feat: DIAMOnD

[tristan-f-r]

Another disease module mining algorithm to complement #240 (#109). Closes #244.

Note that, in compliance with #246, DIAMOnD is MIT licensed.

[tristan-f-r]

Moving this back to a draft because I need to handle rank information.

[tristan-f-r]

This will be updated after #261 gets merged.

[annaritz]

New image (tagged as v1 and as latest for DIAMOnD now pushed to DockerHub: https://hub.docker.com/repository/docker/reedcompbio/diamond/general

Some minor comments/questions, but otherwise looks straightforward.

[annaritz]

Still a little unclear how seed weights are incorporated (should they be prizes, then?)

Also, do you know why tests are failing?

[tristan-f-r]

Oh! The diamond images need to be pushed with linux/amd64 instead of just linux/arm64.

I just went through DIAMoND, and I'll update the description to reflect this, but DIAMOnD does not take any weights per node/edge. However, DIAMOnD does assign an integer prize to all prizes, to change how much prizes affect the overall pathway construction. Specifically [in code], prizes are used to assign edge weights. alpha = 1 itself does nothing, but an alpha greater than one gives more attention to the seed nodes than other nodes after the first DIAMOnD reconstruction iteration.

DIAMOnD can actually be modified (in code, not during wrapping) to specifically take in prizes per node.

[annaritz]

Good to know that DIAMOnD could be modified (but let's not do that now, that would be a different alg).

I pushed the linux/amd64 versions with latest and v1 tags. That means that the linux/arm64 versions no longer have those tags (should that matter?).

https://hub.docker.com/repository/docker/reedcompbio/diamond/tags

[tristan-f-r]

There is an actual CI failure! I am not sure exactly, but I believe DIAMOnD fails reconstruction if the provided n is one less than the network size. The last two commits address this and wrap the error with a nice display message wrapping the original opaque error.

[annaritz]

Ok, after thinking about the alpha parameter I get it now (it's something passed to DIAMOnD, not something that requires input file modification). That's where my confusion was - the code and the comments are all correct.

All tests pass with the linux/amd64 images! I'm approving for merging.