Extract ChEMBL version associated per compound

[czodrowskilab]

To whom it may concern,

is there a way to extract something like a publication/experimental date associated with a ChEMBL compound?

For benchmarking studies, we would like to use time-split validation. Therefore, it would be of tremendous help for us to get an association between ChEMBL version (or publication/experimental date) and a ChEMBL compound?

Best regards, Paul

[dominiquesydow]

Hi!

Thanks for the chembl_webresource_client - we are using it at @volkamerlab a lot!

My question is related to @czodrowskilab's question here (I think), so I'll add it here.

Is there a field for the date when

• a compound or
• a bioactivity

was added to ChEMBL?

I know of the document_year field, which I can fetch for bioactivities:

bioactivities = bioactivities_api.filter(
    target_chembl_id='CHEMBL941').only(
        'activity_id',  
        'document_year'
     )
)

Given the ChEMBL database scheme, I guess it works as follows: From doc_id (field) in activities (table) go to doc_id (field) in docs (table) and extract year (field). This probably also works similarly for compounds.

However, the document's year

• does not necessarily did to equal the bioactivity or compound deposition date in ChEMBL and
• does not need to be the same for compounds and bioactivities, right?

Thus, back to my question: What is the best way to filter compound or bioactivity entries in ChEMBL by their deposition date?

Thank you for your time.

[dominiquesydow]

I am realizing: The deposition date probably equal the ChEMBL version in which the compound/bioactivity was added.

Is there a way to access the version? In the ChEMBL scheme, the version table seems not to be connected with any other table: https://www.ebi.ac.uk/chembl/db_schema

[dominiquesydow]

Update on this matter: The ChEMBL support team (https://www.ebi.ac.uk/support/) let me know that it is currently not possible to extract data from the web services (or the interface) for a previous ChEMBL release.

[eloyfelix]

Hi, sorry for not having replied this before. As Dominique says, this is currently not possible but we are working towards supporting it on future ChEMBL versions.

[iwwwish]

Dear ChEMBL team,

Thank you for chembl_webresource_client. Is there any update on whether the version could be directly/indirectly linked to an activity record?

Like @czodrowskilab, I am also interested in splitting the activity data for a target by chembl version in which it first appeared (to mimic time-split).

Thanks, Vishal

[BZdrazil]

We are working on implementing time stamps for deposited data sets. We are exploring how to deal with updates of documents now. Tentatively, we'll release this with CHEMBL 34.

[BZdrazil]

However, time stamps for data sets are already available ( and have always been) via the VERSION table which includes CREATION_DATE for every ChEMBL release. By querying for which release a document has been added, you'll get your time stamps. After we've solved the question about updated documents, we plan to make that information more easily accessible, likely via a new table.