How do we know what we don’t know: Using the Neuroscience Information Framework to reveal knowledge gaps
1. How do we know what we don’tHow do we know what we don’t
know: Using the Neuroscienceknow: Using the Neuroscience
Information Framework to revealInformation Framework to reveal
knowledge gapsknowledge gaps
Maryann E. Martone, Ph. D.
University of California, San Diego
Tools for Integrating and Planning Experiments in
Neuroscience-UCLA March 11, 2014
2. We say this to each other all the time,
but we set up systems for scholarly
advancement and communication that
are the antithesis of integration
Whole brain data
(20 um
microscopic MRI)
Mosiac LM
images (1 GB+)
Conventional LM
images
Individual cell
morphologies
EM volumes &
reconstructions
Solved molecular
structures
No single technology serves
these all equally well.
Multiple data types;
multiple scales; multiple
databases
A data integration problemA data integration problem
3. • NIF is an initiative of the NIH Blueprint consortium of institutesNIF is an initiative of the NIH Blueprint consortium of institutes
– What types of resources (data, tools, materials, services) are available to theWhat types of resources (data, tools, materials, services) are available to the
neuroscience community?neuroscience community?
– How many are there?How many are there?
– What domains do they cover? What domains do they not cover?What domains do they cover? What domains do they not cover?
– Where are they?Where are they?
• Web sitesWeb sites
• DatabasesDatabases
• LiteratureLiterature
• Supplementary materialSupplementary material
– Who uses them?Who uses them?
– Who creates them?Who creates them?
– How can we find them?How can we find them?
– How can we make them better in the future?How can we make them better in the future?
http://neuinfo.org
• PDF filesPDF files
• Desk drawersDesk drawers
4. Old Model: Single type of content; singleOld Model: Single type of content; single
mode of distributionmode of distribution
ScholarScholar
LibraryLibrary
Scholar
PublisherPublisher
Systems for cataloging, standards, and citation in placeSystems for cataloging, standards, and citation in place
6. The duality of modern scholarship
Observation: Those who build information systems from the
machine side don’t understand the requirements of the
human very well
Those who build information systems from the human side,
don’t understand requirements of machines very well
Scholarship requires the ability to cite and track usage of scholarly
artifacts. In our current mode of working, there is no way to track
artifacts as they move through the ecosystem; no way to incrementally
add human expertise; no way to look across the entirety
Scholarship requires the ability to cite and track usage of scholarly
artifacts. In our current mode of working, there is no way to track
artifacts as they move through the ecosystem; no way to incrementally
add human expertise; no way to look across the entirety
7. Whither neuroscience information?Whither neuroscience information?
∞
What is easily machine
processable and accessible
What is easily machine
processable and accessible
What is potentially knowableWhat is potentially knowable
What is known:
Literature, images, human
knowledge
What is known:
Literature, images, human
knowledge
Unstructured;
Natural language
processing, entity
recognition, image
processing and
analysis; paywalls
communication
Abstracts vs full
text vs tables etc
8. NIF: A New Type of Entity for New Modes ofNIF: A New Type of Entity for New Modes of
Scientific DisseminationScientific Dissemination
• NIF’s mission is to maximize the awareness of, access to
and utility of research resources produced worldwide to
enable better science and promote efficient use
– NIF unites neuroscience information without respect to
domain, funding agency, institute or community
– NIF is like a “Pub Med” for all biomedical resources and a “Pub
Med Central” for databases
– Makes them searchable from a single interface
– Practical and cost-effective; tries to be sensible
– Learned a lot about the effective data sharing
The Neuroscience Information Framework is an initiative of the
NIH Blueprint consortium of institutes http://neuinfo.org
The Neuroscience Information Framework is an initiative of the
NIH Blueprint consortium of institutes http://neuinfo.org
10. Data Federation: Deep searchData Federation: Deep search
http://neuinfo.org
With the thousands of databases and other information sources
available, simple descriptive metadata will not suffice
With the thousands of databases and other information sources
available, simple descriptive metadata will not suffice
11. A unified framework for neuroscienceA unified framework for neuroscience
Hippocampus OR “Cornu Ammonis” OR
“Ammon’s horn”
Hippocampus OR “Cornu Ammonis” OR
“Ammon’s horn”
NIF queries > 200 databases; ~400 million recordsNIF queries > 200 databases; ~400 million records
12. NIF Semantic Framework: NIFSTD ontologyNIF Semantic Framework: NIFSTD ontology
• NIF uses ontologies to help navigate across and unify neuroscience
resources
• Ontologies are built from community ontologies cross integration with
other domains
NIFSTDNIFSTD
OrganismOrganism
NS FunctionNS FunctionMoleculeMolecule InvestigationInvestigationSubcellular
structure
Subcellular
structure
MacromoleculeMacromolecule GeneGene
Molecule DescriptorsMolecule Descriptors
TechniquesTechniques
ReagentReagent ProtocolsProtocols
CellCell
ResourceResource InstrumentInstrument
DysfunctionDysfunction QualityQualityAnatomical
Structure
Anatomical
Structure
13. Purkinje
Cell
Axon
Terminal
Axon
Dendritic
Tree
Dendritic
Spine
Dendrite
Cell body
Cerebellar
cortex
Bringing knowledge to data: Ontologies as frameworkBringing knowledge to data: Ontologies as framework
There is little obvious connection between
data sets taken at different scales using
different microscopies without an explicit
representation of the biological objects that
the data represent
There is little obvious connection between
data sets taken at different scales using
different microscopies without an explicit
representation of the biological objects that
the data represent
14. : C: C
Neurolex: > 1 million triples
Dr. Yi Zeng: Chinese neural knowledge base
NIF Cell Graph
This is your brain on
computers
15. Ontologies as a data integration frameworkOntologies as a data integration framework
•NIF Connectivity: 7 databases containing connectivity primary data or claims
from literature on connectivity between brain regions
•Brain Architecture Management System (rodent)
•Temporal lobe.com (rodent)
•Connectome Wiki (human)
•Brain Maps (various)
•CoCoMac (primate cortex)
•UCLA Multimodal database (Human fMRI)
•Avian Brain Connectivity Database (Bird)
•Total: 1800 unique brain terms (excluding Avian)
•Number of exact terms used in > 1 database: 42
•Number of synonym matches: 99
•Number of 1st
order partonomy matches: 385
16. 0
1-10
11-100
>101
Open World-Closed World: Mapping the knowledge - data space
Data Sources
NIF lets us ask: where isn’t there data? What isn’t studied? Why?NIF lets us ask: where isn’t there data? What isn’t studied? Why?
18. ““The Data Homunculus”The Data Homunculus”
Funding drives representation in the data spaceFunding drives representation in the data space
19. Neurolex.org: A computableNeurolex.org: A computable
lexicon for neurosciencelexicon for neuroscience
http://neurolex.org Larson et al, Frontiers in Neuroinformatics, 2013Larson et al, Frontiers in Neuroinformatics, 2013
•Semantic MediaWiki
•Provide a simple interface
for defining the concepts
required
•Light weight semantics
•Community based:
•Anyone can contribute their
terms, concepts, things
•Anyone can edit
•Anyone can link
•Accessible: searched by Google
•Growing into a significant
knowledge base for
neuroscience
•25,000 concepts
Demo D03
200,000
edits
150
contributors
200,000
edits
150
contributors
21. Structural LexiconStructural Lexicon
The scourge of neuroanatomical nomenclatureThe scourge of neuroanatomical nomenclature
• Problem: Neuroscientists have a
myriad number of ways to
parcellate the brain
– Brains are made up of networks
that do not respect gross
anatomical boundaries
– Partonomies are generally along
multiple axes:
• Volummetric (species
dependent): NeuroNames
• Functional (Swanson)
• Developmental
• Cytoarchitectural
– Partonomies are often weak
• Arbitrary but defensible
Program on Ontologies for Neural Structures, INCF-
creating a computable lexicon for neural structures
Program on Ontologies for Neural Structures, INCF-
creating a computable lexicon for neural structures
23. Structural Lexicon in NeurolexStructural Lexicon in Neurolex
Brain
Region
Brain
Region
Brain
Parcel
Brain
Parcel
•Trans-species
•“Stateless”, i.e. no universal defining
criteria
•General structures and partonomies
based on Neuroanatomy 101
Partially overlaps
e.g., Hippocampus, Dentate gyrus
•Species specific
•Specific reference
•Defining criteria
•Sometimes partonomy;
sometimes not
e.g., Hippocampus of ABA2009
24. ““When I use a word...it means what I choose itWhen I use a word...it means what I choose it
to mean”to mean”
25. Neurolex NeuronNeurolex Neuron
• Led by Dr. Gordon
Shepherd
• > 30 world wide
experts
• Simple set of
properties
• Consistent naming
scheme
• Integrated with
Structural Lexicon
• Used for annotation in
other resources, e.g.,
NeuroElectro
26. ““You have broken links”You have broken links”
Red Links: Information is missing (or misspelled)Red Links: Information is missing (or misspelled)
27. Location of Cell Soma
Location of dendrites
Location of local axon
arbor
28. Analysis of Red Links in the Neuron RegistryAnalysis of Red Links in the Neuron Registry
• Analysis of red links
tells us where
instructions aren’t
clear, the information
isn’t available, or the
model insufficient
– Conceptualization not
clear
• what is most important
thing about local axon
terminals?
– Tool doesn’t capture
all details
Social networks and community sites let us learn things from the
collective behavior of contributors INCF/HBP Knowledge Space
Social networks and community sites let us learn things from the
collective behavior of contributors INCF/HBP Knowledge Space
29. Re-inventing Narrative: Do I have to write inRe-inventing Narrative: Do I have to write in
triples?triples?
• Not all entities are well-enough specified that they
lend themselves to deep annotation
– And, as we’ve seen in the previous example, we probably
don’t want to pretend that they are
• But…sometimes they are
– Semantic annotation of research papers to make them
“machine-interpretable” has been a goal of many
– Can we update the way that authors produce manuscripts
so that they are easier to process?
• NIF pilot project: Semantic annotation of entities
that researchers would understand
30. The problem: How many papers were
published that used my: antibody
Paz et al,
J Neurosci, 2010
31. Now, go find the antibody
http://www.millipore.com/searchsummary.do?tabValue=&q=gfap Nov 12,
32. Jan 15, 2014A catalog number is not a persistent identifierA catalog number is not a persistent identifier
34. If we can’t do it,
neither can the robot
• Automated text mining tools were not
deployed on this problem, because too few
antibodies were able to be automatically
identified
• We are asking authors to change their ways,
instead!
• Almost all antibodies were identified with the
company name, city and state, but the
information is useless if the goal is to identify
the antibody used
35. The Resource Identification InitiativeThe Resource Identification Initiative
• NIF, FORCE11 and
partners
– Led by Anita
Bandrowski and
Melissa Haendel
• Identify 3 types of
research resources
– Antibodies
– Genetically
modified animals
– Software
http://force11.org/Resource_identification_initiativehttp://force11.org/Resource_identification_initiative
36. Musings: You can’t do that!Musings: You can’t do that!
• Two powerful trends in the 21st
century:
– Networking machines and networking people
– Moving science into a machine-accessible platform has been a challenge
• Mechanistically
• Culturally
• Sociologically
• “A foolish consistency is the hobgoblin of little minds”
– When you have a lot of data and information in an accessible form, we
can start to look at actual practices and trends
– Focusing on the “negative space”, i.e., what we don’t know, reveals
glimpses into sources of bias and confusion
• When we scratch the surface of science, we find uncertainty and confusion
– Not a failure, but an opportunity
• Sometimes we can be precise, i.e., which reagents we used
• Sometimes, we can’t so we should set up systems so we can learn from that
37. Next Steps: Neurolex to Knowledge SpaceNext Steps: Neurolex to Knowledge Space
Data SpaceData Space
Laboratory
Space
Laboratory
Space
Knowledge
Space
Knowledge
Space
BAMS
LexiconLexicon
EncyclopediaEncyclopedia
39. What is the “completeness” of our knowledge?What is the “completeness” of our knowledge?
Neocortex
Olfactory bulb
Neostriatum
Cochlear nucleus
All neurons with cell bodies in the same brain region are grouped
together
All neurons with cell bodies in the same brain region are grouped
together
Properties in Neurolex
•Simple set of
properties that can
be reasonably
supplied with a
minimal amount of
effort
40. The case of the meanest journal in the world,
coincidentally having the lowest retraction rate
41. The landscape is messy, diverse and evolving: Data toThe landscape is messy, diverse and evolving: Data to
Knowledge – Knowledge to DataKnowledge – Knowledge to Data
NIF favors a hybrid, tiered,
federated system
• Domain knowledge
– Ontologies
• Claims, models and
observations
– Virtuoso RDF triples
– Model repositories
• Data
– Data federation
– Spatial data
– Workflows
• Narrative
– Full text access
NeuronNeuron Brain partBrain part DiseaseDisease
OrganismOrganism GeneGene
Caudate projects to
Snpc
Caudate projects to
Snpc Grm1 is upregulated
in chronic cocaine
Grm1 is upregulated
in chronic cocaine
Betz cells
degenerate in ALS
Betz cells
degenerate in ALS
NIF provides the tentacles that connect the pieces: a
new type of entity for 21st
century science
NIF provides the tentacles that connect the pieces: a
new type of entity for 21st
century science
TechniqueTechnique
PeoplePeople
42. Data about the subthalamusData about the subthalamus
http://neuinfo.org
Hinweis der Redaktion
Queue movie after this. Would be nice to visually pull this together with an animated view.
Current model: Scholars are producing multiple types of research objects; each goes to their own infrastructure with little coordination among them.
Consumer no longer exclusively a scholar: General public wants access to what they pay for; automated agents are accessing first and mining the content.