Enabling simultaneous analysis of multiple cohort studies without accessing the full

dataset: a BRISSKit use case

Dr Jonathan Tedds jat26@le.ac.uk @jtedds

Senior Research Fellow, Health Informatics & Interdisciplinary Research Group,Department of Health Sciences (University of Leicester)

PI #BRISSKit http://www.brisskit.le.ac.uk

http://www.astrogrid.org(April 2008 1st public release)

Data Reuse: asking new questions

Hubble Space Telescope• Papers based upon reuse of archived observations now exceed those based on the use

described in the original proposal.– http://archive.stsci.edu/hst/bibliography/pubstat.html

• See also work by Piwowar & Vision re life sciences: “Data reuse and the open data citation advantage”– http://peerj.com/preprints/1/

Science as an Open Enterprise ReportWhy open?• As a first step towards this intelligent openness,

data that underpin a journal article should be made concurrently available in an accessible database

• We are now on the brink of an achievable aim: for all science literature to be online, for all of the data to be online and for the two to be interoperable. [p.7]

• Royal Society June 2012, Science as an Open Enterprise, http://royalsociety.org/policy/projects/science-public-enterprise/report/

• Issues linking data to the scientific record:– Data persistence– Data and metadata quality– Attribution and credit for data producers

• Geoffrey Boulton (Edinburgh), Lead author:– “Science has been sleepwalking into crisis of

replicability...and of the credibility of science”– “Publishing articles without making the data

available is scientific malpractice”

BRISSKit context: The I4Health goal of applying knowledge engineering to close the

‘ICT gap’ between research and healthcare (Beck, T. et al 2012)

Biomedical Research Infrastructure Software Service Kit

A vision for cloud-based open source research applications #BRISSKit

http://www.brisskit.le.ac.uk

http://www.brisskit.le.ac.uk

BRISSKit USPs Integrated support for core research processes

Well-established mature open source applications as protoyped in Cardiovascular, Respiratory, Cancer Theme Biobank: UK customised

A platform for seamless management and integration between applications

An API allows integration with existing clinical systems

Easy set up, use and administration through browser (including on mobile devices)

Capability of being hosted in any compliant cloud provider including UHL (NHS information governance)

www.brisskit.le.ac.uk Email: brisskit@le.ac.uk

BRISSKit Community & Hack Event, Oct 2012

http://www.brisskit.le.ac.uk/node/35

BRISSKit Information Governance& Security Management Work Stream

- Dr Andrew Burnham leading

1. Information Governance Toolkit - analysis of Department of Health (DoH/NHS) IGT requirements vs. BRISSKit organisation/project and services/toolsa) Hosted Secondary Use Team/project (Hosted IGT)b) Acute Trust (Acute Trust IGT)

2. IG Training Tool (NHS – University is registered)

3. Pseudonymisation requirements

4. Data Management Plan

5. IT Security & standards – Penetration Testing & Security Testing

6. Other NHS Standards/Requirements:- Care Records Guarantee- NHS Constitution- NHS Records Management- Patient Safety DSCN 14/2009, 18/2009

The semantic bridge

?

OBiBa Onyx

Records participantconsent, questionnairedata and primaryspecimen IDs

i2b2

Cohort selection and data querying

Bio-ontology!

• Deploy solutions in international bio banking initiatives

• Investment through Prof Paul Burton (Health Sciences at Leicester/Bristol) & international collaborations

• Building on strong informatics expertise at University of Leicester in partnership with the University Hospitals Leicester Trust• Cardiovascular, Respiratory & Lifestyle

BRUs• Cancer Theme Biobank• Genomics etc

BRISSKit and Bio Banking

Contemporary biobanking: meeting the “data” challenge

Large data sets, why bother?

• Sample size• Depth of phenotyping• Quality of measurementAll critical

How big is BIG?• The direct effect of a gene• 2,000 cases minimum, 10,000 cases better

• Environmental and life-style factors• Highly context specific: from hundreds to tens of

thousands of cases• Gene-lifestyle and gene-gene “interactions”• Absolute minimum 10,000, usually need at least

20,000, a comprehensive platform needs at least 50,000• Scientifically fundamental

The bottom line Effective data access is crucial Effective joint analysis is essential too (integration) Fundamental challenges• Scientific harmonization• Restriction on access to individual level data• Streamlined access to multiple data sets

Central to the integrative aims of P3G, PHOEBE, BioSHaRE-eu etc

Also fundamental to the aims of potential BRISSKit users

18

Horizontally partitioned data

Data

PREVENDPREVEND

1958BC1958BC

Data

KORAGENKORAGEN

Joint centralanalysis

Data

FINRISKFINRISK

Data

How can we undertake a full joint analysis using multiple data sources if the data cannot physically be pooled? Ethico-legal constraints Physical size of the data objects Intellectual property issues

DataSHIELD: a novel solutionTake analysis to data not data to analysis

One step analyses: simple

Iterative analyses: parallel processes linked together by entirely non-identifying summary statistics

Typically produces mathematically identical results to fitting a single model to all the data held in one pooled data set

Take analysis to data not data to analysis

One step analyses: simple

Iterative analyses: parallel processes linked together by entirely non-identifying summary statistics

Typically produces mathematically identical results to fitting a single model to all the data held in one pooled data set

R

R

R RAnalysis

Computer

Web services

Web servicesWeb services

Data computer OpalFinrisk

OpalFinrisk

OpalPrevend

OpalPrevend

Opal1958BC

Opal1958BC

Data computer Data computer

BioSHaREweb site

BioSHaREweb site

Web services

Horizontal DataSHIELD

R

R

R RAnalysis

Computer

Web services

Web servicesWeb services

Data computer OpalFinrisk

OpalFinrisk

OpalPrevend

OpalPrevend

Opal1958BC

Opal1958BC

Data computer Data computer

BioSHaREweb site

BioSHaREweb site

Web services

Horizontal DataSHIELD

Opal includes• DataSHIELD• DataSHaPER• Researcher ID

R

R

R RAnalysis

Computer

Web services

Web servicesWeb services

Data computer OpalFinrisk

OpalFinrisk

OpalPrevend

OpalPrevend

Opal1958BC

Opal1958BC

Data computer Data computer

Horizontal DataSHIELD

BioSHaREweb site

BioSHaREweb site

Web services

Work in progress:• Embed Opal in BRISSKit • ALSPAC• MRC e-HIRCS• +more…

Opal1958BC

Opal1958BC

Opal1958BC

Opal1958BC

BRISSKit gains• DataSHIELD• DataSHaPER• Researcher ID

Opal1958BC

Opal1958BC

Opal gains• Direct interface with more tools• I2B2 functionality• Potential for

enhanced user interface

BRISSKit gains• DataSHIELD• DataSHaPER• Researcher ID

Opal1958BC

Opal1958BC

BRISSKit gains• DataSHIELD• DataSHaPER• Researcher ID

Opal gains• Direct interface with more tools• I2B2 functionality• Potential for

enhanced user interface

Everybody gains• Enhanced combined functionality - better science• Bigger user group - greater portability• Greater potential to become a sustainable standard

Opal1958BC

Opal1958BC

BRISSKit gains• DataSHIELD• DataSHaPER• Researcher ID

Everybody gains• Enhanced combined functionality - better science• Bigger user group - greater portability• Greater potential to become a sustainable standard

Opal gains• Direct interface with more tools• I2B2 functionality• Potential for

enhanced user interface

Enhanced joint analysis with• Ethico-legal constraints e.g.US/Europe biobanks• Intellectual property issues e.g. H3AFRICA

The bottom line Effective data access is crucial Effective joint analysis is essential too (integration) Fundamental challenges• Scientific harmonization• Restriction on access to individual level data• Streamlined access to multiple data sets

Central to the integrative aims of P3G, PHOEBE, BioSHaRE-eu etc

Also fundamental to the aims of potential BRISSKit users

29