New deals on data - Generating open knowledge based on closed data

Talk hold at the "Blockchain for Science 2018" (Berlin).

Transcript

1. New deals on data –
   Generating open knowledge
   based on closed data
   Konrad U. Förstner
   ZB MED – Information Center for Life Sciences, Cologne, Germany &
   TH Köln, Cologne Germany
   November 5th, 2018, Blockchain for Science Con
   

   View Slide

2. Disclaimer
   I have no to connection to any of the companies
   that I will be metioned here.
   I present my perspective as a bioinformatician
   and open science enthusiast.
   https://www.flickr.com/photos/redjar/113823307/ – CC-BY by flickr user redjar
   

   View Slide

3. Open [data|source|*] should be the default in science.
   This is simply good scientific practice.
   https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
   

   View Slide

4. ... but ...
   https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
   

   View Slide

5. There are cases where privacy migh be
   a higher good than openess.
   Certain data should not be linked to individuals.
   https://commons.wikimedia.org/wiki/File:Masks_in_Venice.jpg CC-BY by Wikipedia user Rasevic
   

   View Slide

6. Behavioral data
   https://www.flickr.com/photos/andrikoolme/27729630943 – CC-BY by flickr user andrikoolme
   

   View Slide

7. Socio-economical status
   https://commons.wikimedia.org/wiki/File:Assorted_United_States_coins.jpg
   

   View Slide

8. Medical records
   https://de.wikipedia.org/wiki/Datei:Blood_pressure_measurement.jpg
   

   View Slide

9. Genome / Exome / SNPs
   https://de.wikipedia.org/wiki/Datei:DNA_human_male_chromosomes.gif
   

   View Slide

10. View Slide

11. Microbiome information
    https://de.wikipedia.org/wiki/Datei:DNA_human_male_chromosomes.gif
    

    View Slide

12. https://de.wikipedia.org/wiki/Datei:DNA_human_male_chromosomes.gif
    

    View Slide

13. Having access to the such data of a large popuplation
    would significantly help research and
    to extend our medicial knowledge.
    https://de.wikipedia.org/wiki/Datei:Crowd_at_Knebworth_House_-_Rolling_Stones_1976.jpg CC-BY by Wikimedia Commons Ibirapuera
    

    View Slide

14. On the other hand the data can be misused for
    systematic discrimination due to
    political, ideological and commercial interests.
    https://www.flickr.com/photos/22394551@N03/2226095398 CC-BY by flickr user viZZZual.com
    

    View Slide

15. We have moral dillemma.
    Protect individual rights
    or
    push the scientific progress.
    https://commons.wikimedia.org/wiki/File:Apothecary%27s_balance_with... CC-BY by Wikimedia Commons user Fæ
    

    View Slide

16. Similar dilemmata from other research domains
    • Financial data of organisations
    • Energy consumption recording of devices
    • Location data of vehicles
    https://commons.wikimedia.org/wiki/File:Apothecary%27s_balance_with... CC-BY by Wikimedia Commons user Fæ
    

    View Slide

17. How can this dillemma be solved?
    https://commons.wikimedia.org/wiki/File:Apothecary%27s_balance_with... CC-BY by Wikimedia Commons user Fæ
    

    View Slide

18. Can we use closed data to generate (open) knowledge?
    https://commons.wikimedia.org/wiki/File:Eiserne_Truhe_Museum_Senftenberg.jpg PD
    

    View Slide

19. Can we research based on black boxed data
    that is at least reproducible?
    https://commons.wikimedia.org/wiki/File:Eiserne_Truhe_Museum_Senftenberg.jpg PD
    

    View Slide

20. Can we train machine learning models on closed data?
    https://commons.wikimedia.org/wiki/File:Eiserne_Truhe_Museum_Senftenberg.jpg PD
    

    View Slide

21. Or can we at least use the data to generate hypthesis
    that then can be tested with complementary methods?
    https://commons.wikimedia.org/wiki/File:Eiserne_Truhe_Museum_Senftenberg.jpg PD
    

    View Slide

22. Genomics England
    • Aims to hold 100,000 full genomes
    • Data processing in closed data centers
    • Only results leave the center via an ”airlock”
    https://de.wikipedia.org/wiki/Datei:Crowd_at_Knebworth_House_-_Rolling_Stones_1976.jpg CC-BY by Wikimedia Commons Ibirapuera
    

    View Slide

23. Personal Health Train (PHT)
    • Data stations – (”FAIRports”)
    • Trains – Workflows that can work on the data
    provided to them
    https://de.wikipedia.org/wiki/Datei:Crowd_at_Knebworth_House_-_Rolling_Stones_1976.jpg CC-BY by Wikimedia Commons Ibirapuera
    

    View Slide

24. • Locked system
    • Trust of the platform required
    https://de.wikipedia.org/wiki/Datei:Crowd_at_Knebworth_House_-_Rolling_Stones_1976.jpg CC-BY by Wikimedia Commons Ibirapuera
    

    View Slide

25. (This slide was modified for online deposition - simply click on
    the link below; It is a news article that describes how 23andMe
    and other are selling genomic data to pharma industry.)
    https://www.businessinsider.de/dna-testing-delete-your-data-23andme-ancestry-2018-7
    

    View Slide

26. Promises of blockchain-based,
    decentralized data marketplaces
    • owners have control over their data and can stay
    anonymous
    • standardisation of data
    • people can be incentivized to share the data
    • traceability (especially for pharmaceutical
    companies interesting)
    https://www.flickr.com/photos/katerha/4592429363 – CC-BY by flick user katerha
    

    View Slide

27. Data market place
    Data
    consumer
    Access to data
    Token
    Data
    owners
    

    View Slide

28. Concepts of underlying solutions
    • Fully Homomorphic Encryption (FHE)
    • Multi-party Computation (MPC)
    • Trusted Execution Environment (TEE) like Intel SGX
    https://unsplash.com/@toddquackenbush?photo=IClZBVw5W5A - PD
    

    View Slide

29. Blockchain
    Data
    consumer
    Secure
    compute
    node
    Private
    data
    storage
    Data
    owner
    

    View Slide

30. General purpose blockchain-based solutions
    • Ocean protocol
    • Enigma (secret contracts)
    • Ekiden protocol (Oasis Labs)
    • OpenMind
    https://unsplash.com/@toddquackenbush?photo=IClZBVw5W5A - PD
    

    View Slide

31. Blockchain-based solutions for healthcare data
    • Nebula (by George Church)
    • Longenesis
    • Luna DNA
    • phrOS (Personal Health Record Operating System)
    • EncrypGen
    https://unsplash.com/@toddquackenbush?photo=IClZBVw5W5A - PD
    

    View Slide

32. Will these data market places improve science?
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

33. This has huge potential!
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

34. ... but ...
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

35. Currently lot of white papers available –
    nothing openly testable.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

36. Discussion unfortunately driven
    by companies not academics.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

37. High risk –
    you won’t get your genome back once it leaked.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

38. https://doi.org/10.1126/science.1229566
    

    View Slide

39. The suggested systems have very high complexity.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

40. Problem of different legal systems.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

41. Implications for data owner/seller might be not clear
    – education needed.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

42. Data stored off-chain
    =
    outsourcing of one important problem
    (suggestion like Dropbox metioned –
    IMO quite a bad idea)
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

43. How to avoid false statements in surveys
    to become interesting for data consumers?
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

44. Bottom line:
    Very promising,
    but a long and hard way to go.
    https://www.flickr.com/photos/subcircle/500995147 – CC-BY by flickr user subcircle
    

    View Slide

45. What are your questions?
    konrad.foerstner.org / @konradfoerstner
    zbmed.de / @ZB_MED
    th-koeln.de / @th_koeln
    https://www.flickr.com/photos/nateone/3768979925/ – CC-BY by flick user nateone
    

    View Slide