Open pgaudet opened 7 months ago
balhoff
commented
6 months ago @pgaudet we do have a check that ensures if there is a pH 7.3 mapping, it must be used. We do allow other terms that aren't mentioned in the mapping file. CHEBI:35780 doesn't have a pH 7.3 mapping in the file.
pgaudet
commented
6 months ago OK, not sure why CHEBI:35780 was in the ontology then - unless this showed up in reports and I hadn't noticed?
deustp01
commented
6 months ago ChEBI terms used in the ontology should be limited to the pH7.3 list
A practical limitation is that when ChEBI has entries for multiple ionization states of a molecule, it does not always identify the one that is the prodominant microspecies at pH 7.3, e.g., phosphatidylcholine (CHEBI:64482) is annotated as the conjugate base of phosphatidylcholine(1+) (CHEBI:49183), but not flagged as the predominant form at pH 7.3, so knowledge external to the ChEBI annotation is needed to identify it. Is that knowledge assembled into a pH 7.3 list someplace?
balhoff
commented
6 months ago OK, not sure why CHEBI:35780 was in the ontology then - unless this showed up in reports and I hadn't noticed?
I don't think it would appear in any reports. Since it isn't mapped to anything in the pH 7.3 file, it is considered to be allowed. I think we have to allow terms that aren't mentioned in the file, since it includes various general terms.
balhoff
commented
6 months ago TODO: @balhoff check how many CHEBI terms we use that are not in the 7.3 mapping file.
balhoff
commented
6 months ago @pgaudet here are 842 CHEBI terms we use that are not mapped to any pH 7.3 value:
deustp01
commented
6 months ago @vanaukenk This file is not yet the basis of a reliable drop-down list to point curators building GO-CAMs to appropriately charged forms of small molecules, I think because of information missing from the ChEBI data that Jim is mining here.
Consider CHEBI:12777 vitamin A. This instance is not a single chemical, but a class that includes molecules like CHEBI:17898 all-trans-retinal that has no ionizable groups at anything near a physiological pH, but also all-trans-retinoic acid (CHEBI:15367) (uncharged) which can ionize to yield all-trans-retinoate (CHEBI:35291) (with a net charge of -1). I expect, though ChEBI provides no indication, that the latter, retinoate, is the major form at pH 7.3. (RHEA builds reactions with the ionized form, not the un-ionized one, e.g., RHEA:42080). Bottom line: more ChEBI annotation (or more propagation of information catalogued elsewhere, e.g., RHEA, into ChEBI) is needed before a useful list for pH-aware GO curation will fall out.
pgaudet
commented
6 months ago Can I change in the ontology the ones for which we do have a 'RHEA-approved'? for example:
pgaudet
commented
6 months ago Google spreadsheet of all the non-pH7.3 chemicals: https://docs.google.com/spreadsheets/d/196YJakCj77UmQklmNb5WpE2i-8PCVxHpFd587G6JceM/edit
deustp01
commented
6 months ago What does the yellow shading on some cells of the spreadsheet mean?
pgaudet
commented
6 months ago These are some of the ones we should easily with the RHEA term.
cmungall
commented
6 months ago I would consider mapping from a stereochemically agnostic form to a stereochemically positive term a separate step. IMO the stereochemically agnostic terms are not useful and actively misleading.
It looks like CHEBI treats atoms differently. I have never understood atoms in CHEBI. In any case it looks like we can just derive these mappings from RHEA. Ideally this would be done upstream of us in GO and we would just get one SOT file that maps CHEBI terms to the representative member.
Of course none of this should be necessary for us in GO. We have the same needs as everyone from CHEBI.
pgaudet
commented
6 months ago After discussion with Elisabeth Coudert from RHEA, I changed all 'inorganic phosphate' to 'phosphate' in term labels and definitions.
deustp01
commented
6 months ago See also ChEBI issue 3823 "Document use of has_major_microspecies_at_pH_7_3 relationship type and how to traverse is-a in CHEBI "
pgaudet
commented
6 months ago New equivalent classes:
suzialeksander
commented
5 months ago from @pgaudet's request, SGD needs a handful of CHEBIs to finish the P2GO import. added to new sheet in the Gsheet.
CHEBI:28593 CHEBI:47519 CHEBI:4883 CHEBI:75442
cmungall
commented
5 months ago I would like to help out here but I can't figure out things from reading this issue top to bottom.
This issue seems to be primarily about filtering the set of CHEBI terms in the ontology. I see PRs that normalize CHEBI terms in the edit file to the ph7.3 form. This is all good! I'm hoping this wasn't done manually, robot or oak could have been used here! let me know if you need help here with the rest.
@suzialeksander, I don't understand the relationship of your ask to the issue here though. It seems that SGD has annotations like
SGD S000005009 AQR1 involved_in GO:0055085 PMID:11922628 IMP P Plasma membrane transporter of the major facilitator superfamily YNL065W protein taxon:559292 20190418 SGD has_input(CHEBI:28593) UniProtKB:P53943
To express the concept of "quinidine transmembrane transport". Because there is no reference to quinidine in the ontology, it doesn't make it into the subset of CHEBI used in Noctua, and hence you get false positive errors - is this correct?
Is the ask here to do on of the following:
quinidine transmembrane transport in the ontology, and push down the SGD annotationsI don't understand the relationship of the google sheet to all of this. I hope there isn't some Noctua QC check that is looking up things in a google sheet...
UPDATE
My current understanding: go-lego has CHEBI terms from the union of the ontology chebi import, plus terms from the Rhea chebi_pH7_3_mapping.tsv file. Some chemicals like quinidine are neither referenced in the ontology nor in the rhea file, so they don't make it into go-lego-chebi-import, and thus are flagged in the annotation QC report.
I think the immediate fix here is to have a small manually curated inclusion list where we can list these chemicals and have them be included in the subset that is available to curators. We can of course signal upstream to rhea that we would like quinidine and others added which might even trigger adding the protonated forms into CHEBI, but we presumably do not want to be blocked on this. CHEBI's protonation strategy is not scalable so we can expect many more omissions like this.
However, this is a patch on an overly complicated system.
Hi @balhoff ChEBI terms used in the ontology should be limited to the pH7.3 list (see https://github.com/geneontology/go-ontology/issues/27012). Can you add a check for this?
Thanks, Pascale