Improve automated mappings

[matentzn]

Here are some example problems with the automated mappings (check major MRE mapping issue):

You can find for example this entry in there:

MRE:0000100 Exposure to 2,2',4'-tribromodiphenyl ether TICKETED ebi-chebi/ChEBI#3338

Which resulted in the creation of CHEBI:138001 (called 2-bromophenyl 2,4-dibromophenyl ether).

The class is correctly in ECTO - but how would rdfmatcher, or any tool, be smart enough to do this match? To do this properly, we need to add manual matches to the pipeline? Or are there any other approaches we can use here?

[diatomsRcool]

chemical names are a known disaster - I did some mapping here ages ago that maybe we can use - if I can find it.

[cmungall]

We don't need to do any specific chemical name matching

The general principle with our modular ontology building approach is

• we SHOULD use the same primary label in the composed term in the base term
• exceptions MUST be intentional
• if the base class primary label is not used then an exact synonym SHOULD be used
• one exception here is chebi which does not assign synonym scope
• the derived term term MUST use one of label or any kind of synonym from the base term

In this case CHEBI took our request for foo and gave us a bar. They did not add foo as a synonym

Was this intentional on their part?

• If not, then we should make a new ticket to add the synonym
• If so, then it indicates our string "foo" was not good for some reason. perhaps it is ambiguous. Either we need more clarification, and we either need a new term in CHEBI or we need to use a different string ourselves

Either way the matching is doing exactly asintended, lack of matches indicates a gap