Neuroscience Thrust Area Providing Access to Exploding Amount of Brain Data

PROJECT LEADER
Mark Ellisman, Professor of Neuroscience, Director, National Center for Microscopy and Imaging Research, UC San Diego; Leader, Neuroscience thrust area, NPACI

euroscientists want to understand how the brain works. This task requires integrating information from a large number of disciplines, each focusing on different aspects of the brain--from the molecular to thewhole brain and its behavior. In recent years there has been an explosion of information about the brain. New laboratory, experimental, and computational techniques are making it possible to refine and reveal information about biological structures ranging in size from molecules to nerve cells to regions of the brain. Advances have also been made in describing the dynamics of brain function at levels ranging from biochemical processes to nerve cell interactions and even to human thought.

The strategy in constructing the NPACI Neuroscience thrust area has been to target the most relevant and challenging technological impediments to progress in brain research that may be surmounted using NPACI technologies and infrastructure. As more information becomes available, there is a great demand for approaches to handling large-scale data, methods for accessing and merging information from databases, and advanced tools for data exploration and modeling.

Modeling programs, for example, are being called upon to incorporate more neurons, each increasingly realistic. Such large-scale federated databases and more intricate simulations constitute important testbeds for NPACI's high-performance computation and data facilities. "These needs have defined the projects the Neuroscience thrust area is pursuing," said Mark Ellisman, professor of neuroscience at UC San Diego and the Neuroscience thrust area leader. "This issue of ENVISION describes the challenges we face and the progress we hope to make."

Figure 2: Cricket Nerves

FEDERATING BRAIN DATA

Neuroinformatics--techniques for sharing and extending the knowledge base of neuroscience--is a science in its infancy. "We have been collecting all kinds of data from brain cells, tissue, and whole brains for many years, in many laboratories," said Ellisman, leader of the Federating Brain Data project. "Our goal in this project is to develop such databases and the tools to access and analyze them in a high-speed, networked, collaborative environment." Ellisman is also director of the National Center for Microscopy and Imaging Research, an NIH Research Resource located at UC San Diego.

"The molecular geneticists are way ahead here," said Gwen A. Jacobs, co-director of the Center for Computational Biology (CCB) at Montana State University. "Their data are relatively simple, fundamentally text-based lists of base or amino acid sequences. Neuroscientists have 3-D image data, data from time series measurements of neural function, as well as behavioral data from the organisms whose nervous systems are under study," Jacobs said. "Our data collections span multiple levels in spatial resolution, geometry, and time. Moreover, there are no readily available, off-the-shelf tools with which to examine and cross-correlate data from one set to another."

The Montana State group has very large data sets deriving from their study of the nervous system of the cricket. "Crickets supply a simple sensory system, standard from one individual to the next, unlike the human brain," Jacobs said. "With an anatomical database from the cricket, we have a baseline system for exploring neuronal interactions across a species on the same coordinate system." (Figure 1)

Jacobs, John P. Miller, co-director of the CCB, and other members of the Montana State lab, including Sandy Pittendrigh and Josef Svitak, have developed a database focused on the structure and function of the invertebrate nervous system. It incorporates a sophisticated 3-D visualization interface for posing database queries. Montana State is coordinating discussion of database issues among the project members.

A brain-mapping group at UCLA, led by Arthur W. Toga, and three groups at Washington University in Saint Louis, led by Jerome Cox, Marcus Raichle, and David C. Van Essen, focus on data from the more complex brains of macaques and humans, which have more variable structures across individuals. These groups have developed data structures and exchange formats for large-volume cryoanatomical data sets (Figure 2).

As an initial step in facilitating data exchange between sites, NPACI support has enabled upgrades to equipment in the participating laboratories. The Toga lab at UCLA added 650 gigabytes of disk storage to its SGI Reality Monster, and a 400 gigabyte data cache was implemented at Washington University, where the groups have begun installing a vBNS link to their labs.

THE SUMS OF THE PARTS

Van Essen, Heather Drury, and James Dickson at Washington have developed a Surface Management System (SuMS), which is a combined database and graphical user interface for dealing with explicit surface reconstructions and associated experimental data obtained in studies of the mammalian cerebral cortex. The SuMS interface is Java-based for machine independence and uses Sybase System 11. Recent advances in computerized neuroanatomy make it practical to generate surface reconstructions from individual hemispheres using structural MRI or cryosection images.

"Surface reconstructions are needed to localize neuroimaging or neuroanatomical data accurately in relation to the unique pattern of cortical folds associated with each experimental hemisphere," Van Essen said. In addition, to compensate for individual variability in cortical folding, it is desirable to bring data into register on a common substrate--a surface-based atlas--using warping methods that respect the topology of the cortical surface.

SuMS, currently in an early prototyping stage, is designed for use in conjunction with software for surface visualization and reconstruction (CARET) and surface flattening (FLATTEN).

The Montana State investigators and the Ellisman group, including Stephen Young and Maryann Martone, have circulated a detailed questionnaire to the Neuroscience thrust area participants, and a Web site will soon be created at Montana State to catalog responses as well as to encourage responses from other members of the neuroscience community. The project is also working with the Data-Intensive Computing thrust area to develop a prototype Neuroscience Data Interchange System based on the SDSC Storage Resource Broker.

The work being carried out by the project members covers a very wide variety of neural systems and preparations, fixed and living, from invertebrates to mammals, and an even wider variety of imaging techniques--electrophysiological recordings from neurons, light microscopy, electron microscopy, two-photon laser scanning microscopy, and functional magnetic resonance imaging, to name a few. "Our aim is ultimately to include all these databases in a superstructure, or federation, of databases with common means of access and query," Ellisman said.

NEW COLLABORATION WITH CALTECH

The recent participation of Scott E. Fraser, Anna Rosen Professor of Biology and director of the Center for Biological Imaging, and his colleague Russell E. Jacobs at Caltech, will take advantage of this group's focus on melding complementary imaging modalities. Two-photon laser scanning and functional magnetic resonance imaging techniques are yielding time-resolved information on the emergence of biological order in the developing embryonic central nervous system of species including quail, mouse, and monkey.

The Caltech researchers expect the work to result in interactive atlases of neuronal development and a set of registration and alignment tools for multiple images of the same species in the same stage of development. "This is very important to us all," Montana State's Jacobs said, "because in addition to helping us understand how the brain areas and connections are formed, we know that the external environment during development before and after birth plays a very large role in the normal 'wiring' of brains."

"By working with the NPACI thrust areas, we are bringing together elements of a discipline that were formerly isolated from one another," Ellisman said. "The job of the brain is to process information from and redistribute it across the nervous system, and thus neuroscientists have tasks before them that resemble the tasks performed by the object of their studies. To make progress, they must be able to integrate the information they have acquired. The improvements in computational infrastructure dictated by this scientific necessity will be of use to scientists in many other disciplines as well." --MM