Ref to the line:

Hi Biothings team. I am trying to retrieve records in large batches using multiple inchi keys. THe swagger doc says that you can submit up to 1000 keys in a post request at a time. It seems that no matter how many I supply, the results include 10 records.

example post I am making with 17 unique inchi keys that all resolve individually

fields = 'drugbank.targets,drugbank.drugbank_id,unii.unii,drugcentral.drug_use,drugcentral.bioactivity',

ids = 'FQXGHZNSUOHCLO-IZLXSQMJSA-N,MKWYFZFMAMBPQK-UHFFFAOYSA-J,SXYZQZLHAIHKKY-GSTUPEFVSA-N,WMKGGPCROCCUDY-HEEUSZRZSA-N,XXKJBDTZZBQBCP-MRVPVSSYSA-N,WQVJHHACXVLGBL-UHFFFAOYSA-N,FLOSMHQXBMRNHR-UHFFFAOYSA-N,FLOSMHQXBMRNHR-QPJJXVBHSA-N,ZMJBYMUCKBYSCP-UHFFFAOYSA-N,IMONTRJLAWHYGT-REGVOWLASA-N,MSRILKIQRXUYCT-UHFFFAOYSA-M,ZJEFYLVGGFISGT-UHFFFAOYSA-L,ZJEFYLVGGFISGT-VRZXRVJBSA-L,UFUVLHLTWXBHGZ-KUWMELJBSA-N,UFUVLHLTWXBHGZ-MGZQPHGTSA-N,GGWBHVILAJZWKJ-KJEVSKRMSA-N,GGWBHVILAJZWKJ-CHHCPSLASA-N' 

        url = 'http://mychem.info/v1/drug'
        data = {
            'ids': ids,
            'fields': fields
        }

Cheers

http://drugcentral.org/download

Add this to the landing page query examples section at the bottom:

Get data source metadata:
GET http://mychem.info/metadata

Get the list of all fields:
GET http://mychem.info/metadata/fields

This applicable to mygene.info and myvariant.info APIs as well.

Load rxnorm. scrape API, example:
https://rxnav.nlm.nih.gov/REST/rxcui/595060/allProperties.xml?prop=all

drug-gene interaction database

downloads page: http://www.dgidb.org/downloads

http://tanlab.ucdenver.edu/DSigDB/DSigDBv1.0/geneSearch.html

Tried: http://mychem.info/v1/query?q=drugbank.targets.uniprot:P24941&fields=_id&size=20

Shows total number of responses is 137

But it's just only returning 10 results.

I tried different sizes, e.g. 2, it works. However, putting any number larger than 10 wouldn't work.

First, the biothings.api commit point in requirements_web.txt is incorrect.

The commit point in the requirements_web.txt file is:
3933f82392042b1446456e13700f51dea9b4c975

However, the following exception is thrown at run time:
mychem.info | Traceback (most recent call last):
mychem.info | File "bin/hub.py", line 62, in
mychem.info | from biothings.hub.databuild.syncer import ThrottledESJsonDiffSyncer, ThrottledESJsonDiffSelfContainedSyncer

I manually checked the biothings.api commit point and the class ThrottledESJsonDiffSyncer is missing in this commit. Which commit point should be used?

Further, the following packages are missing in the requirements file:
aiocron
IPython
pympler

See: https://www.ebi.ac.uk/chembldb/index.php/compound/inspect/CHEMBL1431

http://mychem.info/v1/chem/XZWYZXLIPXDOLR-UHFFFAOYSA-N?fields=chembl

Pubchem flat files located here: ftp://ftp.ncbi.nlm.nih.gov/pubchem/Compound/CURRENT-Full/XML/ don't have CAS numbers.
But they are accessible from PubChemRDF
example
along with lots of other identifiers

If I do a simple query q=siltuximab, I get 5 results, with these identifiers and keys:

57894-421 ['_id', '_score', 'ndc']
57894-420 ['_id', '_score', 'ndc']
CHEMBL1743070 ['_id', '_score', 'chembl', 'drugcentral']
DB09036 ['_id', '_score', 'drugbank']
T4H8FMA7IM ['_id', '_score', 'aeolus', 'unii']

The way I actually want to query this data is by asking for compounds that have a particular aeolus outcome. So if I come in and query for a particular outcome, and it matches siluximab, I will get back only aeolus and unii information. I won't get chembl or drugcentral, making it hard to give this compound an identifier that I can integrate other data with.

I don't know if this is a general feature or if I just found one, but it seemed in testing that I often didn't get either a chembl or chebi node when querying by aeolus.

Current external vocab:

"external_vocabulary": {
"atc": "N07XX02(riluzole)",
"ndfrt": "N0000148421(RILUZOLE)",
"rxnorm": "35623(Riluzole)",
"umls": "C0073379(Riluzole)"
},

should be:

"external_vocabulary": {
"atc": "N07XX02",
"ndfrt": "N0000148421",
"rxnorm": "35623",
"umls": "C0073379"
},

mychem.info should support both http and https like MyGene.info and MyVariant.info

"bioactivity": {
"act_comment": "Mechanism of Action; CHEMBL2362997; PROTEIN COMPLEX GROUP",
"act_source": "CHEMBL",
"action_type": "ANTAGONIST",
"gene_name": "CHRNA1|CHRNB1|CHRND|CHRNE|CHRNG",
"moa": "1",
"moa_source": "SCIENTIFIC LITERATURE",
"swissprot": "ACHA_HUMAN|ACHB_HUMAN|ACHD_HUMAN|ACHE_HUMAN|ACHG_HUMAN",
"target": "Muscle-type nicotinic acetylcholine receptor",
"target_class": "Ion channel",
"uniprot_id": "P02708|P07510|P11230|Q04844|Q07001"
},

Query: http://mychem.info/v1/drug/CHEMBL503?fields=chebi.pubchem_database_links

There are two problems with this field:

• The results should not be in CURIEs, e.g. (prefix:value), it should be just the value itself, which is 53232.

• PubChem Compound ID (CID) and PubChem Substance ID (SID) are two different IDs, they should be put in two separate fields, so the current JSON structure should be:
  { 'pubchem_database_link': { 'CID': 53232, 'SID': 26697338 } }

http://mychem.info/v1/query?q=drugbank.name:riluzole&fields=drugcentral.xref

This line:

The values of pharmgkb.type field look like just "Drug" and "Prodrug", better to index it as keyword. Is it somehow recognized in the inspector as "text" field?

http://www.hmdb.ca/metabolites/HMDB0000012

• drugcentral collection was updated, new structure => run inspector for new mapping (done in 501a9a1)
• sider: use keylookup decorator to convert pubchem ID to InchiKey + rerun mapping generation (parser has changed ?) : mapping done in 67a449d (not keylookup)

For example, drugbank.id field should be included in "all", but does not seems included:

Either http://mychem.info/v1/chem/DB01076 or http://mychem.info/v1/query?q=drugbank.id:DB01076 works, but not this:

http://mychem.info/v1/query?q=DB01076

We should double check all fields need to be included in "all".

CMap should provide gene expression measurements associated with chemical compounds. In theory, that data should be accessible via https://clue.io/cmap, but on first glance it is unclear how one can actually access these data (via either downloadable files or API). Needs someone to dig through the links (including the GEO records mentioned in https://clue.io/GEO-guide and this user guide).

Also, I believe CMap and LINCS are a partially overlapping set of data. Would also be good to import LINCS -- more info probably available at http://www.lincsproject.org/.

This ticket is partially motivated by a request from the alpha translator team, so they should be notified when we have more info...

Sample data:

http://mychem.info/v1/query?q=_exists_:aeolus.outcomes&fields=aeolus.outcomes

this field "aeolus.outcomes.meddra_code" was labelled as "aeolus.outcomes.code" in the mapping:

aeolus/aeolus_upload.py#L74

should change the mapping to "meddra_code", and this query should work:

q=aeolus.outcomes.meddra_code:10028813

Current structure:

{
  "molecule_synonym": "DNDI996469",
  "syn_type": "OTHER",
  "synonyms": "DNDI996469"
},
{
  "molecule_synonym": "DNDI996469", 
  "syn_type": "RESEARCH_CODE",
  "synonyms": "DNDI996469"
}
]

1. Should be grouped based on "syn_type"
2. "syn_type" should serve as the key
3. "molecule_synonym" and "synonyms" fields look redundant

Proposed structure:

{
  "research_code": "DND1996469",
  "other": "DND1996469"
}

In current MyChem.info, drugbank ID is sometimes represented as "drugbank_id" while sometimes represented as "id".

Example of "drugbank_id": http://mychem.info/v1/drug/PMXMIIMHBWHSKN-UHFFFAOYSA-N?fields=drugbank.drugbank_id,drugbank.id

Example of "id": http://mychem.info/v1/query?q=_exists_:drugbank.id&fields=drugbank.id,drugbank.drugbank_id

We should uniform it. So all "drugbank_id" should be converted to "id".

Possible lines to look at in the code:

TODO:
* update the license text for pharmgkb data source
* see if any additional data we can include

1. Fields documented in the mapping file but returning zero hit:
   drugbank.pharmacology.snp_effects.effects.rs-id
   drugbank.pharmacology.snp_effects.effects.gene-symbol
   ...and many others

2. Missing 'gene_name' field
   Take an example of Maraviroc, a CCR5-targeted drug
   https://www.drugbank.ca/drugs/DB04835#targets
   Above is the link to this drug annotation information provided by drugbank website,
   there is a field called 'Gene Name' in the table, however, that field is missing in c.biothings.io

The Roles and ChEBI ontology data is missing from the chebi response

Example:
http://mychem.info/v1/chem/XZWYZXLIPXDOLR-UHFFFAOYSA-N?fields=chebi
http://www.ebi.ac.uk/chebi/searchId.do?chebiId=CHEBI%3A6801

for drugcentral.drug_use.indication and contraindication, if there is more than one indication, you get a list. But if there is only one, then you don't get a list, you just get the item.

That's fine, but in terms of parsing the output it means that you have to do a little extra logic. From my perspective, it would be nicer to return a list of one item.

CAS is missing from SDF file but is present (along with many other xrefs) in the obo/owl files:
ftp://ftp.ebi.ac.uk/pub/databases/chebi/ontology/chebi.obo.gz

[Term]
id: CHEBI:15578
name: (25S)-5beta-spirostan-3beta-ol
alt_id: CHEBI:178
alt_id: CHEBI:10854
alt_id: CHEBI:18537
subset: 3_STAR
synonym: "(3beta,5beta,25S)-spirostan-3-ol" RELATED [ChemIDplus]
...
xref: KNApSAcK:C00003590 
xref: Beilstein:91757 "Beilstein"
xref: KEGG:C03963 
xref: CAS:126-19-2 "KEGG COMPOUND"
xref: CAS:126-19-2 "ChemIDplus"
is_a: CHEBI:26606

1. Data should be grouped based on xref_src
2. CURIEs should be converted to values if it's the standard representation by the source
   Example:
   "cross_references": {
   "xref_id": 170466478,
   "xref_name": "SID: 170466478",
   "xref_src": "PubChem"
   }

=>

"cross_references": {
"pubchem_sid": "170466478"
}

https://comptox.epa.gov/dashboard

Looks like they do some aggregation among other resources, but they also have new data of their own.

Downloadable files here: https://comptox.epa.gov/dashboard/downloads

for example, see http://mychem.info/v1/chem/KTUFNOKKBVMGRW-UHFFFAOYSA-N?fields=drugbank,drugcentral

screenshot:

http://www.guidetopharmacology.org/download.jsp

CC-BY-SA

main page is at https://www.rhea-db.org/. provides useful information on chemical reactions (mapped to chebi IDs)

Many possible formats on their download page, but will need some work to figure out which is the right one to use.

Hi,

Really awesome service and concept.

I noticed the 'non-commercial' use flag on mychem.info that isn't present on mygene.info. Could you comment/blog on why the difference? Is there was any way (perhaps an endpoint with limited datasets, or an alternative service) that it could be used in a commercial environment? I understand that there is a vast difference in the history of licensing between biology and chemistry data - As I am sure you are aware, non-commercial licenses cause a lot of complications for integration efforts.

Thanks,

Iain

There are 4245 drugs in the aeolus data set. All have a rxcui. 3043 can be matched to an InchiKey or UNII using the data from the FDA link.
Look into using Wikidata to do additional normalization.

Wanted to follow up with a question we discussed offline about AEOLUS data, which describe adverse drug reactions (ADRs), using meddra terms to codify the adverse outcomes. Specifically, I noted the inclusion of 'outcomes' in the mydrug AEOLUS data that seem to be primary indications for a drug (i.e. what it is used to treat), rather than adverse drug reactions.

For example, the mydrug results for imatinib, which is used to treat various leukemias, and outcomes include things like "Leukaemia", "Chronic lymphocytic leukaemia", "Acute leukaemia".

In your parsing of AEOLUS data, was there any metadata that would tell you what medrra terms represent indications for a particular drug so these could be filtered from the results, or at least tagged as being indications rather than adverse reactions to the drug? This would be a significant improvement to the dataset you generate.

UMLS provide IDs for many biological entity types, including chemicals and drugs:

https://www.nlm.nih.gov/research/umls/

Similar to this UMLS <-> gene mapping for MyGene.info:
https://github.com/biothings/mygene.info/tree/master/src/dataload/sources/umls

, we can import the mapping of UMLS <-> chemicals/drugs into MyChem.info.

Currently dumper informs about a new release and it needs the zip file to be manually downloaded and put on to the server. As of release 5.1.0, website shows a command we can use to download it using curl:

https://www.drugbank.ca/releases/5-1-0

curl -Lfv -o filename.zip -u EMAIL:PASSWORD https://www.drugbank.ca/releases/5-1-0/downloads/all-full-database

Similar to how mygene.info/v3/query behavior. By default, the matching hits return a subset of "essential fields" by default, instead of all available fields. This will return hits much smaller and faster.

Users can always get the specific fields they need by specifying "fields" parameter, or passing fields=all to return all available field (the current default)

I have a vague recollection of mentioning this issue before, but I forget the response. So just adding it here.

I was expecting this query:
http://mychem.info/v1/query?q=imatinib

to return a superset of records relative to this fielded query:
http://mychem.info/v1/query?q=drugbank.name%3Aimatinib

But the first query currently returns no results. I'm guessing this has to do with what fields are indexed by default. I'll leave it you guys to decide whether this is the desired behavior or something to be changed....

See: https://github.com/biothings/mychem.info/blob/master/src/hub/dataload/sources/pubchem/pubchem_parser.py#L21

Should consider remove this code and add a decorator for the pubchem parser.

Pubchem ID in mychem.info should have unified name
"pharmgkb.pubchem_compound": "132971",
"pubchem.cid": "CID132971"

The following fields need to be fixed in the parser level to conform to the same data type

7. drugcentral.drug_use.indication.snomed_concept_id: int/float -> int (should be int, need to look at why it's float in some cases)

This is very minor and low priority.

I found myself many times to click into data folder instead when I actually wanted to go to hub folder. Probably just because the name "data" confused me.

Don't know if that's just me or not. One suggestion would be to move /src/data to the top-level /data, considering this folder is not the application code itself, just some data relevant to the application.

And this could be applicable for other biothings API code bases.