1 of 21

Repurpos.us: A fully open and expandable drug repurposing portal

https://commons.wikimedia.org/wiki/File:Wikidata-logo-en.svg

Sebastian Burgstaller-Muehlbacher, PhD

​

@sebotic

2 of 21

Drug repurposing

Promises:

  • Can bring down drug development cost
  • Can speed up drug development process

​

Prominent examples:

  • Sildenafil
  • Thalidomide
  • Avastin/Lucentis

​

​

Source: Scientific American

3 of 21

Drug Repositioning Approaches

Experimental:

  • Screens/trials with approved drugs or well examined compounds

Computational:

  • Compound similarity
  • Disease symptom similarity
  • Drug side effect similarity
  • Gene expression and pathway analysis
  • Metapath analysis and other network based methods
  • ...

​

​

4 of 21

Drug Repurposing Data Challenges

  • Many computational results are not being published, although the algorithm is.
  • Recapitulating computational results is hard, because data not open/accessible, e.g.:
    • Electronic health records
    • Certain databases
    • Full text papers
  • Computational results don’t make it into public databases, stay in the supplements of scientific publications.

​

5 of 21

"By 2029, computers will have human-level intelligence"

�-- Ray Kurzweil, Director of Engineering @ Google & Futurist

https://www.facebook.com/SXSWFestival/videos/10154414699178994/

6 of 21

Integrating computational predictions with existing data

  • Use the open graph database Wikidata for high quality data
    • Compounds, diseases, targets as stable nodes
    • Interactions, e.g. drug indications as stable edges
  • Inject data from computational predictions into the Wikidata namespace
    • Map nodes of predicted data to Wikidata nodes
    • Introduce predicted edges into new graph

​

​

Allows us to integrate computationally predicted drug repositioning data with high quality biomedical data!

​

7 of 21

8 of 21

Wikidata as core infrastructure

X

Your app

9 of 21

Wikidata for drug repurposing

10 of 21

SPARQL endpoint namespace extension for computational predictions

11 of 21

Compound space for Wikidata and drug repositioning

ChEBI

Currently, 156K unique chemical compounds and drugs + 6.9K diseases from Disease Ontology

12 of 21

13 of 21

14 of 21

15 of 21

Same compound, including predictions from Gottlieb et al, 2011

16 of 21

Summary

  • Developed a strategy for storing and reusing computational drug repositioning data.
  • Integrated with the rich and high quality data on compounds, targets and diseases which we had created in Wikidata.
  • Developed a prototype web-portal (http://repurpose.us) which allows user interaction with the data for further drug repositioning efforts.
  • Most importantly: We can store experimental and computational results in our framework.

​

17 of 21

Acknowledgments

​

Andrew Su

Benjamin Good

Gregg Stupp

Tim Putman

Julia Turner

(TSRI)�

​

​

​

Gang Fu

Evan Bolton

(NIH, PubChem)

Andra Waagmeester�(Micelio.be)

Elvira Mitraka

Lynn Schriml

(Disease Ontology, U Baltimore)

​

18 of 21

A Wikidata statement

https://commons.wikimedia.org/wiki/File:Wikidata_statement.svg

19 of 21

Wikidata

Two types of entities:

​

    • About 3,500 edge types (aka Properties)

​

    • Currently ~ 28 million nodes (aka Items)

​

​

​

Nodes and edges can both be populated with statements describing their nature.

​

20 of 21

Unique Features of Wikidata

  • Completely free, even for commercial usage (CC0).
  • Granular: Single values with references.
  • Everybody can contribute.
  • Single value item history.
  • A repository for data on all domains of knowledge.
  • Full integration with the semantic web.
  • Essentially: A giant graph of knowledge.

21 of 21

A Wikidata statement