My slides at HeartBD2K weekly technical conference call.

1. Chunlei Wu, Ph.D.
cwu@scripps.edu
@chunleiwu
Associate Professor of Molecular Medicine
Dept. of Molecular Experimental Medicine
The Scripps Research Institute
La Jolla, CA, USA
01/22/2016
From MyGene.info and MyVariant.info towards BioThings API

2. As a
MyGene.info and MyVariant.info recap
Annotations
Gene
Variant
(Aggregated)
(high-performance)
(real-time) Web Service

3. So many variant annotation resources
dbNSFP
The Exome Aggregation
Consortium (ExAC)

4. Annotations centered around bio-entities
Gene
G
Variant
V
Pathway
P
D
Metabolite
M
Disease

5. Simple JSON-based Aggregation mechanism
{
"_id": "chr1:g.196659237C>T",
"cadd": { … },
"clinvar": { … },
"cosmic": { … },
"dbsnp": { … },
"dbnsfp": { … },
"evs": { … },
"emv": { … },
"mutdb": { … },
"gwassnp": { … },
"snpedia": { … },
"wellderly": { … }
}
{
"_id": "chr1:g.196659237C>T",
“dbsnp": {
"snpclass": "single",
"rsid": "rs1061170",
"func": "missense"
}
}
{
"_id": "chr1:g.196659237C>T",
“cosmic": {
"tumor_site": "breast",
"mut_freq": 0.49,
}
}
{
"_id": "chr1:g.196659237C>T",
“dbnsfp": {
“sift": {
"breast“: “tolerated”,
“val”: 1
}
}
}
“cadd” “clinvar” “evs” “mutdb”
…

6. Keep data always up-to-date
Each data source is updated individually. Colors
indicate their different updating schedules.
Schematic view of MyVariant.info architecture

7. High-performance web service APIs
Schematic view of MyVariant.info architecture

8. MyVariant.info for the end users:
http://MyVariant.info
(currently v1 API, two endpoints)
http://MyVariant.info/v1/query?q=<query>
any query term(s)
matching variant hits
http://MyVariant.info/v1/variant/<variantid>
hgvs id(s)
matching variant object(s)
Both supports batch-mode via POST
Simple API. No sign-up. No API key.
Try our live API , and documentations

9. MyGene.info for the end users:
http://MyGene.info
(currently v2 API, two endpoints)
http://MyGene.info/v2/query?q=<query>
any query term(s)
matching gene hits
http://MyGene.info/v2/gene/<geneid>
gene id(s)
matching gene object(s)
Both supports batch-mode via POST
Simple API. No sign-up. No API key.
Try our live API , and documentations

10. MyGene.info usage updates
last
year
this
year
2M
3MMonthly hits in Millions

11. Usage spikes (5M hits/day) during X-Mas 2014

12. 30%9%
35%
26%
Increased clients adoption
Requests by MyGene.info clients
Highlights:
• mygene Python client usage now surpasses BioGPS usage
• mygene R client usage now increased to 9% from <1%
10/07/2015-01/05/2016

13. 30%9%
35%
26%
Increased clients adoption
mygene Python client hosted in PyPI
mygene R client hosted in Bioconductor

14. MyVariant.info updates
Total over 334 Millions of annotated variants
The Exome Aggregation Consortium (ExAC)
New additions:
dbNSFP
Updated:

15. MyVariant.info updates
30%
68%
2%
10/07/2015-01/05/2016
1 Million requests in 3 months

16. MyVariant.info official Python/R Clients
myvariant Python client hosted in PyPI
(initial release in Aug 2015)
myvariant R client hosted in Bioconductor
(initial release in Oct 2015)

17. A Node.js client made by a user with passion

18. Next?
MyVariant.info
MyGene.info

19. Make our APIs serve Linked Data
via

20. Why Linked Data?
Gene
G
Variant
V
Pathway
P
D
Metabolite
M
Disease

21. Linked Data for data aggregation
MyVariant.info
V
Another Variant API
V
V

22. Linked Data for data aggregation
MyVariant.info Another Variant API
{
"_id": "chr1:g.196659237C>T",
“cosmic": {
"tumor_site": "breast",
"mut_freq": 0.49,
},
"clinvar": {…},
"dbsnp": {…},
…
}
{
"pop": "GWD",
"nobs": 226,
"freq": 0.371681415929,
…
}
{
"_id": "chr1:g.196659237C>T",
“cosmic": {
"tumor_site": "breast",
"mut_freq": 0.49,
},
"clinvar": {…},
"dbsnp": {…},
"new_src": {
"pop": "GWD",
"nobs": 226,
"freq": 0.371681415929
},
…
}

23. JSON + context = JSON-LD
{
"@context": {
"clinvar": "http://schema.myvariant.info/datasource/clinvar",
"rcv": "http://schema.myvariant.info/datanode/rcv",
"gene": "http://schema.myvariant.info/datanode/gene",
"_id": "@id"
},
"_id": "chr6:g.26093141G>A",
"clinvar": {
"@context": {
"uniprot": "http://identifiers.org/uniprot/",
"omim": "http://identifiers.org/omim/"
},
"chrom": "6",
"alt": "A",
"ref": "G",
"allele_id": 15048,
"rsid": "rs1800562",
"rcv": {
"@context": {
"accession": "http://identifer.org/clinvar"
},
"accession": "RCV000000020",
"origin": "germline",
"clinical_significance": "risk factor"
},
"gene": {
"@context": {
"symbol": "http://identifiers.org/hgnc.symbol/"
},
"id": "3077",
"symbol": "HFE"
},
"omim": "613609.0001",
"variant_id": 9
}
}

24. Processed JSON-LD
<chr6:g.26093141G>A> <http://schema.myvariant.info/datasource/clinvar> _:b0 .
_:b0 <http://identifiers.org/omim/> "613609.0001" .
_:b0 <http://schema.myvariant.info/datanode/gene> _:b1 .
_:b0 <http://schema.myvariant.info/datanode/rcv> _:b2 .
_:b1 <http://identifiers.org/hgnc.symbol/> "HFE" .
_:b2 <http://identifer.org/clinvar> "RCV000000020" .
JSON-LD N-Quads output:
{
"@id": "chr6:g.26093141G>A",
"http://schema.myvariant.info/datasource/clinvar": {
"http://identifiers.org/omim/": "613609.0001",
"http://schema.myvariant.info/datanode/gene": {
"http://identifiers.org/hgnc.symbol/": "HFE"
},
"http://schema.myvariant.info/datanode/rcv": {
"http://identifer.org/clinvar": "RCV000000020"
}
}
}
JSON-LD compacted output:

25. In a nut-shell, what JSON-LD context does?
Marks values in a JSON object to defined URIs
"http://identifer.org/clinvar"
→clinvar.rcv.accession

26. JSON-LD context makes your data
"Linkable"
"Linked"
Downstream
processing libraries

27. A Python library for processing JSON-LD data
In [1]: fetch_value_source_for_variant("chr6:g.26093141G>A","http://identifiers.org/dbsnp/")
Out[1]:
['rs1800562 http://schema.myvarint.info/datasource/dbnsfp',
'rs1800562 http://schema.myvarint.info/datasource/clinvar',
'rs1800562 http://schema.myvarint.info/datasource/dbsnp',
'rs1800562 http://schema.myvarint.info/datasource/evs',
'rs1800562 http://schema.myvarint.info/datasource/gwassnps',
'rs1800562 http://schema.myvarint.info/datasource/mutdb']
By Kevin Xin

28. Need to define an API specs
• Output as a JSON object with a defined _id.
• "jsonld=true/false" toggle for the inclusion of JSON-LD
context.
• Support the retrieval of a single entity via GET
(use case: individual data aggregation on the fly)
• Support the retrieval of a list of entities via POST
(use case: routine data aggregation in batches)
• Output should indicate the entity existence:
GET /variant/<unknown_id>  404
POST /variant/ id1, <unknown_id>, id3 
[id1: {…},
<unknown_id>: "notfound",
id3: {…}]
to enable data exchange via JSON-LD

29. BioThings
API
MyVariant.info
MyGene.info
By Cyrus Afrasiabi

30. BioThings API
MyVariant.info
MyGene.info
JSON data
aggregation
mechanism
High-
performance
query engine
Well-designed
REST API
pattern
JSON-LD
enabled
Linked Data
Data-updating scheduler
Python/R clients
…

31. Data-sharing via Web API is trending
Making a single web service is trivial,
but making a sustainable/scalable
web API is non-trivial.
We would like to help other groups to
create their own hosted web API for
sharing their data.

32. Action item 1: BioThings API whitepaper
Also the action item from last BD2K CA
consortium meeting and the API working
group from last year's NIH BD2K AHM

33. Action item 2: BioThings API framework
NIH commons
Infrastructure as a Service:
Software as a Service:
BioThings API

34. Action item 3: expansion to other "BioThings"
D
Disease
D
Drugs
MyDrug.info MyDisease.info
need an alt. name here

35. Acknowledgement
Funding and Support
U54GM114833
U01HG008473
Washtington U:
Ben Ainscough
Obi Griffith
TSRI:
Andrew Su
Jiwen Xin
Cyrus Afrasiabi
Ginger Tsueng
Adam Mark
Greg Stupp
Tim Putman
STSI:
Eric Topol
Ali Torkamani
Galina Erikson
U. Washington:
Sean Mooney
Moritz Juchler
Nikhil Gopal
OICR:
Robin Haw
UC Berkeley:
Chris Mungall
UCSD:
Trish Whetzel
MyVariant.info MyGene.info