Add functions for integrating CITE-seq data

[HelloWorldLTY]

Dear Authors,

Hi all, I have added functions to implement the functions of integrating gene expression information and protein information from CITE-seq data. I have offered two options to model protein data, either using 1. NBmixture or 2. Normal Distirbution after preprocessing with CLR. It seems that the later option works better. I also provide a method to create the gudiance graph between genes and proteins.

Sincerely,
Tianyu

[Jeff1995]

Hi @HelloWorldLTY, thank you for the contribution!

I have reviewed the changes and understood the new functionalities and approaches. Below are some further changes that need to be addressed before we can merge.

1. The unit tests are failing because of a new argument init_fea_emb in scglue.models.sc.GraphEncoder.__init__(). It needs a default value of None, otherwise the standard GLUE model would stop working. Meanwhile, I suppose the forward function does not need this argument?
2. The scglue.utils.generate_prot_guidance_graph() function should be moved to the scglue.genomics module, just like other guidance graph functions.
3. The scglue.utils.clr() function is a duplicate of muon.prot.pp.clr, so I'd suggest removing this and use muon directly.
4. The NBMixtureDataDecoder returns one of the NB components randomly per Bernoulli sample rather than a genuine NB mixture. While this might also work in practice, a cleaner approach would require defining an NB mixture distribution and implement a mixture log_prob. I suppose totalVI did implement one, but we may also achieve the same goal using torch.distributions.MixtureSameFamily and NegativeBinomial.
5. The new functions need type hints and documentations in a style consistent with the current code base.

Please let me know if you'd like me to address any of these problems or if I've misunderstood some of the code. I would be glad to help!

[HelloWorldLTY]

Hi, sorry for my late reply. I have made the modification based on your comments. For NBMixture model, I used categorical and nb distribution from torch to implement. Please let me know if you have other questions.

[HelloWorldLTY]

It seems like a random issue. Schneegans/dynamic-badges-action#36

[Jeff1995]

Thx! This was an awkward issue due to a custom coverage badge that has since been broken. I've now switched to codecov and it should be working fine.

It seems like a random issue. Schneegans/dynamic-badges-action#36

[codecov]

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

Thanks for integrating Codecov - We've got you covered ☂️

[HelloWorldLTY]

Thank you Zhijie!