Danila

The described use case seems to work for MuData objects. If there's a particular scenario that still results in an error, please follow up with some steps to reproduce it!

Hey @emdann, This stems from the necessity of `.update()` — and the fact that by default, the columns are copied from individual modalities. We might change this behaviour in `v0.3`...

This should be fixed by the new API in v0.3 (`.update(pull=False)`), which will become the default one in the next versions.

For a start, from v0.3 (#56) it will adopt the `src` layout!

Now updated to support the latest pandas!

MOFA specifically and muon in general expect multimodal data. Here mudata is a more flexible container, and it is one of the design challenges how to fail gracefully... For the...

Sounds like those are different modalities so the default `axis=0` should be more appropriate to apply multimodal methods. To note here, most multimodal methods won't account for feature relationships, i.e....

Thanks both! On the consistency: there was a recent discussion on trying to be more aligned with the version support across packages. Broadly speaking, [the `SPEC 0`](https://scientific-python.org/specs/spec-0000/) seems to be...

Let me know if that works, @jflusche, we have tried to [improve compatibility across versions](https://github.com/scverse/mudata/blob/main/src/mudata/_core/compat.py) in the latest releases.

Just stumbled upon this issue, and as it's pretty recent, wanted to give a pointer to yet another experimental approach to only load parts of the data — https://github.com/scverse/shadows. Please...