Transcript Pharmaceutical R&D and the role of semantics in information

Pharmaceutical R&D and the role
of semantics in information
management and decisionmaking
Otto Ritter
AstraZeneca R&D Boston
W3C Workshop on
Semantic Web for Life Sciences
27-28 October, 2004
Drug R&D – complex, costly & risky
information-driven enterprise
$$
Target ID
Target Val.
Biology
Screening Optimize
Chemistry
Pre-clinical Clinical
Development
2
~ 10 years
~ $1B
odds < 1/1000
Reality vs. Ideal State
3
Project vs. Business Perspectives
Challenge
Business
Knowledge
Problem
Knowledge
Scientific
Knowledge
Technological
Knowledge
uncertainty
B
benefit
A
C
cost
4
Many Maps, Models, Mappings
functional
& structural
spaces
conceptual
categories
INDIVIDUAL
ENTITY
context
attributes
(some context-dependent)
5
models
Heterosemantic Networks and Decision
Support

Find optimal routes
between entities, based on
evidence

Extend evidence-based
routes with technological
options (cost, risk)

Extend optimal plans,
based on science and
technology, into a lattice of
business options (real
options valuation)
6
From Molecular and Biomedical Information
Pathways to “R&D Pathways”
 Typical project routes
 Time, cost, attrition &
transition probabilities
 Model fitting for different
contexts (e.g., disease area,
target or lead molecular class,
…)
 Simulation, ranking of options
 Joint portfolio & infrastructure
optimization
7
Where we need (semantic and syntactic)
information integration
 Problem statement
… definition
 Representation
… language, formalism
 Integration/Implementation
… data, methods
 Modeling
… model, theory
 Evaluation of
… confidence feasibility
 Simulation of
… answers consequences
 Analysis
… options, conclusions
 Interpretation
… reference to reality
 Decisions
… impact on reality
8
Lessons learned so far
 Decouple form (syntax) from meaning (semantics)
 Allow for multiple interpretations & conflicts
 Reuse generic (form-oriented) components
 Operational definition for identity
 Explicit representation of context
 Decision support analysis presents a special case of
intelligent information integration across the science,
technology and business domains
9
Needs & Opportunities
 Large-scale and high-throughput data integration,
mining, model building and verification, interpretation &
reasoning over complex, dynamic, hetero-semantic
domains
 “Workflows of workflows”, driven by the meaning,
sensitive to context, and smart about uncertainty
 Stack of high-level declarative languages. Orthogonal
representations of concepts, logical and physical
structure, UI services and views (extension of the
Model-View-Control paradigm)
10