The Neuroscience Information Framework:The present and future of neuroscience data sharing

The Neuroscience Information
Framework:
The present and future of neuroscience
data sharing
Maryann Martone, Ph. D.
University of California, San Diego

The
Encyclopedia
of Life
A…
Access to data has
changed over the
years
Tim Berner-s Lee: Web of data
Wikipedia defines Linked Data as "a term used
to describe a recommended best practice for
exposing, sharing, and connecting pieces of
data, information, and knowledge on the
SemanticWeb using URIs and RDF.”
http://linkeddata.org/
Genban
k
PDB

The mountain of data problem
Would like to be able to find:
 What is known****:
 What is the average diameter of a Purkinje
neuron
 Is GRM1 expressed In cerebral cortex?
 What are the projections of hippocampus
 What genes have been found to be
upregulated in chronic drug abuse in adults
 What studies used my monoclonal mouse
antibody against GAD in humans?
 Find all instances of spines that
contain membrane-bound organelles
 ****by combining data from different
sources and different groups
 What is not known:
 Connections among data
 Gaps in knowledge
Required Components:
– Query interface
– Search strategies
– Data sources
– Infrastructure
– Results display
– Trust
– Context
– Analysis tools
– Tools for translating existing
content into linkable form
– Tools for creating new data ready
to be linked

Where would you rather look?
Unstructured vs structured data
Publishing data in the literature/ web pages vs databases and tables

Scale
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 multi-scale data problem
A data federation problem

Two organizing frameworks for
knowledge
Knowledge in space and spatial relationships
(the “where”)
Knowledge in words, terminologies and
logical relationships (the “what”)

Assembling data into coherent
models
Snavely et al. Scene Reconstruction andVisualization from Community Photo
Collections

What if...
 The Matterhorn could be 15 different things?
 There were 6 billion Matterhorns, all more or less different from one
another?
 The Roman Coliseum was called by 45 different names?
 The photo represented 1/1,000,000 of the whole with no context?
 Photos weren’t annotated at all or were tagged “1” or “mm45”?
 The statue of liberty was represented as a mathematical equation? Or a
scatter plot?
1

Cerebral peduncle
Internal capsule
Corticospinal tract
Every brain is
different;
terminology is
used
inconsistently;
there are many
names for the
same structure

Curators vs researchers
• Example of segmented object names from
CCDB for a Node of Ranvier:
• Mitochondria1
• Shwannlowermerge
• U.L.Cisternae
• Crop
• Loop7_lower
• Blue
• Alex
• Lysosomme_3
http://ccdb.ucsd.edu
•Alex left
•Program used to
create
annotations
obsolete

The Neuroscience Information Framework: Discovery and
utilization of web-based resources for neuroscience
 A portal for finding and
using neuroscience
resources
 A consistent framework for
describing resources
 Provides simultaneous
search of multiple types of
information, organized by
category
 Supported by an expansive
ontology for neuroscience
 Utilizes advanced
technologies to search the
“hidden web”
http://neuinfo.org
UCSD,Yale, CalTech, George Mason, Washington Univ
Supported by NIH Blueprint

Scale
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 federation problem

How many resources are there?
•NIF Registry: A
catalog of
neuroscience-relevant
resources
•> 3500 currently
described
•> 1500 databases
•Another 4000
awaiting curation
•And we are finding
more every day

NIF Data Federation
 Too many databases to visit
 Capturing content in a few keywords is difficult if not impossible
 Each is organized differently; different UI’s, data models and tools
 NIF provides tools for databases to register their content to NIF
 Access to deep content; currently searches over 35 million records
from > 65 different databases
 Web services, schema registration,XML-based description, RDF
 Organized according to level of nervous system and data type, e.g.,
brain activation foci
 Enhanced keyword query interface
 Link to host resource
 Accompanied by a tutorial
 Defines common data models for similar data

HippocampusOR “CornuAmmonis” OR
“Ammon’s horn” Query expansion: Synonyms
and related concepts
Boolean queries
Data sources
categorized by
“data type” and
level of nervous
system
Simplified views of
complex data
sources
Tutorials for using
full resource when
getting there from
NIF
Link back to
record in
original
source

NIF data federation...
 Simultaneous access to multiple sources of information through a
concept-based interface
 Unique resource for asking certain types of questions
 e.g., what rat strains have been most commonly used in research
 Indexes content in the hidden web not currently well served by search engines
 A set of tools for making resources available through the NIF
 A platform for data integration
 Simplified and neuroscience-centered views of very complicated resources
 An ontology for enhanced query and integration
 A wealth of real information on the practical issues of search across and
integration of data in the neurosciences
 Share experiences through publications, presentations, blogs and with other projects
 Developing annotation standards that help with search
 Provide best practices for resource creators

What are the connections of the
hippocampus?
Connects to
Synapsed with
Synapsed by
Input region
innervates
Axon innervates
Projects toCellular contact
Subcellular contact
Source site
Target site
Each resource implements a different, though related model;
systems are complex and difficult to learn, in many cases

Is GRM1 in cerebral cortex?
 NIF system allows easy search over multiple sources of information
 But, we have difficulty finding data
 Well known difficulties in search
 Inconsistent and sparse annotation of scientific data
 Many different names for the same thing
 The same name means many things
 “Hidden semantics”: 1 = male; 1 = present; 1=mouse
Allen Brain Atlas
MGD
Gensat

Cerebral Cortex
Atlas Children Parent
Genepaint Neocortex, Olfactory cortex (Olfactory
bulb; piriform cortex), hippocampus
Telencephalon
Allen Brain Atlas Cortical plate, Olfactory areas,
Hippocampal Formation
Cerebrum
MBAT (cortex) Hippocampus, Olfactory, Frontal,
Perirhinal cortex, entorhinal cortex
Forebrain
GENSAT Not defined Telencephalon
BrainInfo frontal lobe, insula, temporal lobe,
limbic lobe, occipital lobe
Telencephalon
Brainmaps
Entorhinal, insular, 6, 8, 4, A SII 17,
Prp, SI
Telencephalon

Result
•We are not publishing data in a
form that is easy to integrate
•What we mean isn’t clear to a
search engine (or even to a
human)
•We use many different data
structures to say the same
thing
•We don’t provide crucial
information
•Searching and navigating across
individual resources takes an
inordinate amount of human effort
Tempus PecuniaEst Painting by Richard
Harpum

NIF: Minimum requirements to use shared
data
 You have to be able to find it
 Accessible through the web
 Structured or semi-structured
 Annotations
 You have to be able to use it
 Data type specified and in a usable form
 You have to know what the data mean
 Semantics
 Identity
 1 = integer, time scale, male, left hemisphere
 Context: Experimental metadata
Reporting neuroscience data within a consistent framework helps enormously

Whole Brain Catalog
Stephen Larson, Mark Ellisman http://wholebraincatalog.org
Uses 3D
game
engine to
bring
together
multiple
data types
within a
common
framewor
k

Purkinje
Cell
Axon
Terminal
Axon
Dendritic
Tree
Dendritic
Spine
Dendrite
Cell body
Cerebellar
cortex
Multiscale integration is not obvious
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

What is an ontology?
Brain
Cerebellum
Purkinje Cell Layer
Purkinje cell
neuron
has a
has a
has a
is a
 Ontology: an explicit, formal
representation of concepts and
relationships among them
within a particular domain that
expresses human knowledge in a
machine readable form
 Branch of philosophy: a theory
of what is
 e.g., Gene ontologies

What ontology isn’t
(or shouldn’t be)
 A rigid top-down fixed hierarchy for
limiting expression in the
neurosciences
 Not about restricting expression but
how to express meaning clearly and
in a machine readable form
 A bottomless resource-eating pit
that consumes dollars and returns
nothing
 A cure-all for all our problems
 A completely solved area
 Applied vs theoretical
 Easy to understand Mike Bergman

What can ontology do for us?
 Express neuroscience concepts in a way that is machine readable
 Synonyms, lexical variants
 Definitions
 Provide means of disambiguation of strings
 Nucleus part of cell; nucleus part of brain; nucleus part of atom
 Rules by which a class is defined, e.g., a GABAergic neuron is neuron that
releases GABA as a neurotransmitter
 Properties
 Quantities
 Provide universals for navigating across different data sources
 Semantic “index”
 Perform reasoning
 Link data through relationships not just one-to-one mappings
 Provide the basis for concept-based queries to probe and mine data
 As a branch of philosophy, make us think about the nature of the
things we are trying to describe, e.g., synapse is a site

Linking datatypes to semantics: What is
the average diameter of a Purkinje
neuron dendrite?
 Branch structure not a tree,
not a set of blood vessels, not
a road map but a DENDRITE
 Because anyone who uses
Neurolucida uses the same
concepts: axon, dendrite, cell
body, dendritic spine,
information systems can
combine the data together in
meaningful ways
 Neurolucida doesn’t, however,
tell you that dendrite belongs
to a neuron of a particular
type or whether this dendrite
is a neural dendrite at all
( (Color Yellow) ; [10,1]
(Dendrite)
( 5.04 -44.40 -89.00 1.32) ; Root
( 3.39 -44.40 -89.00 1.32) ; R, 1
(
( 2.81 -45.10 -90.00 0.91) ; R-1, 1
( 2.81 -45.18 -90.00 0.91) ; R-1, 2
( 1.90 -46.01 -90.00 0.91) ; R-1, 3
( 1.82 -46.09 -90.00 0.91) ; R-1, 4
( 0.91 -46.59 -90.00 0.91) ; R-1, 5
( 0.41 -46.83 -92.50 0.91) ; R-1, 6
(
( -0.66 -46.92 -88.50 0.74) ; R-1-1, 1
( -0.74 -46.92 -88.50 0.74) ; R-1-1, 2
( -2.15 -47.25 -88.00 0.74) ; R-1-1, 3
( -2.15 -47.33 -88.00 0.74) ; R-1-1, 4
( -3.06 -47.00 -87.00 0.74) ; R-1-1, 5
( -4.05 -46.92 -86.00 0.74) ; R-1-1, 6
Output of Neurolucida neuron trace

“A rose by any other name...”:
 Identity:
 Entities are uniquely identifiable
 Name is a meaningless numerical identifier (URI: Uniform resource identifier)
 Any number of human readable labels can be assigned to it
 Definition:
 Genera: is a type of (cell, anatomical structure, cell part)
 Differentia: “has a” A set of properties that distinguish among members of that
class
 Can include necessary and sufficient conditions
 Implementation: How is this definition expressed
 Depending on the nature of the concept or entity and the needs of the
information system, we can say more or fewer things
 Different languages; can express different things about the concept that can be
computed upon
 OWLW3C standard, RDF

Entity recognition: Are you the M Martone
who...
The Gene Wiki: community intelligence applied to human gene annotation.
Huss JW 3rd, Lindenbaum P, Martone M, Roberts D, Pizarro A, Valafar F, Hogenesch
JB, Su AI. Nucleic Acids Res. 2010 Jan;38(Database issue):D633-9.
Ontologies for Neuroscience:What are they and What are they Good for? Larson SD,
Martone ME. Front Neurosci. 2009 May;3(1):60-7. Epub 2009 May 1.
Three-dimensional electron microscopy reveals new details of membrane systems for
Ca2+ signaling in the heart. HayashiT, Martone ME,Yu Z,Thor A, Doi M, Holst MJ,
Ellisman MH, Hoshijima M. J Cell Sci. 2009 Apr 1;122(Pt 7):1005-13.
Traumatic brain injury and the goals of care.Martone M. Hastings Cent Rep. 2006 Mar-
Apr;36(2):3.
Three-dimensional pattern of enkephalin-like immunoreactivity in the caudate nucleus of the
cat.Groves PM, Martone M,Young SJ, Armstrong DM. J Neurosci. 1988 Mar;8(3):892-900.
Some analyses of forgetting of pictorial material in amnesic and demented
patients.Martone M, Butters N,Trauner D. J Clin Exp Neuropsychol. 1986 Jun;8(3):161-78.

ID: 555 55 5555
 Full URI-
http://usagov/ss#555555555
 Label: Maryann Elizabeth
Martone
 Synonym: ME Martone, M
Martone, Maryann
 Abbreviation: MEM
 Is a
 Has a
 Is that entity which has these
properties
M Martone
Dept of
Psychiatry,
UCSD
MH
Ellisman
Publications
BostonVA
Hospital
Text mining algorithms can discover a lot of things
about me

NIFSTD: Comprehensive Ontology
 NIF covers multiple structural scales and domains of relevance to neuroscience
 Aggregate of community ontologies with some extensions for neuroscience, e.g., Gene Ontology, Chebi,
Protein Ontology
 Simple, basic “is a : hierarchies that can be used “as is” or to form the building blocks for more complex
representations
NIFSTD
Organism
NS FunctionMolecule Investigation
Subcellular
structure
Macromolecule Gene
Molecule Descriptors
Techniques
Reagent Protocols
Cell
Resource Instrument
Dysfunction Quality
Anatomical
Structure

Query across resources: Snca
and striatum
NIF uses the NIFSTD ontologies to query across sources that use very
different terminologies, symbolic notations and levels of granularity

Entity mapping
BIRNLex_435 Brodmann.3
Explicit mapping of database content helps disambiguate non-unique and
custom terminology

Concept-based search: search by meaning
 SearchGoogle: GABAergic neuron
 Search NIF: GABAergic neuron
 NIF automatically searches for types of GABAergic
neurons
Types of GABAergic
neurons

NIF #1: You have to be able to
find it...
 What genes are upregulated by drugs of abuse in the adult
mouse?
Morphine
Increased
expression
Adult Mouse

Integration of knowledge based on relationships
Looking for commonalities and distinctions among animal
models and human conditions based on phenotypes
Sarah Maynard, Chris Mungall, Suzie Lewis NINDS
Thalamus
Cellular inclusion
Midline nuclear
group
Lewy Body
Paracentral nucleus
Cellular inclusion

Building ontologies: modified
OBO Foundry principles
 NIF has adopted certain practices which we have found
make it easier to build and work with ontologies in
neuroscience
 Unique numerical identifers for class names
 Single asserted hierarchies
 Avoid multiple inheritance
 Use community ontologies
 One ontology per domain
Open Bio Ontologies

Asserted vs defined classes: the
power of explicit semantics
 Asserted class: Purkinje cell is a type of neuron
 Why? Because I said so!
 Defined class: Purkinje cell is a GABAergic neuron
 http://ontology.neuinfo.org/NIF/BiomaterialEntities/NIF-Neuron-
NT-Bridge.owl#nlx_neuron_nt_090803
 Why? Because it is a member of the class Neuron that releases
neurotransmitter GABA
 Logical definition based on properties
 Membership in the class is computed by reasoners based on the
satisfaction of a set of conditions
 Makes building ontologies tractable because you don’t have to
create multiple hierarchies; you can infer them

Reclassification of a flat hierarchy based on logical definitions
The principle of
single inheritance
•Each class belongs to
only a single asserted
hierarchy that is
generally fairly
uninteresing
•Through the
assignments of
properties and
restrictions, each class
may belong to many
defined hierarchies
•The criteria for
membership in that
class is explicit
•Easier bookkeeping

The case for shared ontologies
Brain
Cerebellum
Cerebellar
Cortex
Cerebellar Purkinje
cell
Purkinje neuron
Purkinje cell
soma
Purkinje cell
layer
Cerebellar
cortex
IP3
Cerebellum
•To create the
linkages requires
mapping
•Mapping is
usually
incomplete and
not always
possible
•Can’t take
advantage of
others’ workTop down anatomy ontology Cell centered anatomy ontology

Cerebellum
Purkinje cell
soma
Cerebellum
Purkinje cell
dendrite
Cerebellum
Purkinje cell axon
(Cell part
ontology)
Cerebellum granule cell
layer (Anatomy ontology)
Cerebellum Purkinje
cell layer
Cerebellum
molecular layer
Has
part
Has
part
Has
part
Is part of
Is part of
Is part of
Shared building blocks: Knowledge base is enriched
Calbindin IP3
(CHEBI:16595)
Cerebellum
Purkinje neuron
(Cell Ontology)
Cerebellar cortex
Has part
Has part
Has part

Access to shared ontologies
 Neuroscience Information Framework
(http://neuinfo.org): Ontologies available as
OWL file, RDF and throughWeb Services
 https://confluence.crbs.ucsd.edu/display/NIF/
OntoQuestMain.
 NCBO Bioportal
(http://bioportal.bioontology.org/): Repository
of ontologies for biomedical research
 199 ontologies (including NIFSTD)
 Contains many mappings
 Provides annotation services
 INCF Program on Ontologies for Neural
Structures
 Neuronal RegistryTask Force
 Description of neural properties
 Structural Lexicon
 Description of properties across scales

Building or expanding ontologies
Michael Bergman

NeuroLexWiki
http://neurolex.org Stephen Larson
SemanticWiki: provides community
interface for viewing, enhancing and
modifying NIFSTD ontologies
•Provide a simple
framework for
defining the
concepts required
•Cell, Part of
brain,
subcellular
structure,
molecule
•On demand
•Assign permanent
URI
•Ontologists/knowle
dge engineers build
in complexity
•Tries to teach and
adhere to basic best
practices

Define by rules: Generate multiple
classifications programmatically

Enriching the knowledge base
Members of this class automatically
generated according to a rule expressed in
a standard query language

Inferring the Mesoscale
 The NIFSTD is expressed in
OWL (Web Ontology
Language)
 Supports reasoning and inference
 Through integration with
other ontologies covering
gross anatomy and molecular
entities, we are working to
create inferences across scales
 Analyze locally; infer globally
Larson and Martone, 2007Stephen Larson

Inferencing across scales: Compare
statements
1. Look brain region up in NeuroLex
2. Look up cells contained in the brain region
3. Find those cells that are known to project out
of that brain region
4. Look up the neurotransmitters for those cells
5. Determine whether those neurotransmitters
are known to be excitatory or inhibitory
6. Report the projection as excitatory or
inhibitory, and report the entire chain of logic
with links back to the wiki pages where they
were made
7. Make sure user can get back to each statement
in the logic chain to edit it if they think it is
wrong
Stephen Larson
CHEBI:18243

A semantic web for neuroscience? Good idea.
So all I have to do is...
 Express your data in RDF?
 Well...
 Which RDF
 Bio2RDF, BioRDF, Linked Data, Open Data, SemanticWeb
 Use URI’s for all data elements
 Well...
 What exactly does that mean?
 Shared Names, BioRDF, my own?
 Use shared ontologies?
 Well...
 Which ones?
 I don’t have one
 They’re not stable
 They take too long
 I’d rather share your toothbrush
 Wait forWatson 3.0
Effective data sharing is still an act of will

We do know some things
NIF Blog
1. Register your resource
with NIF!!!!
2: Mindfulness
 Resource providers: Mindfulness that your
resource is contributing data to a global
federation
 Link to shared ontology identifiers where
possible
 Stable and unique identifiers for data
 Explicit semantics
 Database, model, atlas
 Researchers: Mindfulness when publishing
data that it is to be consumed by machines
and not just your colleagues
 Accession numbers for genes and species
 Catalog numbers for reagents
 Provide supplemental data in a form where it is
is easy to re-use

Many thanks to...
Amarnath Gupta, UCSD, Co Investigator
Jeff Grethe, UCSD, Co Investigator
Anita Bandrowski, NIF Curator
Gordon Shepherd,Yale University
Perry Miller
Luis Marenco
DavidVan Essen,Washington University
Erin Reid
Paul Sternberg, CalTech
ArunRangarajan
Hans Michael Muller
GiorgioAscoli,George Mason University
SrideviPolavarum
FahimImam, NIF Ontology Engineer
Karen Skinner, NIH, Program Officer
Mark Ellisman
Lee Hornbrook
Kara Lu
VadimAstakhov
XufeiQian
Chris Condit
Stephen Larson
Sarah Maynard
Bill Bug

The Neuroscience Information Framework:The present and future of neuroscience data sharing

More Related Content

What's hot (20)

Viewers also liked (17)

Similar to The Neuroscience Information Framework:The present and future of neuroscience data sharing (20)

More from Neuroscience Information Framework (20)

Recently uploaded (20)

The Neuroscience Information Framework:The present and future of neuroscience data sharing

Editor's Notes