Drug discovery and chemical representation

[agitter]

Work in progress. Ref #689.

I plan to add a few new references on graph convolutions (without much discussion) to complement our new section on PPI graphs in #638. This also adds a reference to partially address the bioRxiv comment per #654.

[agitter]

@dhimmel ef2b76c passed the build yesterday but failed today when I reran it to try to diagnose my more recent build failures.

Have you seen this error before:

Traceback (most recent call last):
  File "/home/travis/miniconda/envs/manubot/bin/pandoc-fignos", line 11, in <module>
    sys.exit(main())
  File "/home/travis/miniconda/envs/manubot/lib/python3.6/site-packages/pandoc_fignos.py", line 415, in main
    plusname, starname, 'figure')
TypeError: replace_refs_factory() missing 1 required positional argument: 'target'
Error running filter pandoc-fignos:
Filter returned error status 1

[agitter]

This may need some additional work but is ready for feedback and review. I didn't go into detail about the new graph methods because the reviewer already thinks this section is too technical. I primarily sought to separate the representation learning section to break up the long ligand-based subsection and emphasize the graph-based methods. These were requested by the same reviewer and are interesting algorithms beyond the virtual screening domain.

I could cut some text if we'd like to shorten this.

This PR adds #555, #641, #741, #762, #773, and #776.

[dhimmel]

Nothing stands out to me as problematic in this PR's diff.

@tomduck (creator of pandoc-fignos), any idea what this error could be caused by:

TypeError: replace_refs_factory() missing 1 required positional argument: 'target'
Error running filter pandoc-fignos:

We recently upgraded pandoc-fignos, pandoc-tablenos, and pandoc-egnos to v1.0 in 6c2020f. @agitter do you know whether the successful builds used the older version of pandoc-fignos?

[tomduck]

The API for pandoc-xnos was slightly changed. You need to update it as well.

[tomduck]

Actually, it looks like you may be using the new pandoc-xnos but an old pandoc-fignos. Can you upgrade both to the most recent version?

Note that I am in the process of preparing another release, which will go out within the next hour.

[dhimmel]

Looks like our build used pandoc-eqnos==1.0.0, pandoc-fignos==1.0.0, pandoc-tablenos==1.0.0, and pandoc-xnos==0.15.

Actually, it looks like you may be using the new pandoc-xnos but an old pandoc-fignos. Can you upgrade both to the most recent version?

Will do soon, after the next release is out. I think the issue may be that we pinned fignos, egnos, and tablenos, but not xnos. Therefore, the error would have first occurred when a higher version pandoc-xnos started getting installed.

[tomduck]

Yes, that makes sense @dhimmel. Is there something I should be doing on my end to prevent such a problem in the future?

[dhimmel]

As per here, I think you could specify something like pandoc-xnos>=0.13,<0.15, in setup.py of pandoc-fignos. However, I'm not sure how you would know, when making pandoc-fignos==1.0 that it would stop working at pandoc-xnos==0.15, which presumably hadn't been released yet.

[tomduck]

Yes, that is exactly the problem. I will think on it and come up with a solution.

[cgreene]

This looks good to me - just one suggested change. Marking approve so that you feel free to merge after that minor edit.

[cgreene]

I think you need to either remove The or add networks or models.

[agitter]

do you know whether the successful builds used the older version of pandoc-fignos

@dhimmel I didn't explicitly change anything in our environment between when the build passed and failed. However, our last successful build on master did use pandoc-xnos version 0.14.

It looks like we have a root cause now and can merge this once we pull in manubot/rootstock#99 Thanks for the help @tomduck.

[dhimmel]

@agitter, I just merged #778 so you should be good to go.