biothings / mychem.info

MyChem.info: A BioThings API for chemical/drug annotations
http://mychem.info
Apache License 2.0
16 stars 14 forks source link

ChEBI records not properly associated with other documents #86

Closed zcqian closed 3 years ago

zcqian commented 3 years ago

It seems that some ChEBI aren't properly indexed, for instance CHEBI:81931, CHEBI:63038, etc. (both listed in their entity of the month), are not indexed with the respective InChIKey as its _id.

In this case, users cannot find related data except for ChEBI when using something like https://mychem.info/v1/chem/CHEBI%3A81931 (e.g. no PubChem info), but all other records are shown when using https://mychem.info/v1/chem/CID%3A61021

kevinxin90 commented 3 years ago

@newgene

Currently, seems all CHEBI records are not properly merged with other MyChem docs. The reason is in current data transform module for MyChem.info, all edges from CHEBI ID are removed. See line 67-76 in the keylook.py file.

And I was looking at the commit history of the Keylookup.py file, in this commit, @sirloon removed/commented out all edges from CHEBI. I believe that's the reason why CHEBI docs are not merged correctly with other documents. I don't recall any discussion that lead to this decision. Do you recall any? @newgene

newgene commented 3 years ago

@kevinxin90 don't remember any reason for commenting out chebi edges. Let's add them back and reload CHEBI.

kevinxin90 commented 3 years ago

Fixed in https://github.com/biothings/mychem.info/pull/87/commits/838a794194e193c8fd478384fb38357ca333ebe4