map pathways column in big trait table

Knowledge-Graph-Hub / kg-microbe

https://knowledge-graph-hub.github.io/kg-microbe/index.html

BSD 3-Clause "New" or "Revised" License

15 stars 3 forks source link

map pathways column in big trait table #5

Open cmungall opened 3 years ago

cmungall commented 3 years ago

Split from #2.

1 trithionate_oxidation 1 carbonmonoxide_oxidation 1 tetrathionate_oxidation, iron_reduction 1 pyrrhotite_oxidation 1 galena_oxidation 1 thiocyanate_oxidation 1 carbonylsulfide_oxidation ... 182 sulfur_reduction 186 thiosulfate_reduction 353 aerobic_chemo_heterotrophy 366 fermentation 371 denitrification 400 nitrite_reduction 983 NA 1420 nitrate_reduction

These should all map to GO

cmungall commented 3 years ago

To work on this ticket, make PRs on pathways.sssom.tsv in schemas/

realmarcin commented 3 years ago

Transferring earlier discussions here:

Original ideas was to use GO terms for NER mapping. Per Harshad this ran into two issues: 1) underscores delimiters in trait table, and 2) some different wording for pathways which included context of metabolism mode (eg aerobic).
1. is trivial but 2. may require some manual curation. We may also need to compose terms from say ECOCORE metabolism mode together with GO function.
In the future we could infer some links between carbon growth substrates, metabolism mode, and pathways but may be more useful in the context of linking with KG-Pathways.

realmarcin commented 3 years ago

Placeholder for ongoing work, edit as needed.

Merge NER results into SSOM mapping file is a next step.

Then update KG build workflow to load mappings from SSOM file instead of hard-coded in ingest script.

cmungall commented 3 years ago

Let's start with

[x] automated mapping to GO. I did this using rdfmatch, but runner may give additional hints
[ ] automated mapping to KEGG pathways. @hrshdhgd is working on this

once this is done, we will assess the situation. If KEGG has higher coverage than GO, this is useful information I can take back to GO and figure out why GO has poorer coverage here

If neither gives us the coverage we need we need to look a bit deeper. Is this a matter of insufficient precomposed terms in the ontology? Is there a systematic thing we can address, e.g. by adding synonyms for GO?

hrshdhgd commented 3 years ago

I did a quick OGER run through the KEGG Pathways with GO as dictionary (output file).

Total number of KEGG pathways: 542 Total number of KEGG pathways tagged by GO: 352 Total Number of pathways in KEGG absent in GO: 190