Goals of data/meta-data exchange

• Understandable by humans
  
• Understandable by machines
  
• Analyzable in a single-study context
  
• Analyzable across multiple studies
  
• Re-purposeable beyond collection context

What is required of “Standards”?

• Domain/Application Relevance
  
• Semantic expressiveness
  • Concepts
  • Relationships/context
  • Layered representation
  • Information models  meta-data, relationships, context
  • Data or terminology models instance-level bindings
  • Data type models  complex semantics beyond “standard” data types

How are “relevant standards” developed?

• Bottom-up development ensures concrete instantiation and situational relevance
  
  • development ensures concrete instantiation and situational relevance
• Top-down development ensures maximum breadth and coherence
  
  • Usually very time-consuming
  • Often results in broader scope then required in many contexts
  • Often requires inordinate amount of time on “edge cases”
• Synthesis of the two approaches provides maximum practicality and reuse
  • Operationally difficult from methodology/tooling perspective

Barriers to cross-standard interoperability:

• • Differences in implementation technologies, e.g. proprietary databases
  
  
• Syntactic/Lexical
  • Differences in transport serialization, e.g. difficulty of combining XML schemas
  
  
• Semantic
  • Difficulty in computationally detecting “semantic equivalence”

Semantic Web Value Proposition

• Common representation between “design-time” and “run-time” (graph patterns)
  • Eliminates “impedance mismatch” that normally occurs between these two contexts
  
  
• Emphasis on “pure semantics”
  • Reduces technology barriers
  • SPARQL end-points (a broadly-applicable standard)
  • Reduces syntax and lexical barriers
  • Simplified model and syntax  higher level of abstraction  immunity to serialization brittleness

Semantic Web Value Proposition (2)

• Tools and infrastructure designed to promote reuse, discovery, amalgamation, harmonization of data and meta-data from multiple contexts
  • “Point-wise” semantic harmonization
  • Reuse made easy: “Interoperability emerges from laziness…”
  
  
• Significant amount of life sciences data already available
  • Linked Clinical Trials (http://linkedct.org)
  • Open Linked Data (http://linkeddata.org)
  • Linked Structured Product Labels (http://purl.org/net/nlprepository/linkedSPLs)

HL7 and CDISC

• Both organizations have developed standards that attempt to support data/meta-data
  • collection
    
  • submission
    
  • analysis
    
  • re-purposing
    
  • cross-study exchange

HL7 and CDISC Sandards

• However, both HL7 V3 messages or CDA documents and CDISC standards (SDTM or ODM) fall short of achieving the goals of standards for data exchange and re-purposing
  • At minimum, these standards currently have issues with XML serialization brittleness
    
  • limited cross-study utility 
    
  • need for “top-down” governance to explicitly specify representation and encoding choices that 
    
  • limit innovation and inherent “new study” variability

W3C Proposal

• A general approach to the use of multiple standards that collectively support standardized data/meta-data exchange and re-purposing
  • Standards-as-is*
    
  • Standards-in-context*
    
  • Interoperability across standards and the data collected using them*

*3-phase project plan of CDISC2RDF project

1. Standards-as-is

• Use Semantic Web standards to define standards for meta-data/data exchange
  • URIs for all elements in a given standard
  • Expressed using RDF and SKOS
  • Standards become addressable and queryable
  
  
• A substantial amount of the work has been done…
  • HL7 “O-RIM”
  • HL7 FHIR
  • BRIDG 3.2
  • SNOMED-CT
  • NCIt
  • CDISC2RDF

2. Standards-in-context

• Link Semantic Web expressions of one standard to those of another
  • Maximize cross-standard utilization based on semantic equivalences
    
  • Focus on cross-standard semantic incompatibilities as targets for semantic harmonization

3. Interoperability-across-standards

• Example: a “TermInfo” example
  http://www.w3.org/wiki/HCLS/ClinicalObservationsInteroperability/TermInfo

• W3C Proposal: Semantic Web-enabled standards and submission format  cross-trial focus on semantic equivalence  enable re-purposing
  
• Example: CDISC2RDF
  • RDF representations of all CDISC standards
  • schemas, models, and transform processes for CDISC content standards
  • completion by Q4 2012
  • ODM not yet included but sh/could be
  
  
• Example: HL7 FHIR
  • REST interface (default)
  • RDF expression of RIM bindings SPARQL endpoints  data interchange with other Semantic Web-enabled standards  CDISC SHARE, ODM, etc.
  
  

Summary

• A Semantic Web-based approach to standards development
  • All standards represented using Semantic Web standards
  • Supports collaborative (top-down) development and
  • Supports context-grounded (bottom-up) development
  
  
• A Semantic Web-based approach to standards use means:
  • Data/meta-data (including relationships) transmitted using Semantic Web standards
  
  
• Getting started
  • Meta-data represented using Semantic Web standards
  • e.g. represent ODM in RDF
  • Submit first study in RDF
  • The Easy Path has started…”Interoperability emerges from laziness…”