Problem: Genome Data Held in Silos, Unshared, not Standardized for Exchange

Transcription

1 Problem: Genome Data Held in Silos, Unshared, not Standardized for Exchange No one institute has enough on its own to make progress. Every researcher and clinician should be able to compare their genome data to others.

2 We need a public ledger for sharing GATTTATCTGCTCTCGTTG GAAGTACAAAATTCATTAAT GCTATGCACAAAATCTGTAG C TAGTGTCCCATCTATTT

3 Alliance for data sharing UC Santa Cruz Genomics Institute

4 Enabling Responsible Sharing of Genomic and Clinical Data GA4GH Founded on June 5, 2013 More than 400 insitu:onal members From more than 40 countries, Approx. 1/3 are companies Mission: to enable rapid progress in biomedicine Strategy: support major driver projects create and maintain interoperability of technology plajorm standards develop guidelines and harmonizing procedures for privacy and ethics in the interna:onal regulatory context engage stakeholders across sectors to encourage the responsible and voluntary sharing of data and of methods

5 Cancer Global Data Sharing Cancer is driven by muta8ons in DNA. Precision treatment of cancer depends on knowledge of these muta8ons Start with a Pilot: -build a mechanism for recording cancer DNA muta:ons and clinical informa:on from millions of cancer pa:ent par:cipants across the world -Ini:ally called the Ac:onable Cancer Genome Ini:a:ve -Cancer is the right place to start. Once this is working, similar technology could be used to share DNA informa:on for other diseases genomicsandhealth.org

6 Example proposed public record gene: BRAF variant: V600E Pa:ent ID: 163a fa-4705-bc\-d264c4cff796 Gender: Male Ethnicity: White Caucasian Age at Diagnosis: 57 Tumor Classifica:on: non-small-cell lung carcinoma (MeSH D002289) Tissue or organ of origin: Lung Tumor morphology: Squamous (epidermoid)

7 Data Sharing Specifica:ons Data open and available to all Ubiquitously accessible on the Internet Can scale to accept dona:ons from 1000s of sources Not maintained by any central authority or :ed to any single country, loca:on or ins:tu:on Not corrup:ble Protects par:cipant privacy Stable design so that it may be used by many 3 rd party applica:on programs ( apps )

8 This is accomplished with a Shared Public Ledger Ethereum: foundation Ripple: Hyperledger: hyperledger IBM Open BlockChain: MIT Enigma project enigma.media.mit.edu AirBnB (proposed):

9 Simplest Shared Public Ledger Record of transac8ons over :me Transac:on is adding informa:on to the database Special case: New informa:on marks previous informa:on as out-of-date It is only possible to add more transac:ons, no transac:on is ever erased or altered 1000s of copies of the ledger all over the world are kept in sync while addi:onal transac:ons come in from mul:ple sources by miners A shared public ledger keeps track of data provenance, i.e. when and how data was entered and updated, so users have the reputa:on/ reliability informa:on they need to filter out data they don t want

10 Ethereum Cancer KnowLedger Pilot Website at findpubs.org See:

11 Who will use the shared public ledger? Shared Public Ledger???? Data Users are: Professional Researchers Ci:zen Scien:sts Clinicians Developers of molecular dx, drugs, decision analysis tools Payers Pa:ents/par:cipants Regulatory agencies and treatment guideline organiza:ons

12 Where do the data come from???????? Ul:mately, all data comes from individual par8cipants who wish to share their gene:c and clinical informa:on for research or improvement of medicine??

13 How do data enter the ledger? Researchers Clinicians par:cipants? Shared Public Ledger Individuals Developers

14 A par:cipant works with a Trusted Steward to add their informa:on to the ledger

15 Possible Trusted Stewards Medical research ins:tu:ons (e.g. GENIE ins:tu:ons) Hospitals and clinics Pa:ent registry services and clinical trial recruitment organiza:ons Pa:ent agency advocate groups possibly providing service to allow pa:ents to maintain agency over data AND par:cipate in clinical trials (e.g. Sage Trust and Gene:c Alliance) Gene:c tes:ng companies All trusted stewards use the same somware (provided by GA4GH) to add informa:on to the ledger System is designed to support thousands of stewards globally

16 Trusted Stewards Data Users Medical Clinics Researchers par:cipants Pa:ent Advocacy Groups Pa:ent Registries Shared Public Ledger Clinicians Developers Tes:ng Companies Individuals

17 What goes into the public ledger and what stays with the steward? Public Ledger Par:cipant s gene:c variants in selected genes ~1 dozen broad, noniden:fying clinical features Steward s iden:ty and contact info Random numerical ID for par:cipant Steward Par:cipant personal iden:fying informa:on and staged consent Par:cipant s extended clinical and gene:c info Possibly iden:fying Par:cipant s instruc:ons for sharing addl info with qualified researchers w/o recontact Par:cipant s instruc:ons for recontact

18 Example: par:cipant -> public ledger Par:cipant visits doctor at a medical clinic Doctor orders gene:c test Doctor suggests data dona:on through steward (possibly her own ins:tu:on) Test results come back

19 Test Results

20 Par:cipant visits steward Steward records from the par:cipant: personal informa:on test results addi:onal gene:c and clinical data consent to donate to database addi:onal sharing and recontact preferences Steward appends par:cipant s publicly sharable gene:c/clinical data to public ledger

21 Trusted Stewards Data Users Medical Clinics Researchers par:cipants Pa:ent Advocacy Groups Pa:ent Registries Shared Public Ledger Clinicians Developers Tes:ng Companies Individuals

22 Recontact Data user discovers muta:ons of interest by using 3 rd party app on the public ledger and wants more informa:on about the par:cipant that provided in the ledger Data user contacts par:cipant s steward Steward has info about under what circumstances the par:cipant will share addi:onal data or agrees to be recontacted As appropriate, steward will supply addi:onal informa:on to data user or set up contact between user and par:cipant

23 Trusted Stewards Data Users Medical Clinics Third-Party App Researchers par:cipants Pa:ent Advocacy Groups Pa:ent Registries Shared Public Ledger Clinicians Developers Tes:ng Companies Individuals

24 How do two stewards know if they have a par:cipant in common? On the public ledger, a par:cipant is iden:fied with a random number Each steward securely stores personal iden:fiable informa:on for their par:cipants; only they can associate par:cipant with a public random number Personal iden:fiable informa:on is compared between the two stewards without revealing any informa:on except which par:cipants they have in common by a cryptographic trick (secure mul:party computa:on) To make it comparable, personal information could be collected by the NIH NDAR GUID standard

25 Demo: secure mul:party computa:on to compute private set intersec:on Internet Homomorphic encryption Steward A 100,000 participants 10 overlap with B Steward B 100,000 participants 10 overlap with A Neither server sees the personal information on the other. Only the checksum identifying the participants in common is visible, nothing else Runtime: 10 seconds over a transatlantic link, single CPU Implementation and experiments by Max Haeussler

26 Who can be a steward? Any en:ty that: Has a legi:mate permanent contact Follows the rules Has enough par:cipants to prevent par:cipant reiden:fica:on by steward (small stewards can be anonymously pooled) All stewards will have Internet ra:ngs; these can be available on a ledger (e.g. like AirBnB); users can filter out unreliable steward data

27 Who pays for all this? System can be designed and implemented for a few million dollars; long term maintenance is the only issue Governments or philanthropies (possibly associated with hospitals, pa:ent advocacy groups, etc.) could supply general funding Taxes on gene:c tests could provide revenue Stewards can be mo:vated to secure data dona:ons either by altruism or commissions from data users 3 rd party app developers can charge for use of their tools or sell adver:sing to support their efforts and to support the public ledger

28 Summary A completely decentralized, public database is possible, while s:ll protec:ng privacy We need this because trust is local No single state government or private organiza:on can/should own or control all the world s gene:c data Once launched, a shared public ledger grows and is maintained organically by the global community because it benefits them, much like the Internet itself