Representing AGAC in VSM?

[stcruy]

–> Question – (Broader than NLP – On how to encode knowledge in VSM):

I was wondering how an AGAC-annotated sentence would be shown via VSM?

–> Answer:

I took a couple of AGAC examples from this paper: doi:10.18653/v1/D19-5710. I hope these are relevant?

► The paper's Figure 3 illustrates "two protein-truncating DNMs in SHROOM3",
which means: "two de novo mutations that truncate the protein, in human SHROOM3".

If you follow the link below, you go to the ‘VSM demo’ page and it will automatically show the VSM-sentence that is encoded in the URL.
This is still a quite literally translated VSM-sentence. I made it manually. I looked up valid identifiers for each term, including URIs from CHEBI, UniProt, and WordNet. (I used WordNet for "truncating" (the verb "truncate"), and "in" (it means the same as the verb "pertain to")).
See: example 1.

I think this would actually be simpler to say like this: "two DNMs that truncate SHROOM3". Because SHROOM3 has a UniProt-ID, so the extra term "protein" was redundant.
In VSM: example 2.

► Figure 4 describes something that a human curator would enter, if I understand correctly?
It says: "25805808, SHROOM3, LOF, Neural tube defects".
This could correspond to: example 3.

Or perhaps better as in example 4, which places the modified protein as the subject instead.

► I also made a VSM-template version from that last VSM-sentence. Curators could use this, after linking the interface to the relevant dictionary modules that provide autocomplete-data: example 5.

For more about VSM-templates: see the causalBuilder web-app where you can click some checkboxes to generate different VSM structures.