`ecodive`: A new, dependency-free package for diversity metrics (from `rbiom` dev)

[dansmith01]

Hello mia developers,

I'm the developer of rbiom and wanted to let you know that all alpha and beta diversity functions have been moved from rbiom into a new, dedicated package: ecodive (https://github.com/cmmr/ecodive). This new package offers a more comprehensive collection of metrics, is highly optimized for speed and memory, and has zero R dependencies.

For now, rbiom will call ecodive behind the scenes, so your existing code will continue to work. However, we recommend transitioning to ecodive directly to ensure future compatibility, as some rbiom function parameters may be deprecated soon.

I would be happy to help with this transition and can submit a pull request to port the code if you'd like.

Thanks for your consideration.

Best,
Dan

[TuomasBorman]

Hello @dansmith01, thanks for letting us know!

I guess the following issue, would be fixed by using ecodive: #756

@jepasan Can you check?

-Tuomas

[TuomasBorman]

ecodive could be also cited in OMA online book

[jepasan]

Seems like it should be simple enough to work into our code, I'll look into it as soon as I can.

[dansmith01]

Heads-up - ecodive v2.1.0 is now on CRAN.

I recommend using this version for the integration work. Please be aware of one breaking change: the rescale parameter is replaced by norm. Setting norm = 'binary' will produce unweighted distance matrices, equivalent to rbiom's weighted = FALSE. Apologies for the timing on this — it should be the last breaking change for a while.

[antagomir]

ecodive could be also cited in OMA online book

Yes, and in general it might be a good idea to use this as default compared to corresponding vegan functions, at least based on the speed & memory benchmarks: https://cmmr.github.io/ecodive/articles/benchmark.html

-> some work needed, worth an open issue?

[jepasan]

I created a pull request for replacing the rbiom unifrac code. It was simple enough that I don't think replacing the vegan functions would be too difficult either, it'll just take more work because vegan is used in so many more places.
I also tested their Faith's alpha diversity function, and that seems to produce the same results as our implementation too, at comparable speeds, so that's another thing that could potentially be replaced.

[antagomir]

@jepasan now as the Unifrac has been done, would you have a chance to look at the other dissimilarity indices as well?

We would like to use the most fast & robust implementations by default I guess. When improved alternatives for vegan exist, we could use those.

It might still be useful to support access to vegan as well because it is so widely used. But then users could specify they want to use it, index names could indicate this, e.g. vegan::bray or bray_vegan?

Accelerated Faith is currently from our own (co-authored) publication. Unless there are really improved methods I would not see reason to do extra work to update our already implemented method to another with similar performance.

[antagomir]

@0xMuluh can you check that ecodive has been cited in OMA online book. Shall we also cite/mention in the manuscript (if not yet)

[jepasan]

@antagomir Sure, I'll have a look at them.