Interest in building an R wrapper for the AlphaGenome API

Community
1

Hi AlphaGenome team,

I’ve been following the project and wanted to ask if you’d be open to collaborating on a small R package that wraps the AlphaGenome
API for R users in genomics.

My idea is to keep it lightweight and closely aligned with your existing API/docs, mainly to make common tasks easier for people who
work mostly in R.

If this sounds interesting, I’d be happy to share a rough outline and keep it consistent with whatever conventions you prefer.

Best,
Himanshu

2

Category: Systems-Level Biological Intelligence (SLBI

3

Hi @Himanshu_Bhandary, welcome to the forum!

Yes I’m sure there would be interest in being able to use AlphaGenome from R! I’d suggest taking a look at the underlying API protocol and then writing a C/C++ module that utilizes gRPC to call the backend.

FYI, unfortunately we won’t be able to upstream this to our repo, as we don’t have the bandwidth to maintain additional implementations.

Good luck!

1 Like
4

Hi tward,

Thank you for the encouragement and the technical direction. I’ll get started using the existing repo and documentation.

Since the current SDK is Python-only, an R interface could help reach a large segment of genomics researchers who primarily work in R. I’ll plan to build and maintain this independently.

Once I have a working version, would the AlphaGenome team be open to sharing it with the community (e.g., a forum post or social mention)? I think it could be useful to others in the ecosystem.

Happy to keep you posted on progress.

Best,
Himanshu

5

I had a couple of quick questions to make sure I design this in a way that stays robust over time:

  • Are there any anticipated breaking changes to the API or protobuf definitions that I should be aware of?

  • Are there particular workflows or use cases you’d especially like to see supported in an R interface?

This will help me prioritize the initial design and keep it aligned with real usage.

6
  • There are no planned breaking changes to the RPC protocol, no.
  • I’m not aware of any specific R workflows… I’d probably prioritize variant scoring as this is the most useful from an analysis POV, tho plotting tracks can be useful to get understanding of a particular score.

We can definitely post an announcement on this forum, the Announcements category has some other community developed packages (e.g. PyTorch fine-tuning).