last day (18 days later) » 

2:01 PM
0
Q: Why use both universal identifiers and separately incremented primary keys for strong entities?

atevmThere is a good public example for this: The relational model of ChEMBL (A bioactive molecule database). Here all strong entities (such as molecules) have a database-wide unique identifier which also serves as an universal identifiers: ChEMBL ID (such as CHEMBL25 for ASPIRIN). These strong entiti...

 
I'll have to look into it. I'm more familiar with PubChem.
 
What kid of database do they use in their back end? I downloaded some substance from them and I didn't find any db dump or schema on their FTP, just XML files. It would be awesome to know their database and data model. I also read some of their publications but they did not detail this part of their solution.
 
I don't see chembl_id_lookup in their schema. Which table.column are you referring to?
They use an RDBMS. I believe the myChEMBL dumps are for MySQL. Link to schema was in my prior comment.
 
@MichaelKutz It is in the "GENERAL" (grey) rectangle in the top-middle part on the diagram what you linked. One of its column is "entity_type".
Yes they use an RDBMS for ChEMBL. I referred to PubChem and Entrez when I said I don't known the database technology in use.
 
Long gone are the days of using proprietary DB Technologies for storing bioinformatic/cheminformatics Even then, I believe the MDL ISIS/Server used Oracle as its back-end. (MDL moved to the MDL Direct data cartridge for indexing chemical structures)
 
2:06 PM
I agree I use PostgreSQL with RDKit cartridge to represent molecules and their fingerprints.
Now I am having a meeting, in a hour (hopefully) I'll write further details
 
I've been out of the Pharma industry for 6+ years. MSI turned to Accelrys near the time I started. Accelrys merged with MDL (to form Biovia - iirc) just before I left.
 
 
1 hour later…
3:13 PM
Thank you. I checked on MDL, yes it uses Oracle. It is somewhat similar to our system, but not entirely. However it is unlikely that I can persuade my supervisors to buy it.
My original question however is not specific to pharmacology. It is a valid problem in any research oriented OLAP data model, which plans to provide universal identifiers externally and use integer surrogates for strong entities.
But, aside I am really happy to meet someone with your experience in this field, because in my company I am the most experienced in biomedical data and knowledge engineering, which is a problem because I do not consider myself experienced enough.
 

  last day (18 days later) »