transmart community meeting 5-7 nov 13 - session 3: transmart and the one mind for research data...
Post on 19-Jan-2015
264 Views
Preview:
DESCRIPTION
TRANSCRIPT
www.1mind4research.org
An Open-Science Framework• Generate and Share Data:
– Use a portal system to access global coherent datasets – Build an inter-operable infrastructure for collation, curation and hosting of
datasets for the R&D community– Common standards and methods for the collection of the world’s largest
biomarker repositories for brain disorders
• Generate and Share Tools:– Collectively develop next-generation animal models, imaging tools, stem cell
lines, biosensors, and more…
• Generate and Share Models: – Apply advanced clinical informatics, computational biology and visualization
tools to create dynamic interactive Disease Maps and inform Clinical Decisions Support and Practice Guidelines
• Create Insight: – Tap Global Expertise (multi-disciplinary & global) – crowd-sourced solutions
THE WAY FORWARD……….
THE CHALLENGE
DATA SHARING IS POORLY INCENTIVIZED
DATA AND TOOL SHARING IS NOT SIMPLE ENOUGH
???
DATA SETS ARE BALKANIZED AND HARD TO FIND
THE APOLLO PROGRAM
PORTAL ARCHITECTURE
8
Engagement Tools• Facebook-like Activity Stream• LinkedIn-like social networking• Patient Engagement platforms
Analytics Marketplace• Access to the world’s most advance Big Data Analytics Tools• API allows both Open-Source and Proprietary Software to Plug-In
Links Data – Experts - Analysis
Data Hub• OS Infrastructure for Comprehensive Data Management• Tiered Resource Ingestion Pipeline, DataSpace, Object Models,
Search, Web Services, Encyclopedias, Ontologies
CASE STUDY
tranSMART CASE STUDY: TRACK-TBIPrepared for ONE MIND
Sirimon O’Charoen, PhDManager, Translational Medicinesirimon.ocharoen@thomsonreuters.com
• ~1.7 Million Reported TBI’s in US / Year– ~52,000 Deaths ----------------------------------
– ~275,000 Severe TBIs (sTBIs) -------------
– ~1.37 M Mild TBIs (mTBIs)------------
– ?? M Unreported TBIs ---------------
• ~ 360,000 servicemen/veterans with war-related TBI
• BUT: TBIs can happen to anyone:– ~3.8 million sports related TBIs/yr in US (3.6 M unreported)
– ~300,000 vehicular TBI’s / yr in US (17% of TBIs)
– 35.2% of reported TBIs from falls
• TBI is a contributing factor to a third (30.5%) of all injury-related deaths in the United States.
Traumatic Brain Injury Numbers
11
3.4M
The Goal: Establishing the Natural History of the TBI Spectrum
12
Accelerate the understanding, both
clinically and mechanistically, of the natural course of TBI
Develop a new classification of
TBI
Identify predictive
measures of outcomes
Develop new diagnostic tools
Identify new targets for therapeutic intervention
Develop new treatments, preventions and cures and move the discoveries into clinical practice for
improved patient outcomes
In order to:
TRACK-TBI MULTICENTER INITIATIVE
• Transforming Research and Clinical Knowledge in Traumatic Brain Injury (TRACK-TBI) Pilot– Multicenter Implementation of the Common Data
Elements for Traumatic Brain Injury
• Aims– Test and validate Traumatic Brain Injury Common
Data Elements (TBI-CDEs)
– Collect neuroimaging, proteomic, and genetic biomarkers, clinical and outcome data
TRACK-TBI DATA
• Approximately 650 patients
• Mostly mild TBI
• Principle Investigators– UCSF, San Francisco
General Hospital• Geoffrey Manley, MD, PhD
– University of Pittsburgh Medical Center• David Okonkwo, MD, PhD
– University Medical Center Brackenridge• Alex Valadka, MD
– Mount Sinai Rehabilitation Center• Wayne Gordon, PhD
Yue JK, et al., J Neurotrauma, 2013 Jul 2.
TRACK-TBI DATA IN tranSMART
16
TRACK-TBI Data Curation Process Delivery in tranSMART
Ready for analysis- Exploratory analysis (tranSMART features)- Advanced workflow (tranSMART features)- Custom analysis (database access or R-
plugin)
QuesGen DatabaseSPSS
Collaborators
Goal: Curate/upload TRACK-TBI Data for data sharing and analysis at researcher’s fingertip
CDEs MappingData QC
ETL
tranSMART USER INTERFACE
• Data navigation: folder-structure• Analysis: drag & drop
• Trainings were provided for all collaborators• Same curated dataset can be accessed
without compressing and emailing files
ANALYSIS EXAMPLES
• 2 Publications focus on– Follow-up prediction
– GFAP-BDP as a diagnostic biomarker
• How do analyses in tranSMART confirm the published results?
• How quickly we can generate new hypotheses and perform exploratory analyses in tranSMART?– Data can be across domains (i.e. genotype/biomarker vs.
outcome, medical history vs. follow-up)
FOLLOW-UP PREDICTION
• Outcome follow-up is challenging*– 3-month GOS-E: 77%
– 6-month GOS-E: 69%
• Rate of successful follow-up significantly depends*– Years of Education
– History of Developmental Disorders
– Toxicology Lab Results at ED
– Homelessness
* Result from
FOLLOW-UP PREDICTION Lost-to-follow up
(6-month)Successful follow up
(69% at 6-month)
Years of Education
History of Developmental Disorders
LIVE DEMO
GFAP-BDP DIAGNOSTIC MARKER
• Glial fibrillary acidic protein (GFAP)– Acute plasma level
– Optimal cutoff 0.68 ng/ml*
• Discriminate*– CT intracranial lesion result
– Outcomes
• Other protein biomarkers may improve prediction
* Result from
GFAP-BDP DIAGNOSTIC MARKER:CT RESULT
Publication Figure 1Subset1: CT negativeSubset2: CT positive
GFAP-BDP DIAGNOSTIC MARKER:OUTCOME GOS-E
Publication Figure 5
GFAP-BDP DIAGNOSTIC MARKER: OPTIMAL CUTOFF
Subset1: GFAP<0.68 ng/mlSubset2: GFAP≥0.68 ng/ml
ACCOMPLISHMENTS
Pilot
• tranSMART implementation
• TRACK-TBI Pilot curation and upload
Evaluation
• User trainings (>25 users)
• Feedbacks• Validation of
publication results in seconds
On-going
• Data access for collaborators
• New data analysis
• Part of One Mind Portal
top related