biogrid platform - qctn marienne hibbert.pdf · life science e-research platform marienne e hibbert...

10
8/20/2008 1 Life science e-research platform Marienne E Hibbert BioGrid Australia Victorian and Australian Government investment First cross-institution, cross- discipline clinical research data integration platform (IBM) Nightly upload of data on-site 2 Nightly upload of data on site No identified health data leaves the site No health data stored centrally Individual records linked via linkage key BioGrid Platform Move from paper .. Data collected at source Clinical outcomes Li k d Bi i Linked Biospecimens, Genomics, Images, Death, Statewide data. Dynamic queries Privacy protected Population Patient Medical Informatics Public Health Informatics Biomedical Informatics: synthesizes knowledge at all levels Biomedical Informatics: synthesizes knowledge at all levels Tissue, organ Cell Molecule, gene Bioinformatics Medical Imaging Genome Epidemiology TITLE OF SLIDE Body copy here. Lorum ipsum id sub quanto cerberus. Collect Data ‘Clinical & Scientific Tools’ Publish And Present Collaborative Research Research Tools Discovery ‘Access Tools’ Analysis Virtual Data Store Why ? Research power: Increase the sample size Increase the potential for research collaborations Link specialist databases covering common diseases: Screening activity - Genetic predisposition Environmental exposures 6 Genomics, proteomics & epigenetics Co morbidities Quality and audit Treatment strategies Outcomes

Upload: dinhnga

Post on 05-Feb-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

1

Life science e-research platformMarienne E Hibbert

BioGrid Australia• Victorian and Australian

Government investment• First cross-institution, cross-

discipline clinical research data integration platform (IBM)

• Nightly upload of data on-site

2

Nightly upload of data on site• No identified health data leaves

the site• No health data stored centrally• Individual records linked via

linkage key

BioGrid Platform

Move from paper ..• Data collected at source• Clinical outcomes

Li k d Bi i• Linked – Biospecimens, Genomics, Images, Death, Statewide data.

• Dynamic queries• Privacy protected

Population

PatientMedicalInformatics

Public HealthInformatics

Biomedical Informatics:synthesizes knowledge at all levels

Biomedical Informatics:synthesizes knowledge at all levels

Tissue, organ

Cell

Molecule, geneBioinformatics

MedicalImaging

GenomeEpidemiology

TITLE OF SLIDE

Body copy here. Lorum ipsum id sub quanto cerberus.

Collect Data‘Clinical & Scientific

Tools’

PublishAnd

Present

CollaborativeResearch

Research ToolsDiscovery

‘Access Tools’Analysis

VirtualData

Store

Why ?• Research power:

– Increase the sample size– Increase the potential for research collaborations

• Link specialist databases covering common diseases:– Screening activity - Genetic predisposition– Environmental exposures

6

– Genomics, proteomics & epigenetics– Co morbidities– Quality and audit– Treatment strategies– Outcomes

Page 2: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

2

Victorian SitesMelbourneAustinWesternPeter MacAlfredSt Vincent’sMonashRCH

Other Australian statesNSW: St Vincent’s

POWTas: RHHAct: Canberra SA: Flinders

Royal AdelaideQueen ElizabethLyell McEwinRACs

Q l d

BioGrid

Scope - Collaborations

RCHRWHBox HillPeninsulaBarwon-tbc BallaratLatrobeBendigoHumeDHSUni of MelbourneWEHILICR

Queensland . . Western Australia ..

Researcher

Sites – International in progressUSA: Michigan

MoffittVanderbiltVenter

Brazil: Sao PauloNew Zealand

BioGrid

Cancer:• Colorectal• Brain• Breast• Lung• Sarcoma• Gynaecology• Prostate

Scope - Datasets Neuroscience:• Epilepsy• Neuropsychiatry• MS• Stroke

Diabetes• Type 1• Type 2

Researcher

• Prostate• Head & Neck• Upper GI• Melanoma• Renal• Prostate• Rare

Other - 1• IBD - Crohns• Cystic Fibrosis• Well womens• Population

Other - 2• Medicare - tbc• Deaths - tbc• DHS

BioGrid

Scope - AnalysisData:• Clinical outcome• Treatment• Genetic (Microarray, Biomarker, Proteomic)• Images• Biospecimen Banks

Researcher

Tools• Bioinformatics• Statistical• Drug Discovery• Image Analysis

High Performance

Computing

Medical Research - the Challenges

• Collection• Large amount of data• Lack of data standards• Lack of interoperability between databases

10

Lack of interoperability between databases• Need a cohesive approach between

disciplines• Ethics, Privacy and Regulation• Who owns the Intellectual Property

Medical Research - the Challenges

• Collection – at the source, clinically useful• Large amount of data – storage cheaper• Lack of data standards – mapping, agreements• Lack of interoperability between databases –

technology

11

technology • Need a cohesive approach between disciplines -

WIIFM• Ethics, Privacy and Regulation – Abide and use

technology• Who owns the Intellectual Property - Agreements

How?

Page 3: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

3

Enabling ResearchEnabling Research

Research• Prediction of appropriate therapy – genomics• Quality audits and feedback • Referral practices

Ch th ibi i t• Chemotherapy prescribing – appropriate dosing

• Testing extrapolation of clinical trials to clinical care

• Cost benefit studies – Colorectal Cancer screening

Diabetes Research questions

• Diabetes and managing its effects:– Investigate the progression of symptoms such as neuropathy

and eye disease in relation to blood glucose control using measurements taken over a 20 year period (Prof. Peter C l M lb H lth)Colman, Melbourne Health)

Diabetes : medical and economic cost of complications

Cardiovascularproblems– Heart– Stroke– PVD Diabetic

Diabeticretinopathy

neuropathyand foot problems

Renaldisease

Intervention StudiesIntervention Studies

MAS

SM

ASS

LOSS OF FIRST PHASE LOSS OF FIRST PHASE INSULIN RESPONSE INSULIN RESPONSE

MULTIPLE ANTIBODY POSITIVEMULTIPLE ANTIBODY POSITIVE

GENETICALLY ATGENETICALLY AT--RISKRISK

DYSGLYCEMIADYSGLYCEMIA

TIMETIME

BET

A C

ELL

MB

ETA

CEL

L M

DIABETES

“PRE”-DIABETES

GENETICPREDISPOSITION

INSULITISBETA CELL INJURY

NEWLY DIAGNOSED DIABETESNEWLY DIAGNOSED DIABETES

CC--Peptide Peptide ββ--cell masscell mass

DYSGLYCEMIADYSGLYCEMIA

Brain Research questions

• Identifying brain tumours– Linking PET images and MRI images to examine

the uptake of radio-labelled amino acids and their possible use as markers to identify brain tumours (Eddie Lau, Peter MacCallum Cancer Centre)

• Investigate brain changes in epilespy– Use of MRI images to investigate longitudinal

changes in the hippocampus of patients with Temporal Lobe Epilepsy (Dr. Sophie Adams, Melbourne Neuropsychiatry Centre)

Page 4: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

4

IF YOU WANTED TO COMPAREIF YOU WANTED TO COMPAREthe volume and shape of the the volume and shape of the

brain of individualbrain of individual

•• EpilepsyEpilepsy•• Dementia Dementia

where would you start?where would you start?

Epilepsy Research questions• Targeting the best drug in epilepsy based on

genetics– Pharmaco-genetic study the effectiveness and side

effects of epilepsy drugs based on a patient’s genetic profile, with the long-term goal of “personalised prescribing” for each patient (Dr. Terry O’Brien and Slave Petrovski, Melbourne University and yMelbourne Health)

SNPsSNPsSingle nucleotide polymorphism.

A DNA sequence variation in a single nucleotide (A/T/G/C) in the genome.

Make up 90% of all human genetic variation (HGP).

SNPs are a great value for biomedical research and for developing pharmaceutical products/medical diagnostics

Epilepsy Drug Therapy

BioGrid allows us to:• Assess genetic markers to refine

d t t t d i i i ADR

Refined by Genetic Knowledge

Key Study Findings:(Petrovski, O’Brien, Sheffield, 2006)

“In 40% of Epilepsy patientstreatment is limited by adverse drug reactions (ADR)”

drug treatment and minimize ADR

• Examine predictors of non-ideal outcomes (psychosocial and seizure recurrence) following surgery for medically refractory epilepsy

• Assess changed neuro cognitive function in newly treated epilepsy patients, and the relationship to genetic variability

Those with CC genotype were 5 times more likely to be resistant to the anti-epilepsy drug (CBZ), and 31% of the CC group were resistant.Those with TT genotype were 4 times more likely to have an adverse reaction to CBZ, and 33% of the TT group had an adverse drug reaction.

Clinical + GenotypicClinical + Genotypic Colorectal Cancer researchPeter GibbsSuzy KosmiderKathryn Field

Julie JohnsNgio MuriguDaniel Compston

Page 5: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

5

Colorectal Cancer question - drug development• BRAF mutation in 10 – 15% these tumours• Developing BRAF inhibitor drug - $ $ $

How effective is a BRAF inhibitor drug inHow effective is a BRAF inhibitor drug in reducing the cancer growth?

• Target the patient group - only possible by linking tumour and clinical databases

Colorectal Cancer Lymph node yieldLarge studies already carried out internationally

Staging for colon cancer

Lymph nodes harvested in patients at a number of BioGrid sites

Median LNY

Median LNY - by age

05

101520

<2121-30

31-4041-50

51-6061-70

71-8081-90

91-100

Age

Med

ian

LNY

0

2

4

6

8

10

12

14

16

1988

199019

921994

1996

199820

0020

0220

0420

0620

08

Year of diagnosis

Med

ian

LNY

Quick reponses …

Journal of the National Cancer Institute

"Re: Residual Treatment Disparities After Oncology Referral for Rectal Cancer"

Field K, Kosmider S, Desai J, Lim L, Barnett F, McLaughlin S, Jones I, Gibbs P

Page 6: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

6

•1992-1999 SEER data

•2716 patients (1931 excluded)

•Patients ≥ 66 yo

•Resection of stage II or III rectal cancer

•Looked at disparities between black and white in receipt of adjuvant chemotherapy

•Younger, fitter black patients LESS likely to receive chemotherapy

Tumor site N Any therapy* recommended, n(%)

Any therapy declined, n(%)

Preop. therapy recommended, n(%)

Preop. therapy declined, n(%)

Colon

<60yo

645

148

313 (48.5)

103 (69.6)

40 (12.8)

5 (4.9)

n/a

n/a

n/a

n/a

Physician recommendation and patient acceptance of adjuvant therapy for stage II and III colorectal cancer according to tumor site, patient age, and timing of therapy

60yo60-75yo>75yo

148305192

103 (69.6)165 (54.1)45 (23.4)

5 (4.9)21 (12.7)13 (31.1)

n/an/an/a

n/an/an/a

Rectum 292 217 (74.3) 9 (4.1) 133 (45.5) 2 (1.5)

•4 hospitals

•Jan 2003 – Feb 2008

Clinical practiseIssues with clinical trials

Patients treated in trials are– Younger (on average 10 years)– Fitter (better performance status)– Fewer co-morbidities– And have normal organ function

than patients in routine clinical practice

Can we extrapolate trial results to the clinic?• Are trial patients the same patient

population as those treated in clinic?

• Can trial results and importantly the observed toxicities be inferred for those treated in clinic?

Trial•X-ACT Study

XELODA

5FUStage 3 colon cancer

(node positive)

–2000pts

Reality• WH and RMH (200pts)• Xeloda • Median age:g

– X-ACT study 62yo (80% pts under 70)– WH and MH 75 years

• Patients requiring a dose reduction– 42% on trial– 79% in reality

Page 7: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

7

Can we extrapolate trial results to the clinic?

• THE ANSWER IS NO• THE ANSWER IS NO

• New prognostic markers for patients undergoing surgery for colorectal cancer are urgently sought

• 378 patients with potentially curative resection of colon cancer between January 2003 and August y g2007.

• Median follow-up was 20.5 months.

Annals of Surgery, Jan 2008

Albumin and Overall Survival Diabetes & Cancer Research question

Diabetes and its impact on colorectal cancer– Investigation of the effect of diabetes on survival and

response to chemo-therapy in patients with colorectal cancer by linking BioGrid’s diabetes and colorectal cancer databases. (Dr. Suzanne Kosmider, BioGrid)

Overall Survival: Diabetes vs No-Diabetes

50% of diabetes patients died by 78 months

Not reached for non-diabetic patients

• Diabetes is a strong risk factor for the development of Colorectal Cancer

• The expected incidence of diabetes would be <20% - in our series 29%

• Poorer survival for diabetes patients with Colorectal Cancer is likely multi-factorial

Page 8: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

8

How is BioGrid novel?Without BioGrid steps With BioGrid : steps1. Determine what data you want to retrieve 1. Determine what data you want to retrieve

2. Locate that data in the various databases 2. Obtain authorization and permission

3. Obtain permission from data owners to access data including identifying

3. Construct a query from the data model to extract the dataaccess data including identifying

information - MOUsextract the data

4. Obtain ethics committee approval which may include more than one site

4. Run the query

5. Run queries on those datasets often at one point in time

5. Refresh data/query when required

6. Manually join the data on various identified demographic data

7. For public domain data, run each data element individually to retrieve the data

SummaryData entered into source system:• Patient details

• Diagnosis

• Treatments : Chemo, Radiotherapy etc.

• Treatment Outcomes

Data used in Clinical care:• Review notes

• Plan treatment

• Communication

SummaryData in de-identified form• Research

• Quality and audit

Linked records

Translate research into improving patient care

Business and admin

Robert Merriel (Chair of Committee)

Marienne Hibbert (Project Director) 

Richard Tate

Vicki Vlekkert

The BioGrid Team

Cancer

Peter Gibbs (CSO)

Jayesh Desai

Suzy Kosmider

Kathryn Field

Julie Johns

N i M iTechnical

Ngio Murigu

Sandy Dupuis

Daniel Compston

Naomi Rafael

Kee Ming Kong

Pranabh Jain

Xiaobin Shen

Nelson Wu

Jana Graenz

Life Sciences

Henry Gasko

Diana Salim

WEHI

[email protected]

Page 9: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

9

Internet

Query

Authorised researchers query the Federated Data Repository for

analysis. They can only view the USI and health data.

Data is loadedinto institute-specific

Local Research Repository nightly

Eg. LRH

De-identified Linked data

BioGrid Linkage key ModelBioGrid Linkage key ModelProbabilistic matchingProbabilistic matching

LRR

FDI

USI

Health Data

BioGrid

Q yIdentifiers

Encrypted USI

Health Data

Breast DB

Identifiers

Health Data

Linkage Server

Identifiers

USI

The identifiers are sent to the Linkageserver where a key is allocated and

sent back to the LRR. The key in the LRR is encrypted for extra security

Patients are allocateda USI using 6 identifiers to

match with existing patients

Hospital IT Services (MH WH NH)

Institution basedLocal Research

Repository

Institution basedsource data

LRR

Data Source ETL

ETL Tools

Databases:• SQL Server

Technology – robust options

d A

rchi

tect

ure

• SQL Server• IBM DB2• Access• OracleFiles:• MS Excel• Other flat files

• IBM Ascential Datastage• IBM DB2 Warehouse Center• Microsoft SQL Server SSIS

• IBM DB2 UDB• Microsoft SQL Server• Oracle

Bio

Gri

•Sun Java CAPS Integration Suite eIndex•Oracle 9iUSI

Metadata DiscoveryServer

XMETA

Data Discovery

Data Discovery and

USI Server

• IBM DB2• IBM Websphere Business Glossary

Data Analytics Server

titut

ion

Opt

ions

Current BioGrid Toolset

Federated Data Integration Server

FDIDB

USIDB

• SAS Enterprise Business Intelligence Server

• SAS Web Report Studio

• IBM Websphere Business Glossary• SAS Enterprise Guide

yQuery Server

Researcher desktop

MS Internet Explorer v6 sp1

• IBM DB2• IBM Websphere Information Integrator

Parti

cipa

ting

Ins

Who ?

Federated D t

Internet

TermsGlossary

QueriesAnalysis

RCH/ MCRI•Cystic Fibrosis•Diabetes•Crohns

Authorised researchers and applications query the Federated Data Repository for analysis.

de-identified

PeterMac•Oncology•Biospecimens•PET images

Alfred•Cystic Fibrosis•Neuroscience•Oncology

Austin

St Vincents•Neuroscience•Diabetes•Oncology

Monash MC•Oncology•Biospecimens•Cystic Fibrosis

Box Hill•Oncology

TasmaniaRHH•Diabetes•Oncology

South AustR Adelaide HQ Elizabeth HFlinders MCLMMC•Oncology•Cystic Fibrosis

ACTCanberra •Oncology

NSWSt VincentsPOW

RWH•Oncology•Diabetes

Bendigo•Oncology

Gippsland•Oncology

Grampians•Oncology

Hume•Oncology

Barwon tbc•Oncology Data

IntegratorVPN - each site

GenBank

UniProt

PubMed

LocusLink

Public Data Sources

AnalysisReports

BioGrid AustraliaHealth through information

Linkage Keys

dataAustin•Oncology•Biospecimens•Diabetes

Western•Oncology•Biospecimens

POW•Oncology•Tissue Bank

QueenslandBrisbane tbc•Oncology•Epilepsy

Data cached on computers at the sites - owned and controlled by site.

DHS•VAED etc

Cabrini Oncology

Oncology

Peninsula•Oncology•Diabetes

Epworth tbc•Oncology AIHW @ MH

• Death + cause

International sites

Melbourne Oncology•Biospecimens•Diabetes•Neuroscience •MRI Images

Governance

Foundation members invited

BioGrid Unincorporated Joint Venture

BioGrid Service

Service Level Agreement including IP Li d

JVA Management Committee Company Board

Appointment inter-relationships as specified in BG Constitution

• Ludwig Institute for Cancer Research• WEHI• Melbourne University • Melbourne Health• Austin Health• Bayside Health• Peter MacCallum Cancer Institute• Western Health• Eastern Health• Southern Health• St Vincent’s Health

Service Company

(Not for ProfitProprietary Company Limited by Guarantee)

License and specification of services to be provided by Service Company to JVA Members

Constitution

NewJoining Member

Simplified Deed of Accession

New Joint Venture Agreementbased upon Collaboration Agreement

Constitution to be approved by

Joint Venture

BioGridService Company – Key Relationship Agreements•

BioGrid Aust

Company Board

New Research Collaborator

Services e.g. For profit company

Project Agreement

Collaborative Research and Development Agreement

If additional services

As required for different

projects

Services provided in accordance with JVA and Service Level Agreement between JVA and BioGrid S i C

54

oG d ustService Company

Constitution

Melbourne Health (Host Organisation)

Project Services Agreement

New Data Provider

New Client

Data Transfer Agreement

Client Services* Agreement

* For a New Client, only applies to additional services not included in standard Collaborative

Research and Development Agreement.

services requiredJVA

MemberService Company

Page 10: BioGrid Platform - QCTN Marienne Hibbert.pdf · Life science e-research platform Marienne E Hibbert BioGrid Australia ... – Increase the sample size ... Federated Data Repository

8/20/2008

10

Approval ProcessResearcher request: Specify •data required•Science of project•Agree to terms and conditions

Custodian Authorisation

Must approve data requested for the project

Management Committee: Must approve• data requested • the project• the researcher

Database administrator

Creates logon for: •Specific data

Reporting and auditingof all queries on the system:• of all users• to all databases

Reports: all queries on the system:• To Management Committee• To Ethics committees as required.• To data custodians as required.

ONLINE ACCESS: to integrated clinical data on patients with cancer (all 12 tumours), diabetes, cystic fibrosis, stroke, asthma and epilepsy from more than 120,000 patients and 1,900,000 clinical records with up to 25 years clinical history.DATA QUALITY: BioGrid systematically collects clinical data from over 30 major

Data at the click(s) of a mouse

hospitals in Victoria and nationally with information such as dates of diagnosis, clinical symptoms, pathology, radiology reports, biospecimen samples, genomic data, MRI images, treatments: drug therapy, radiotherapy and surgical status in a privacy protected way for authorised users.DATA LINKAGE: Once the researcher receives proper authorization to access the BioGrid system, the clinical data from different databases could be analysed across different institutions and disease types as per researcher needs. Questions such as: “What is the probability of a patient with colon cancer developing TYPE II Diabetes?” could be answered in minutes. DATA FLEXIBILITY: The data is updated every night and can be accessed 24/7 via an online portal. It is also available to be easily exported to statistical software of your choice such as Excel, SPSS, etc. For more complicated queries, expert BioGrid staff is available to provide data mining and analysis support.