new curation supporting tests

[Egon Willighagen]

Hi Denise, all,

I have added some tests on Jenkins [0] to report non-metabolites converted into or from metabolites... like a gene->metabolite or metabolite->pathway...

There are some false positives, like metabolites created as breakdown product from proteins... I have to check if a mim-conversion is still the proper annotation for it, but would not surprise me it is... I'm going through the reports but I need to convert the biological of them one by one, so this may take a bit... all, as always, feel free to join in ;)

I already fixed a few ligand-target bindings which where incorrectly annotated with mim-conversion.

Egon

0.https://jenkins.bigcat.unimaas.nl/job/WikiPathways%20Curation%20(WP2RDF%20Curation)/lastCompletedBuild/testReport/nl.unimaas.bigcat.wikipathways.curator/

--

E.L. Willighagen
Department of Bioinformatics - BiGCaT
Maastricht University (http://www.bigcat.unimaas.nl/)
Homepage: http://egonw.github.com/
LinkedIn: http://se.linkedin.com/in/egonw
Blog: http://chem-bla-ics.blogspot.com/
PubList: https://www.zotero.org/egonw
ORCID: 0000-0001-7542-0286
ImpactStory: https://impactstory.org/u/egonwillighagen

[Chris Evelo]

This kind of discussions always makes me wonder what the best way to treat insulin (or snake toxins maybe would be... These are both metabolites (if you look at what kind of function they perform) and gene products.

[Egon Willighagen]

On Tue, Feb 13, 2018 at 8:47 AM, Chris Evelo <chris...@gmail.com> wrote:

This kind of discussions always makes me wonder what the best way to treat insulin (or snake toxins maybe would be... These are both metabolites (if you look at what kind of function they perform) and gene products.

Indeed!

WikiPathways, historically, does not have a biological definition of "Metabolite" but an operational one: do identifiers come from gene/protein databases or small compound databases. Now, insulin is one of those examples where it ends up as a "small" compound and you can find insulin ChEBI IDs. But it has also protein identifiers in UniProt.

This actually makes these tests interesting, as we have the opportunity to model this properly... I have to see if these should be explicit exceptions, or maybe this is something too complex to test with unit tests...

But a query to list all such (true) protein->metabolite mim:conversion's would be nice :)

BTW, I consider Ac -> Protein-Ac (could also be considered a mim:conversion) far outside the scope of this discussion...

Egon

[Egon Willighagen]

Hi all,

a quick update

On Tue, Feb 13, 2018 at 8:57 AM, Egon Willighagen <egon.wil...@gmail.com> wrote:

On Tue, Feb 13, 2018 at 8:47 AM, Chris Evelo <chris...@gmail.com> wrote:

This kind of discussions always makes me wonder what the best way to treat insulin (or snake toxins maybe would be... These are both metabolites (if you look at what kind of function they perform) and gene products.

Indeed!

I still have to check if MIM puts restrictions on the entities being linked (generally, yes, but does it put restrictions on the types of allows source and target for specific interactions?).

There are now two exceptions which were false positives: generation of insulin and bradykinin.

I plan to make SPARQL queries to find such biology (and hope I don't forget, as this is not really top priority, and people are welcome to take the unit test SPARQL and update it for such, and make it available as example on rdf.wikipathways org ;).

There are currently four failing unit tests left... a few are in pathways for which I have recommended an updated homology conversion... (Tina, let me know how you wish to attack that; some of them have minor edits which we may need to check with the original...)

Egon