skinnider do you have a comment?

[skinnider]

Hi @Zethson, so sorry, somehow we must have missed this! Very cool to see the implementation is finished. I will add a link to your implementation in the README now.

[Zethson]

Thank you!