njbernstein

@Yenaled why is that?

Use `loguru`

@davek44 seems like we can close this out

Thoughts?

I've got some code to do this. I'll try to submit a PR soon

I'm happy to implement my method, when we have a consensus. https://github.com/calico/solo/blob/master/solo/hashsolo.py https://www.biorxiv.org/content/10.1101/841981v1

@aditisk I'm the author of this method https://github.com/calico/solo. it should install relatively easily if you have any issues I'm happy to help. The main functionality it doesn't have is `tag_groups`...

I'd like to tackle this. Can someone tell me how we want to store hashing data in an `anndata` object? @flying-sheep @fidelram I'll take back up porting hashsolo to scanpy

@wflynny `hashsolo` allows you to set a prior for your expected rate negatives, singlets, and doublets, which helps quite a bit with the issue you described despite modeling the log...

In supplementary figure 9 of our paper, I did a light comparison of tools using the demuxlet data as ground truth: https://www.cell.com/cms/10.1016/j.cels.2020.05.010/attachment/040c239d-1e70-42a4-8974-9fbd75c65551/mmc1.pdf Which I think is a fine first stab...