New API based on MyChem drugcentral.bioactivity data

[colleenXu]

Related to biothings/biothings_explorer#532 (comment)

MyChem drugcentral.bioactivity data, in subject-association-object format

• We want 1 association per unique combo of subject/object/action_type.
  • Right now in MyChem, all data is aggregated by unique chemical (chem-centric format). So when a chemical has multiple bioactivity, we can't retrieve only the sections that have a specific action_type.
  • MyChem currently has info for 2772 unique chemicals
• We want the gene ID to be UniProtKB (or NCBIGene or ENSEMBL ENSG maybe).
  • Right now, it's UniProtKB but 119 chemicals lack the UniProtKB record...and their bioactivity info looks odd (not acting on genes?)

[colleenXu]

Related issue: biothings/biothings_explorer#316, JQ development work

[rjawesome]

With jq implemented, exisiting API can be used with modifications to smartapi yaml. See this comment

[colleenXu]

Note on the relationships in the resource:

• Some chemicals have many drugcentral.bioactivity elements (>10), and only some of those elements have the action_type field that is most useful for filtering for specific relationships.
• Perhaps act_type value could be used to assume certain relationships, but this has the same issue as Discuss BindingDB issues: related to parser / API? #99; IC50 and Ki might be mean "inhibitory relationship" but that may depend on the actual associated value being large / small... (EDIT: And I found relationships labeled "agonist" that had "IC50" values so...this method may not work)

Example: methotrexate's drugcentral info in MyChem

[colleenXu]

More notes on "action_type" and targets:

• many bioactivity entries seem to be missing "action_types"; this seems to be a bigger issue than for chembl.drug_mechanism (for example, aripiprazole has 45 bioactivity entries but only 6 have action_type values
• many bioactivity entries use non-human targets (for example, out of aripiprazole's 6 bioactivites with action_type values, only 3 are human). The association method / BTE post-processing with JQ may be helpful here, if we want to filter for only human targets (vs non-human orthologs)...

[colleenXu]

Example of current MyChem structure vs association-based structure

(basically repeating #100 (comment) using an example from this resource)

expand here for the example

Background

aripiprazole (CHEMBL1112) has multiple bioactivity entries with different categories (action_types). I'll give two of them as examples:

• PARTIAL AGONIST for D(2) dopamine receptor (P14416)
• ANTAGONIST for 5-hydroxytryptamine receptor 2A (P28223)

In MyChem

these two bioactivity entries are nested inside the drugcentral.bioactivity field of the MyChem record for this chemical: https://mychem.info/v1/chem/60795?fields=drugcentral. This means BTE post-processing (JQ?) is needed to retrieve only the ANTAGONIST bioactivity (or vice versa).

{
    "drugcentral": {
        "bioactivity": "CHEMBL1089636",
        "drug_mechanisms": [
            {"action_type": "PARTIAL_AGONIST", "act_source": "WOMBAT-PK",..., "uniprot": [ {"uniprot_id": "P14416", "gene_symbol": "DRD2", ...} ] },
            {"action_type": "ANTAGONIST", "act_source": "WOMBAT-PK",..., "uniprot": [ {"uniprot_id": "P28223", "gene_symbol": "HTR2A", ...} ] }
        ]
    }
}

association-based structure

However, if we use an association-based structure, we can make two separate records. And these two records can be retrieved separately depending on what association.action_type is set to when querying.

{
    "subject": { "drug_chembl_id": "CHEMBL1112", ...},
    "association": { "action_type": "PARTIAL_AGONIST", "act_source": "WOMBAT-PK",...},
    "object": { "uniprot_id": "P14416", "uniprot_gene_symbol": "DRD2", ...}
},
{
    "subject": { "drug_chembl_id": "CHEMBL1112", ...},
    "association": { "action_type": "ANTAGONIST", "act_source": "WOMBAT-PK",...},
    "object": { "uniprot_id": "P28223", "uniprot_gene_symbol": "HTR2A", ...}
}

[colleenXu]

At the moment, creating a new API is not necessary. Using MyChem's new list_filter function (discussed here biothings/api-respone-transform.js#38 (comment)), I can write operations in the chem -> gene direction that are more specific (however not all records are covered).

Andrew decided that this was fine to try out. I therefore moved this issue to "on-hold"

The edits to the SmartAPI yamls are
for biolink 3.1.1 (currently on dev/CI instances):

• NCATS-Tangerine/translator-api-registry@5d4391a
• NCATS-Tangerine/translator-api-registry@fc2a207
• NCATS-Tangerine/translator-api-registry@c97da47

for biolink 2.4.8 (currently on test/prod instances):

• NCATS-Tangerine/translator-api-registry@81e6e48
• NCATS-Tangerine/translator-api-registry@7e275d8
• NCATS-Tangerine/translator-api-registry@3226da8

[colleenXu]

Closing. If we notice an issue with this (not able to retrieve some desired info in the chem -> gene direction, want to be able to use more specific predicates in the gene -> chem direction), we'll reopen.

[colleenXu]

@newgene

two issues encountered when trying to write operations:

• I'd like to be able to "list_filter" for items that are MISSING the drugcentral.bioactivity.action_type field. At the moment, I have to specify 1 or more string values.
• Getting this issue addressed would help with writing operations: re-index drugcentral.bioactivity.action_type field as keyword, perhaps index Chembl.drug_indications.first_approval mychem.info#153