1 of 10

Gene Wiki Project

Gregory Stupp, Andra Waagmeester

Andrew Su Lab

The Scripps Research Institute

San Diego, CA, USA

2 of 10

wikicite

MyGene.info

3 of 10

4 of 10

What drugs interact with proteins whose genes are associated with prostate cancer?

Physically interacts with

Encoded by

Genetic association

Metformin

Solute carrier family 22 member 3

SLC22A3

Prostate Cancer

https://doi.org/10.1136/amiajnl-2014-002649

Actively being studied!

http://tinyurl.com/m74tnxq

5 of 10

Curated Knowledge

Crowd-sourced Knowledge

Wikidata Crowdsourcing Model for Enhancing DO Curation

Add xref: DOID:0060177 -> MESH:C535328

curation

6 of 10

Gene Wiki Future (Use Cases)

  • Wikibase for non-CC0 data
  • Wikibase for research data
    • Non-Clinical data, experimental data, text-mined data
    • Data (that is not "knowledge") that is not appropriate for wikidata
  • Wikibase for staging, development, testing (shex) bots
  • WDQS with less restrictions on
    • Federated queries
    • Time outs

7 of 10

Implementation (problems and opportunities)

  • Wikibase configuration customization
    • How do i change URIs, ports, etc.
  • Wikidata Integrator support
  • Syncing local properties with properties from another Wikibase
    • Some sort of query interface wrapper/helper that can simplify the property -> skos:exactMatch lookups across wikibases to facilitate writing more legible federated queries
  • Going from wikidata to external wikibases
  • Scalability:
    • Can it handle a billion edges?
    • Multiple synchronous queries?
  • User/data authentication (private data)
  • ...

8 of 10

Lightning Talk: WDI

  • Fixing WDI to work correctly with Wikibase
    • Changing expected URIs
    • Fix fast run mode
    • Fix sparql queries
  • https://github.com/SuLab/WikidataIntegrator

9 of 10

Lightning Talk: Syncing local properties with properties from another Wikibase

How do we link PIDs and QIDs that are the same between Wikidata and local wikibase?

One solution:

10 of 10

Test Implementation: SemmedDB

Text-mined biomedical edges from all of Pubmed Abstracts

  • Scaled down (normalized, grouped version):
    • 200k nodes
    • 10 million edges