/home/precedings/apps/precedings/shared/storage/document/2007/1046/npre20071046-1

[ ]

1st INCF Workshop on Mouse and Rat Brain Digital Atlasing Systems

February 1314, 2007
International Neuroinformatics Coordinating Facility Secretariat
Stockholm, Sweden

Authors

Jyl Boline, Michael Hawrylycz, and Robert W. Williams
Email: rwilliam@nb.utmem.edu

Workshop Organizer

Robert W. Williams, University of Tennessee Health Science Center,
Memphis, Tennessee

Workshop Participants

Jan G. Bjaalie, INCF Secretariat, Stockholm, Sweden
Jyl Boline, Laboratory of Neuro Imaging, UCLA School of Medicine, Los Angeles, USA *Reporteur
Gregor Eichele, Max Planck Institute of Biophysical Chemistry, Göttingen, Germany
Francoise Gofflot, Institut Clinique de la Souris (ICS), Strasbourg, France
Teiichi Furuichi, RIKEN Brain Science Institute, Wako, Japan
Michael Hawrylycz, Allen Institute for Brain Science, Seattle, USA
Andreas Hess, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
G. Allan Johnson, Duke University, Durham, USA
Jonathan Nissanov, Drexel University, Philadelphia, USA
George Paxinos, University of New South Wales, Randwick, Australia
Charles Watson, Curtin University of Technology, Bentley, Australia
Ilya Zaslavsky, University of California, San Diego, USA

Invited but unable to attend

Johan Auwerx, Strasbourg, France
Mihail Bota, University of Southern California, Los Angeles, USA
Maryann Martone, National Center for Microscopy and Imaging Research,
University of California, San Diego, USA
Larry Swanson, University of Southern California, Los Angeles, USA
Arthur Toga, Laboratory of Neuro Imaging, UCLA School of Medicine, Los Angeles, USA

Supported by the INCF Central Fund and the Swedish Foundation for Strategic Research

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ ]

1. Executive Summary

The recommendations of this workshop group center around

the mission of INCF "to contribute to the development of scal-

able, portable, and extensible digital applications that can be

used by neuroscience laboratories to further knowledge of the

human brain and related diseases." Our expectation is that

in the near future, research will involve intense interactions

among scientists and computer networks at the level of ideas

and dynamic reusable data. Neuroscientists will place greater

reliance on stable web-enabled knowledgebases and will move

away from rigid legacy models involving static research sum-

maries and one-way non-digital communication. Atlases and

spatial indexes will play a fundamental role in this shifting

research landscape and will evolve into critical resources for

gathering, securing, analyzing, and communicating research.

In the next decade, the standard bearer of scientific progress--

the primary research paper--is likely to transition from a sleek

and static synopsis of results and conclusions to a more com-

plete, dynamic, and re-computable encapsulation of data and

interpretation. This transition is already evident in the fields of

genomics and bioinformatics in which papers are often point-

ers to massive data sets, analytic tools, and other appendix

material. It is likely that a typical rd-millenium neuroscience

communication/publication will include accessible data sets

with full metadata, far more complete details on experimental

design, and an enriched discussion section including interac-

tive content such as a forum for genuine discussion between

the lead authors and a larger community of reviewers and com-

mentators.

Web-accessible brain atlases and spatial indexes will undoubt-

edly be among the most important tools needed to transition

gracefully from our static synopsis mode of publication to a

data-rich, dynamic, multidimensional mode of scientific inter-

action. Atlases will be key query tools as they appeal to our

preference for visual exploration of complex data sets. Thus,

this is a crucial time for an international community of neu-

roscientists to begin to converge on a lingua franca for digital

neuroscience atlases, less with the goal of enforcing confor-

mity, than with the goal of building resources and tools that

translate among existing and future atlasing systems and their

terminologies. This kind of transnational and translational ac-

tivity ideally matches the mission of the INCF in the domain

of digital atlasing.

With these longer-term objectives in mind, the 1

INCF Mouse

and Rat Atlasing Workshop was held in Stockholm in February

2007. Our first objective was to survey current activities and

plans related to mouse and rat brain digital atlasing systems

and to produce a broad international inventory of resources and

ongoing efforts. The second objective was to review the range

of techniques that are being used to build, normalize, segment,

and label atlases and to examine what aspects of this techni-

cal work are redundant, compatible, and compliant across

platforms. The third objective was to forge an international

network to foster increased collaboration and interoperability

across national, linguistic, and funding barriers and to examine

how to promote international collaborations in the future.

The final and most important objective of this workshop was

to improve the impact of atlasing projects in the near term (

years) while reducing costs and redundancy of these efforts.

Funds spent in support of the INCF secretariat and the national

nodes should ultimately be leveraged many times over. Atlas-

ing efforts in each member country should be able to accom-

plish more as part of a coordinated INCF activity than they

would in isolation. To attain this goal, the INCF secretariat

and the supporting nodes strongly encourage integration, code

reuse, and joint projects.

Researcher goals and platforms differ, and all participants at

this workshop accept, and even encourage, a wide variety of

approaches to brain atlasing. However, all participants also ap-

preciate the enormous benefits that may be gained by combin-

ing and integrating across diverse resources.

Consensus and Agreement

There is a growing need for international digital atlas

based neuroinformatics infrastructure as an organizing

entity for data sharing and relating previously disparate

knowledge bases into a linked system.

The INCF is in a unique position to help coordinate the

creation of standards and tools to harvest and maintain

data and metadata.

The INCF should encourage open standards for protocol,

tools, and even specific data sets and atlases. There are

now several "open access" images of the mouse brain

(Allan Brain Atlas, Franklin and Watson, and NeuroTer-

rain). However, corresponding "open access" D seg-

mentation data sets are still largely unavailable. Efforts

by groups such as those at Drexel (Nissanov and col-

leagues) and UCLA (Toga and colleagues) are leading

the way to open access to D atlases.

There is a need for much more accessible data on con-

nections among CNS regions and neurons incorporated

into atlases to enrich functional network analysis (prefer-

ably open data sets of connections).

Infrastructure and tools developed should be open, gen-

erally available, and easy to use.

Any infrastructure work in this area depends heavily on

database persistence and sustainability and ontologies to

describe relevant data. Any work in this area will be in-

fluenced by the outcome of these upcoming INCF work-

shops.

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ ]

2. Introduction

Publications are currently the primary way in which research-

ers share their findings, data, and metadata with the scientific

community. In the past, the data from these experiments have

rarely been used again and other researchers have often found

it difficult to duplicate an experiment. More recently, there has

been a push to share the original data, metadata, and even the

processed and analyzed data with the rest of the scientific com-

munity. The added value and multiple applications of this are

obvious: evaluation of the initial experiment by other investi-

gators, the use of already collected data in other experiments,

and new meta-analyses. This gives researchers the ability to

examine issues in a way no single laboratory would have the

resources for and to potentially answer questions or even pose

new questions previously impossible.

Sharing data on a large scale with minimal interactions between

the data provider and the data consumer creates the challenge

of how to do this in a usable and meaningful manner. While it

may be relatively easy to reduce the barrier to sharing, it is dif-

ficult to do so to a degree that still facilitates the process of data

rediscovery and use for other purposes. Data repositories fol-

lowing this model have often turned into effective graveyards

for data. Therefore, data discovery and data sharing must have

associated information that will allow a scientist to concentrate

on creating interesting hypothesis, not on the details of gather-

ing the data and tracking down the methods of data collection.

This may make it more difficult for the data provider, but it is

better long-term insurance that their data may be used again

for other purposes.

Much value can be derived from a data sharing system that

finds and integrates data of many different types and from dif-

ferent sources. For this reason, digital atlases have been iden-

tified as potential neuroinformatics frameworks, as they act as

a "map" one may use to traverse the brain and associated data

of different scales, modalities, and sources. However, if we

wish to have an atlas that acts as such a neuroinformatics hub,

it must be more than a map. It must act as a gateway to large

distributed databases of images, volumes, segmentations, and

other types of spatially-registered data. These databases must

be connected through spatial data registries and services, as

well as standard APIs and vocabularies to make them all work

together. Contributing data within this framework requires

tools for registration, image segmentation, spatial selection,

and analysis. Finally, accessing this system requires viewers

and annotation tools with authenticated access. In summary,

there must be a full and complex infrastructure built behind the

atlas as well as intuitive interfaces for interacting with them.

Several investigators have worked on various aspects of this is-

sue, but it is not usually in an individual biological researchers'

best interest to pursue creating an infrastructure that is exten-

sible to others at the neglect of their own unique research. As

of yet, there remains a gap in all the potential resources and a

full extensible sharing infrastructure in this format. Domain

experts in this field would gladly participate in creating such

an infrastructure, but most do not have the technical resources

to build it on their own. What is needed to create such an

infrastructure is an organizing body that can survey current

practices, help generate standards, and aid in the technical con-

struction of this sort of sharing infrastructure.

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ ]

3. Concepts

To frame the discussion and establish the highest precision of

intent it is first useful to establish concrete terminology.

Digital Atlas

An atlas is a collection of maps or manifolds, traditionally

bound into book form, but also found in multimedia formats

(Wikipedia (3/20/07). As technology has advanced so have

brain atlases transformed from passive paper guides to dynamic

databases at the core of software applications (Toga 00). In

this report, we almost exclusively focus on these sophisticated

digital atlases held in either free-standing software tools or

web-enabled hyperlinked neuroinformatics hubs that act as a

gateway to a collection of databases, metadata catalogs, and

related multimedia documents and annotations that are placed

in a common spatial framework and thus can be juxtaposed

and analyzed together.

Data Repository

A central place where data is stored and maintained (Wikipedia

04/07)

Database

A database can be defined as a structured collection of records

or data that is stored in a computer so that a program can con-

sult it to answer queries and incorporates software to make it

accessible in a variety of ways. The records retrieved in an-

swer to queries become information that can be used to make

decisions. (Adapted from Wikipedia 04/26/07 and the Oxford

English Dictionary)

Spatial Database

"We propose a definition of a spatial database system as a data-

base system that offers spatial data types in its data model and

query language and supports spatial data types in its imple-

mentation, providing at least spatial indexing and spatial join

methods."

(Ralf Gting,

http://www.informatik.fernunihagen.de/import/

pi/papers/IntroSpatialDBMS.pdf

)

Image Registration

Image registration is a process of relating and organizing char-

acteristics of two or more images, so that image data obtained

from different measurements can be discovered, compared or

integrated. Image registration may include organizing image

metadata into an image metadata catalog, as well as placing the

images in a common semantic or spatial framework. This may

be contrasted with Semantic registration that involves verify-

ing image metadata and associated labeled delineations against

established formal ontologies or controlled vocabularies, to

ensure commonality of terms used in image description.

Spatial registration (often also called "image registration")

is the process of modifying spatial characteristics of an im-

age dataset to align it to another image dataset, thus placing

different images into a common coordinate reference frame.

Image registration techniques vary across domains. For ex-

ample, different MR images may be spatially co-registered by

tuning image metadata which includes pixel size, dimensions

and orientation of the image. Alternately, an image may be

transformed into alignment with another image by specifying

pairs of fiducial control points, or by relating the image to a set

of anatomic feature delineations. Technically, spatial registra-

tion procedures may involve a linear alignment or a nonlinear

alignment, which actually warps the images into a common

space. Spatial registration ensures that images may be discov-

ered and queried by spatial coordinates, via an anatomic atlas.

(IZ 05/07)

Spatial Registry or Spatial Registration Services

Spatial registry is a component of image metadata catalog. It

contains information about position, orientation and extent of

registered images, and links image spatial metadata with other

image metadata. A spatial registry is typically organized as a

spatial database: it contains polygonal representations of regis-

tered images, maintains spatial indexing of the image polygons,

and supports spatial queries, e.g. `select images intersecting

with a user-defined shape,' `select images whose centroids are

contained within a user-defined shape,' `select images found

within a 3mm sphere around a user-defined point.' The key

concept is that the spatial registry connects image data with

data annotation enabling effect inquiry. (IZ 04/07, MH 6/07)

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ ]

Annotation

Annotation is the process of associating information content

and knowledge with raw data. Annotations may differ in pur-

pose and complexity, ranging from simple text notes made at

a particular point in a document or in an atlas, to multimedia

composite objects that may include user-defined shapes, docu-

ments, hyperlinks, or other annotations. (IZ 04/07, MH 6/07)

Metadata

Metadata is data about or associated with data used to render a

more precise description or record of its significance. An item

of metadata may describe an individual data item or a collec-

tion of data items and is used to facilitate the understanding,

use and management of data. (Adapted from Wikipedia 04/07,

MH 6/07)

Application Programming Interface (API)

An API is a source code interface that a computer system

or program library provides in order to support requests for

services to be made of it by a computer program. (Wikipedia

04/07)

4. Workshop Discussions

Topics of the workshop included:

An inventory and review of the production of digital ro-

dent atlases (short presentations were given by all par-

ticipants), their significance as models, and their diverse

purposes including:

-The display of gene expression and associated im-

age data (Allen Brain Atlas, Cerebellar Development

Transcriptome DataBase, GenePaint, Mouse BIRN

Atlasing Toolkit-MBAT, MousePat, SmartAtlas)

-Anatomic and genetic variation (Neuroterrain and

the Mouse Brain Library)

-Modeling neuronal circuits (Jan Bjaalie, Raphael

Ritz and colleagues)

-Comparative and functional analysis of behavior in

rodents (Andreas Hess)

-The advanced techniques needed to generate high

resolution data as the backbone of these atlases (G.

Allan Johnson, Gregor Eichele)

A discussion of registration and mapping techniques used

to align data to these atlases and the necessity for having

atlases serve as a common frame of reference to which

data is linked

The applications of digital atlases as gateways to diverse

data types, including gene and protein expression, pat-

terns of neuronal connections, time-series records, popu-

lation surveys, quantitative analyses, and species com-

parisons

Access to stable and persistent databases with tools de-

signed to easily upload and share multiple types of data

as well as access and download this data

The importance of defined and interoperable neuroana-

tomic vocabularies and higher order ontologies (George

Paxinos, Charles Watson) to bridge the gap between dif-

ferent atlases, different species, different data modalities,

scales, and as a consistent usable language between hu-

mans and machines

The need for infrastructure, software tools, and techno-

logical expertise to facilitate data sharing

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ ]

The participants of this workshop reviewed the current work in

the area of atlas-based neuroinformatics and developed a set of

recommendations:

We reviewed the challenges of developing, applying, and

translating consistent anatomical terminologies. Several

examples were discussed including:

-The challenge of defining basic structures, such as

the amygdala, across atlases and data collection mo-

dalities. While the standard Paxinos and Swanson

terminologies may be translated at a D stereotaxic

level in adult rodents, how do we handle situations in

which an experiment dictates that an area considered

"amygdala" actually includes adjoining regions?

-To what extent can homologies between CNS struc-

tures be assigned and defined beyond mouse and rat;

mouse and human; mouse and chicken; mouse and

zebrafish?

-How best can terminologies (and atlases) handle the

complexity of development?

We reviewed experimental complexities and advances in

many diverse areas of data collection:

-New magnetic resonance histology methods
-Function magnetic resonance methods applied to ro-

dents

-High throughput transcriptome and proteome meth-

ods in both adults and embryos

We reviewed the values of atlases in examining experi-

mental data and for tying different types of data backed

by databases into a meaningful context including mi-

croscopy and electron microscopy data, other types of

D volumes, surfaces, neural connections, blood vessels,

gene expression levels experimental information

We reviewed the power and problems of mapping data to

digital atlases, an effect exacerbated by the lack of stan-

dardization

5. Recommendations

Through presentations, discussions, and follow up, this work-

ing group came up with two major recommendations for IN-

CF's role and participation toward achieving the goals outlined

above. These are listed below and followed by a more com-

plete description.

1.Encourage standardization and reproducibility in the

development, maintenance and use of digital atlases by

providing and maintaining a framework and/or infra-

structure for digital atlas data sharing.

.Serve in a guidance and organizational role for larger

collaborative mouse and rat projects.

In addition, this group also made the following more general

recommendations, which may be implemented in conjunction

with or separate from the two above.

Promoting International Recommendations and Stan-

dards (following examples of successful international

standards bodies such as Open Geospatial Consortium

http://www.opengeospatial.org/

)

-Make data sharing, standard API, and atlasing soft-

ware tool recommendations standard practice for

code robustness, support for multiple interfaces, and

better usability.

-Best practices for managing and evolving the infra-

structure in an extensible and scalable manner.

-Gather and disseminate best practices learned across

fields in a digestible format.

Resources for Domain Experts:

-Generate and garner financial resources for interna-

tional nodes to build infrastructure and tools.

-Aid in recruiting competent technical help for groups

that wish to share their resources.

Facilitate collaboration:

-Facilitate communication between tool builders and

users.

-Identify groups for collaboration in areas of ex-

panse.

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ 10 ]

-Identify groups for collaboration in other but related

fields.

-Identify relevant technologies from other fields and

parties that may aid in transferring the technology.

-Provide connectivity with people to help facilitate

this collaboration.

5.1

Infrastructure for Data Sharing

The workshop committee recommends that the INCF support a

vision of centralized standardization and infrastructure related

to the development and maintenance of digital rat and mouse

atlases. Rodent models are ideal for developing this infrastruc-

ture, as they are model organisms for human disorders, geneti-

cally defined, and may yield a wealth of diverse data types.

Also, there are many existing atlases from which to generalize

a standard. The ability to pull together these types of informa-

tion both within and across species, greatly facilitates our abil-

ity to understand disease mechanisms and potential therapies.

As the infrastructure for sharing is developed, it is expected

that it will be extensible to other species and data types.

This proposed environment should facilitate data exchange,

content discovery, and open interchange with existing data sets

and formats. This recommendation would take the form of a

potential set of standards for data sets and methods related to

digital atlasing including:

Ontological and format proposals for a set of archival

mouse and rat related neuroinformatics datasets with re-

lated metadata. These are the "canonical" datasets that

provide a state of the art of rodent informatics imaging

and methods.

Description of the access methods for registration, map-

ping, and testing of the canonical datasets with a new test

dataset.

Use of a controlled open source environment with re-

spect to archiving, access, and updating of the canonical

datasets and methods.

The ability to propose or upload other datasets for con-

sideration in the canonical set.

Reporting on performance and testing of the archival ca-

nonical datasets.

5.1.1 Identification of Canonical Datasets

The set of atlases used as the backbone of this project would

likely include a set of digital atlases currently used by the com-

munity or any new high-quality datasets generated with this

purpose in mind. These core datasets provide a new INCF

standard for mapping and registration, whereas transforms en-

abling mapping of existing atlases into this new standard fa-

cilitate connectivity with legacy atlases and applications. The

standard serves as the benchmark for future mapping, registra-

tion, and annotation efforts. Recommendations by this group

for the initial set of canonical datasets and atlases:

Species/strain:

Inbred strains of mice: primary = CBL/J, secondary

strains include 1S1/SvImJ, DBA/J or other of the 1

sequenced strains.

Inbred rat strains including the sequenced Brown Nor-

way BN strain and/or more widely used strains such as

albino Sprague-Dawley.

Datasets: Provide the fundamental image datasets to which

subsequent image registration and annotation would be refer-

enced. Ideally this set should include:

A multiple scan averaged high contrast D MRI volume

(at least 0 micron resolution).

A histology dataset, most likely a Nissl stain (at least 0

micron isotropic resolution, 1 micron in plane) that is re-

constructed into a D image volume.

Angiographic datasets would be collected in order to map

the vascular system of the brain. This would include:

-A MR angiographic dataset (at least 100 micron iso-

tropic).

-A µCT corrosion cast to visualize vessels down to

10-0 micrometers in diameter.

The registered datasets would be used as the default standard

for digital atlas applications and their comparison. Preferably,

these would be from the same animal with an in-situ µCT of

the skull for the lambda and bregma references.

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ 11 ]

Age: Mouse within the range of 8-10 weeks; Rat within the

range of -1 weeks approximately 0 g.

Sex: 1 male and 1 female of each species/strain, dataset,

and age.

Fully delineated in a manner that enables the translation

of known anatomy to significant structural depth (500+

structures). This should include registration mapping

transforms from other widely used annotated atlases.

The inclusion of other types of canonical datasets over time

might include:

Other species and strains

Different ages including embryonic stages

Other annotations

Marker genes

Neural connections

the spinal cord and the eye

Also the group showed a great desire to coordinate the assem-

bly of a new annotated canonical atlas based on the same ani-

mal scanned from CT (for identification of Bregma), MRI, and

histology.

There should be a curation process and methodology for bring-

ing in any new canonical atlases established by the INCF that

includes domain experts. Ideally, as infrastructure is built,

tools would be offered that make it easier for people to com-

ment on existing annotation in these atlases, or add their own.

5.1.2 Linking of Canonical Atlases

In order to use digital rodent atlases as the backbone for a neu-

roinformatics framework they need to be linked in a manner

that will allow tools to navigate across different atlases and

the data associated with them. The following are suggested

methods for connecting these data:

Semantic mapping

A consistent anatomic structure naming convention

should be adopted, which includes both structure names

and abbreviations (as GO has done for gene names). This

has been a significant issue in the development of neuro-

anatomic atlases to date.

Structure names/Ontologies need to be mapped across at-

lases so it is possible to cross different labeling conven-

tions.

There needs to be a solid mapping between the rat and

the mouse. It is likely this mapping will be semantic first,

and spatial over time.

It was suggested that as the group of canonical atlases are

built to look for common features across atlases. These

would be used as the core reference atlas, a set of unam-

biguously defined features and ontology that most scien-

tists would agree with regardless of their preferences for

how the brain should be delineated or named. These may

be used to cross species.

Spatial mapping

Registered to the same space, or explicitly defined co-

ordinate framework (such as stereotaxic coordinates)

which may also be used to define the framework for new

data

Spatially linked using an infrastructure that allows tools

to translate from one atlas to another atlas

5.1.3 Web-accessible Platform for Access to the
Canonical Atlases

The proposed platform would provide a framework and bench-

mark for the evaluation of new digital atlases, datasets, and

methods. Offering maintained and refereed open source datas-

ets and registration methods to these atlases would allow users

to:

Test and map novel mouse and rat datasets against ca-

nonical atlases of their choosing

Improve existing and implement novel algorithms for

mapping and translation between data sets

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ 1 ]

Sharing data within this framework requires data providers to

upload their data using semantic and/or spatial mapping and

may include whole brain, "chunks" of the brain, and several

different data types. Thus, simple and accessible methods to

enable this mapping must be offered. A specific architectural

review of this proposed site and process should be performed

to ensure optimal design and implementation strategies as well

as its relationship to existing repositories/archives and datasets.

The primary role of the INCF in this regard would be to set the

policy and review recommendations for contributed material.

Specific recommendations by this working group include:

This platform would provide a test bed and archival stor-

age system for registration methods against the canonical

datasets.

An architecture and site design review should be held for

the housing of canonical datasets and mapping require-

ments.

INCF would facilitate collection of registration methods

and offer open access to their functionality by working

with domain experts. Ideally, this will be offered through

a simple web-accessible platform that helps step a user

through the registration process in as close to an auto-

matic process as is possible and practical. At the end of

this process, the user must have the ability to automati-

cally "publish" their dataset.

Even with help, this group understands that registration

is still a complex issue, and it may be necessary for INCF

to assist people with registration (even if on a fee-based

basis).

There should be a standard of usability and a curation

process and methodology for bringing in any new regis-

tration processes into the infrastructure that includes im-

age registration domain experts.

5.1.4 Queryable Interfaces

It is vital to this project that data uploaded in this framework

be stored in a manner that enables easy access, rapid content

discovery, and comparison. It is advisable that the design and

development of this interface be performed in concert with the

atlas data upload aspect of the project.

The databases (either new or existing-see section below)

housing either the atlases or the newly uploaded data

should supply data in a format of a common data model

based call-service signature or syntax used for query call

for that data type (API and/or web-services).

Standard APIs will be needed for three levels of interop-

erability:

-To retrieve different types of data from atlas servers
-To query atlas metadata catalogs
-To exchange information between atlas applications

These common data model-based APIs and web-services

should be defined by the experts storing and using that

particular type of data (i.e. MAGE for microarray gene

expression data, see references below). Since there isn't

a standard that has yet been adopted by that community,

it is recommended that a global working group be created

to develop one for that data type. This should include

neuroscientists, database developers, query tool builders,

and people with experience with developing this in other

fields.

Ideally the INCF would also provide a simple visual aid

for web-query of this data. More specifics for the design

of the query tools will also depend on the outcome of the

survey of current software tools and practices.

The issue of allowing users to create their own atlas de-

lineations/annotations is complex and should be consid-

ered, although this would have to be implemented in a

controlled manner. A primary goal of the infrastructure

is to allow researchers flexibility while still operating

within a set of standards that facilitate collaboration and

sharing.

5.1.5 Existing Databases

There are a wide variety of related and supporting databases

and information sources that are presently available and perti-

nent to this effort. Section .1. discusses how common data

model based APIs and web-services will be key tools for op-

erators of databases to may make their own resources avail-

able to the community query tools. Over the long-term, with

the development of the appropriate resources to facilitate data

sharing, different data modalities may be shared through this

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ 1 ]

environment such as electrophysiology, cell imaging, fMRI

and disease phenotypes. To support these data sharing efforts,

there must be tools, methods, and tutorials offered to the public

that allow existing databases to link to this infrastructure.

The INCF should play a key role in this process. As these

standards and methods are developed, INCF should make rec-

ommendations, tools, and tutorials available for how someone

may be able to share their database with this infrastructure in a

manner so that it is accessible for query and access.

Appropriate technical and domain people sponsored by the

INCF may work with individual laboratories to implement the

appropriate steps needed to share their data within this frame-

work.

5.1.6 Software Tools

This proposed infrastructure requires a toolset that facilitates

data upload, data registration, data query, and interoperable re-

sources. As there are already many tools available in the com-

munity in these areas it is possible that tool providers will be

interested in participating in this infrastructure, and that much

existing infrastructure can be leveraged. An INCF approved

study group would develop a plan for this framework based on

a survey of what is currently being developed in this and other

relevant fields. It is also essential that software developed for

this infrastructure be open source and accessible for others to

use and modify for their own needs.

5.1.7 Additional INCF Roles

While this is an ambitious project, this working group sees

some very unique roles the INCF may play in facilitating the

development of this infrastructure:

Information gathering:

-Survey neurobiologists who would use an infrastruc-

ture such as this about what data types should be the

first areas of focus and what queries and integration

scenarios they envision

-Survey of available atlases and potential role they

may fit in this architecture

-Based on these surveys, determine the canonical data

types and atlases and possibly "canonical" databases

or data sources

-Potential survey methods include examination of

current resources and users per week, citation num-

ber, google scholar benchmarks, mailing list feed-

back, informal polling (asking experts and users etc.

and propose alternatives)

Oversee working groups involved in:

-Developing standards for data sharing in the areas

outlined above

-Drawing up a technical specifications document that

outlines how to implement this infrastructure and

process

-Work with international groups to determine stan-

dards for facilitation of integration and implementa-

tion

Develop database resources and ensure longevity and ar-

chival integrity of these resources

Provide resources to aid in programming within the in-

ternational nodes

5.2

Large Data Collection Project

Another role that the INCF is uniquely poised to play is as the

parent organization for large data collection projects similar

to HUGO's role for the human genome project. This might

include communication with funding sources, lobbying, and

potentially housing of the database. The INCF would act as a

critical guide and overseer for large scale high resource proj-

ects for which the value to neurobiology is high but for which

there are concomitant risks.

As a project develops, the INCF could orchestrate planning

groups that would develop concrete road maps and milestone

checkpoints. As components of the project are identified,

INCF would aid in recruitment of the expertise needed to fill

the various roles in the project.

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007

[ 1 ]

5.2.1 Example of a Representative Project

Recommendation for a Spatial Atlas of Gene Expression in

the Rat Nervous System (as outlined by Gregor Eichele)

The rat is a widely used mammalian model system often used

to study a broad spectrum of physiological, neurophysiologic

and behavioral questions. However, much of this research is

being adapted to the mouse, chiefly because of the relative ease

by which genes can be manipulated in the mouse. Much ef-

fort goes into adapting a wide spectrum of assays from larger

mammals to the mouse, but there are significant natural limi-

tations to this such as tissue amounts, continuous monitoring

of physiological parameters, and an intrinsic limitation of the

mouse as a "smart" species. Instead of retooling the highly

developed field of physiology, it seems more practical and less

costly to re-determine the comparable genomic information in

the rat. A first step has been completed with the recent comple-

tion of the rat genome sequencing project. Other techniques

such as RNAi knock-down may make species such as the rat

genetically tractable. Here we propose the production of a gene

expression atlas of the rat nervous system.

The current project proposal is for a novel high resolution

spatially mapped rat genome in situ hybridization atlas. The

availability of key data modalities, such as the wealth of elec-

trophysiological measurements in the rat, makes that organism

the natural choice for a project that synergizes the connection

of genetics, anatomy, and electrophysiology. In particular, the

wealth of cell type specific electrophysiology in the rat such as

the characteristics of ion channels will need to be coupled with

gene expression data to untangle the computational complex-

ity of the brain. Based on existing neurobiological resources,

the genes of the rat genome can be prioritized so as to enable

maximal coverage of significant expression patterns and cell

types. Recent developments in neuroinformatics can be uti-

lized to obtain maximal spatial resolution and image mapping

accuracy.

5.2.2 Advantages of this Project

1.This is chiefly an engineering project that can be subdi-

vided in clearly identifiable deliverables and milestones.

.It is a highly interdisciplinary project covering experi-

mental biology, high throughput lab methods, imaging

and image processing, database technology and anatomi-

cal annotation.

.Many of the goals of this project have been attained or

addressed in a preliminary way in the mouse via the Al-

len Brain Atlas (ABA).

.Several novel experimental and computational strategies

not available at the time when the ABA was generated

can be implemented.

5.There would be enormous benefit if this could become

a multinational project similar to the human genome se-

quencing project.

6.Cost and time frame are predictable and finite.

.The project uses in part existing resources but new ones

could be added (e.g. a new interactive web database

housing the expression patterns).

.Genes can be prioritized by importance based on what is

known from the ABA.

9.The price tag is in the range of 50 million .

5.2.3 Implementation

In order to implement this project, planning groups represent-

ing the relevant components of the scientific community would

need to be established. These groups would then draft the re-

quired concept papers and road maps. The INCF could poten-

tially help in the organization and recruitment of appropriate

parties. Example working groups include:

Rat genomics experts to annotate genes and design tem-

plates for riboprobes

Technology for nd generation tissue sectioning enabling

improved section registration (e.g. high-throughput block

face imaging)

Logistics of tissue collection (stages of development,

strain)

Logistics large-scale template and probe generation

Data collection (in situ hybridization and image capture,

multiplexing for genes allowing detection of multiple

transcripts per tissue section)

Nature Precedings : doi:10.1038/npre.2007.1046.1 : Posted 19 Sep 2007