Closed apriltuesday closed 2 months ago
After a discussion with @DSuveges we have agreed that the most accurate solution is to create a drugs
object. This will be a list of structs with 2 keys:
drugFromSource
: name of the drug at source
drugId
: CHEMBL ID that will be populated by us later on in the drug mapping step, so it will be null in your case
This solution solves the problem of keeping the parity between name and ID. Now that CHEBIs have been removed, you are not affected by this, but it is important for us so that we can map each drug name to the CHEMBL ID after it is mapped.
Thank you very much! Please let us know of any issues
Thanks @ireneisdoomed, this makes sense. I'll make the changes on our end and let you know any questions I have.
As discussed in 22 April meeting, we should make the attribute
drugFromSource
a list when split on/
only (other delimiters have a different semantic in PGKB and should continue to be exploded into separate evidence) and rename the attribute to be pluraldrugsFromSource
.This is to support drug combinations like ivacaftor / lumacaftor.
To remove complexity and because CHEBI IDs aren't being used by OT, we will remove the
drugFromSourceId
attribute entirely.Related OT issue here.