Upload
gully-burns
View
1.506
Download
0
Tags:
Embed Size (px)
DESCRIPTION
This is an introduction to a knowledge engineering methodology called 'Knowledge Engineering from Experimental Design' (KEfED). This methodology provides a powerful, intuitive method for modeling the design of scientific experiments and provides the foundation for work at the Biomedical Knowledge Engineering Group at the Information Sciences Institute (run by Gully Burns)
Citation preview
Knowledge Engineering from Experimental Design
‘KEfED’
Gully APC Burns
Information Sciences InstituteUniversity of Southern California
The Cycle of Scientific Investigation (‘CoSI’)
Knowledge Engineering from Experimental Design
A typical seminar slide
What is an elemental piece of biomedical scientific
knowledge?
For example...
What is an elemental piece of biomedical scientific
knowledge?
The challenge of defining the biomedical semantic
web• Currently consists of a very large number of statements like
‘mice like cheese’– semantics at this level are complicated!
• For example:– “Novel neurotrophic factor CDNF protects midbrain dopamine
neurons in vivo” [Lindholm et al 2007]– “Hippocampo-hypothalamic connections: origin in subicular
cortex, not ammon's horn.” [Swanson & Cowan 1975]– “Intravenous 2-deoxy-D-glucose injection rapidly elevates
levels of the phosphorylated forms of p44/42 mitogen-activated protein kinases (extracellularly regulated kinases 1/2) in rat hypothalamic parvicellular paraventricular neurons.” [Khan & Watts 2004]
• Statements vary in their levels of reliability, specificity. • Existing semantic web approaches involve representations
of argumentation / claim networks• Can we invent a new way to introduce formalism?
Knowledge Engineering from Experimental Design
(‘KEfED’)• There is an implicit reasoning model
employed by scientists to represent their observations based on the way they design experiments– Standardized experimental templates– Parameters [‘Independent Variables’] – Measurements [‘Dependent Variables’]– Calculations [‘Derived Variables’]
Basic KEfED Elements
Logical Element IconActivity
Experimental Object
Parameter
Measurement
Branch
Fork
Dependencies between variables are inherent in the experimental protocol
The KEfED Model is intuitive
KEfED handles complex experimental designs
More Below…
Khan et al. (2007), J. Neurosci. 27:7344-60 [expt 2]
KEfED handles complex designs
Khan et al. (2007), J. Neurosci. 27:7344-60 [expt 2]
Example : Neural Connectivity - Observations
‘anterograde’
‘retrograde’
Tract Tracing Experiments
Neuroanatomical experiments to study neural connectivity.
injection-site
tracer-chemical
labeling-location
labeling-density
labeling-type
Example : Neural Connectivity - Interpretations
Tract Tracing Experiments > Neuroanatomical Elements
Interpretative entities that correspond to facts that may be aggregated into a model
Neuronal Population
cell-bodies
cell-bodies.location
terminal-field.location
terminal-field
‘Neural Connection’
connection-origin
connection-termination
connection-strength
1st look at ‘BioScholar system’: Neural Connectivity Reasoning Tool
Peeking Under the Hood
‘PHAL Injection into SUBv generates labeling in MM’ => ‘SUBv contains neurons that project to MM’(expressed in First-Order-Logic within Powerloom Reasoner)
Computation based on the context of each measurement based on parameters
Crux
• KEfED as the basis for the design of a data repository
• Collaboration with MSU + Science Commons – Funded by MJFF + Kinetics Foundation
to manage data from grantees
• KEfED-editor can as a component in an external web-application
[http://yogo.msu.montana.edu/applications/crux.html]
Using Semantic Web Standards
[https://wiki.birncommunity.org:8443/display/NEWBIRNCC/KEfED+OWL+Model]
OBI
• Use a simplified ‘projection’ with no semantic entailments.
• Seek a simple model with semantics embedded ‘within’ variables
… work in progress here …
• Seek semantic-web-based links to:
– OBI– SWAN / SIOC– ISA-Tab tools
• Domain-specific Reasoning Models (from ‘CoSI’)
– Want to generate hypotheses / predictions that can be expressed as KEfED models?
– $6,000,000 question!
Future Directions
Acknowledgements
Funding– Information Sciences Institute,
seed funding – NIGMS (R01GM083871)– NIMH (R01MH079068)– NSF (#0849977) – Michael J Fox + Kinetics
Foundations – BIRN @ ISI
Neuroscience Team Members – Rick Thompson (USC)– Jessica Turner (MRN)
Neuroscience Contributors– Alan Watts (USC)– Larry Swanson (USC)– Arshad Khan (USC)
Computer Scientist Team – Tom Russ (ISI)– Cartic Ramakrishnan (ISI)– Marcelo Tallis (ISI)– Eduard Hovy (ISI)
Other Team members– Alan Ruttenberg
(ScienceCommons)– Michael Rogan (NYU)– Gwen Jacobs (MSU)– Pol Llovet (MSU)
Computer Scientist Contributors– Hans Chalupsky (ISI)– Jerry Hobbs (ISI)– Yolanda Gil (ISI)– Carl Kesselman (ISI)– Jose Luis Ambite (ISI)