Biomedical Informatics Research Network and

1
Building Multiscale Models of the Nervous System
through the BIRN A Practical Experience
Maryann E. Martone, Ph. D.

• Biomedical Informatics Research Network and
• National Center for Microscopy and Imaging
  Research
• Center for Research in Biological Systems
• University of California, San Diego

2
BIRNBiomedical Informatics Research Network

• National Center for Research Resources
• Establish cyberinfrastructure for storing,
  manipulating and sharing data and resources
• High speed, robust connectivity via Internet2
• Data and computational resources
• Data integration over distributed databases
• Current test beds focused on neuroimaging
• Human MRI
• Human fMRI
• Mouse models of neurological disease
• Involves 30 Research Institutions and 40 research
  groups

Will no longer matter where data, instruments
and computational resources are located
physically
http//nbirn.net
3
Goals of BIRN Project
insert your grid project here

• 21st Century scientific data management
• 20th century biology revolutionized means used to
  acquire data - gave solid footing for old
  theories and provided the experimental bases to
  flesh out the molecular details in many areas of
  science.
• 20th century scientific data management is not so
  very different than 19th century
• Depends on one scientist absorbing as much as
  they can with their limited time knowledge,
  seeking correlations cogently assembling a
  story
• 21st Century data management seeks to grow way
  beyond the single scientist/lab gt meta-analysis
  concept-driven data mining
• Biomolecular sci-data management got there
  through the 1990s - BIRN is trying to deliver
  this capability to other domains of biomedical
  science
• The network is the repository
• Data and knowledge must be machine interpretable
• We are not trying to support business as usual
  (and therein lies the challenge)

Courtesy of Bill Bug
4
BIRN Test Beds

• BIRN Coordinating Center
• Develops, deploys and maintains
  cyberinfrastructure
• Data integration through development of federated
  database tools
• Morphometry BIRN
• Methodology for pooling and analyzing data across
  neuroimaging sites
• anatomical differences and specific memory
  dysfunctions, such as depression, mild
  Alzheimers disease, and mild cognitive
  impairment
• Function BIRN
• Methodology for multi-site fMRI data collection,
  sharing, and integration with focus on
  schizophrenia
• Mouse BIRN
• Building multiscale brain atlases of mouse models
  of human neurological disease
• Integration of multiscale imaging data with
  genomic data

5
Combining MRI data across sites
Before
After

• Calibration standards multisite imaging study
• Human Imaging database
• Metadata standards for neuroimaging data
  transformation

Jovicich et al., Reliability in multi-site
structural MRI studies Effects of gradient
non-linearity correction on phantom and human
data NeuroImage 30 436, 2006 Keator et al., A
general XML schema and SPM toolbox for storage of
neuro-imaging results and anatomical labels.
Neuroinformatics, 4 199, 2006
6
Multiscale Investigation of Dopamine Transporter
KO Mouse

• Duke, UCLA, UCSD, Cal Tech, UT Memphis, Drexel

7
Mouse BIRN Data Integration Framework

• different model from human test beds
• groups work independently but with an eye towards
  sharing

4. Use mediator to navigate and query across
data sources
1. Create multimodal databases
3. Situate the data in a common spatial framework
2. Create conceptual links to a shared ontology
8
Challenges of Data Integration of Distributed Data

• Semantic concordance
• Medium spiny cell vs medium spiny neuron(e)
• C57Bl/6J vs C57BLJ6 vs C57B6J
• Different representations in different databases
• Age 21 months
• Age Date of imaging - date of birth
• Indirect relationships
• Basal Ganglia --gt medium spiny neuron
• Alzheimers disease --gt alpha synuclein
  overexpressor
• Inconsistent or partial relationships
• My amygdala vs your amygdala
• Reconciling different techniques
• Gene expression with microarray vs in situ
  hybridization vs immunocytochemistry

9
Data Integration for BIRN
Find animal models of movement disorders where
the volume of basal ganglia structures are
decreased in old animals
Integrated View
Knowledge Sources
Integrated View Definition
Mediator
Wrapper
Wrapper
Wrapper
Wrapper
UCSD
UCLA
Cal Tech
10
What is an Ontology?

• Way to communicate a shared understanding of a
  field
• representation of terminological knowledge
• concept hierarchy (is-a)
• further semantic relationships between concepts
  (is part of, causes etc.)
• Examples
• GO (Gene Ontology)
• NeuroNames
• Foundational model of anatomy
• Mouse Anatomy (Edinburgh)

Brain
has a
Cerebellum
has a
Purkinje Cell Layer
has a
Purkinje cell
is a
neuron
11
Multiscale Investigation of Neurological Disease
Navigating through Multi-resolution information
Linking animal and human imaging data
brain
Entopeduncular nucleus
Globus pallidus, internal segment
cerebellum
Disease
Animal Model
cerebellar cortex
Interpreting Results
Purkinje cell
Immunolabeling
Microarray
dendritic spine
Technique
Phenotype
12
BIRNLex Lexicon for multiscale investigation of
neurological disease

• BIRN Ontology Task Force
• Lexicon not a terminiology
• Each concept has a human readable definition
• Each concept has a unique identifier
• All synonyms have the same ID
• Current domains
• Neuroanatomy
• Mouse strains
• Neuropsychological assessments
• Next workflows, methods, diseases, animal
  models
• Builds on previous efforts by other groups (no
  intentional reinventing!)
• Foundational ontologies, Peter Foxs paradigm
  classes Neuronames brain structures
• Tried to utilize best practices promoted by the
  National Center for Biomedical Ontologies and the
  OBO Foundry project
• Promotes integration with other efforts
• Promotes transition to fully structured ontology
  with machine-processable semantics

13
BIRN Mediator
BIRN Mediator
14
The Smart Atlas A Grid-based GIS tool for
spatial integration of multiscale distributed
brain data
Ilya Zaslavsky, Joshua Tran, Asif Memon, Willy
Wong, Haiyun He, Amarnath Gupta
SRB
UCSD
15
Multiscale Investigation of Dopamine Transporter
KO Mouse

• Duke, UCLA, UCSD, Cal Tech, UT Memphis, Drexel

16
Space limitations (1)
Context is often difficult to discern across
scales -scale and contrast mechanisms differ
17
Space limitations (2) Multimodal Data
18
No magic technique
ICC
GSCF
RISH
NRISH
Technique will determine what is seen and how
labeling is interpreted still requires expert
knowledge
19
Core domain NeuroanatomyWhat is the
hippocampus?

• Resolution of technique
• Data representation
• Structural vs functional concepts
• Challenge in Human BIRN
• Complicated subcellular anatomy of the nervous
  system

20
Summary of Progress and Lessons Learned

• Data integration is multifaceted
• Calibration protocols
• Spatial and terminological relationships
• Engines to achieve integration
• Its hard
• Technology and people
• Challenge how to let the science and other
  technology development proceed while trying to
  develop these systems (and if it can proceed, why
  do we need the systems?)
• Systems are only now coming on-line
• BIRN databases are being populated BIRNLex v
  1.0 is just about ready mediator ready
• Scope of problem is beyond any single project
• Development with an eye towards integration with
  other efforts
• Creating these resources is helping to clarify
  our thinking about difficult concepts
• Creating the lexicon forces us to grapple with
  murky scientific concepts and try to formalize
  them
• Most biological concepts dont map neatly into
  hierarchies
• Even knocking off the low hanging fruit is
  helpful

21
Challenge Disease Maps
Parkinsons Disease
C0030567
Pathological feature
symptom
akinesia
C027746
neuronal degeneration
C0027746
tremor
rigidity
Lewy Body
C0085200
Dopamine neuron
C0815003
Motor deficit
C0746626
Cell inclusion
C0205708
Substantia nigra
C0175412
neurons
C0027882
Abnormal filaments
cortex
C0007776
glia
C0027836
Basal forebrain
C0041538
Alpha synuclein
C024566
22
Linking the human disease state to the animal
model
Parkinsons Disease
a-synuclein mouse
nuclear inclusion
Cellular phenotype
Pathological feature
Cytoplasmic inclusion
Cellular inclusion
Lewy Body
Alpha synuclein
Filamentous inclusion
neurons
ubiquitin
glia
23
Many Thanks To

• Guy Perkins
• Gina Sosinsky
• Guido Gaietta
• Steven Peltier
• Abel W. Lin
• Joy Sargis
• Tomas Molina
• Lisa Fong
• Lily Chen
• Amarnath Gupta
• Ben Giepmanns
• Josh Tran
• Willy Wong
• Heather Jiles
• Cem Mangir

• Mark Ellisman
• Tom Deerinck
• Bill Bug, Drexel
• Carol Bean, NIH
• Christine Fennema-Notestine
• Diana Martinez-Price
• Jessica Turner, UCI
• Jeff Grethe
• Masako Terada
• Stephan Lamont
• Daniel Rubin, NCBO
• Ying Jones
• Andrea Thor
• Yujun Wang

BIRN OTF