Skip to content


    Skip to content

    NEUROSCIENCE | Contents | Next

    Making Headway in Computational Neuroscience

    PROJECT LEADER
    Mark Ellisman, UC San Diego

    PARTICIPANTS
    Maryann Martone, UC San Diego
    Gwen A. Jacobs, John P. Miller Montana State University
    Scott E. Fraser, Russell E. Jacobs
    Caltech
    Paul Thompson, Arthur W. Toga
    UCLA
    David C. Van Essen, Heather A. Drury, Marcus Raichle, Jerome Cox Washington University

    N euroscience is the study of brains and central nervous systems, ranging from the "simple" nervous systems of invertebrates such as the cricket to the complex neural machinery of humans. It has long been a broad and various science, with at least a dozen well-defined subdisciplines. But over the last two decades, computational methods have cut right across the older divisions. High-performance computation is serving as a new, unifying, and accelerating force in the field. NPACI's Neuroscience thrust area was organized to make more visible and carry forward the promise of computational neuroscience.

    The growth of neuroscience as a discipline is measurable in many ways, but one stark and simple measurement is given by Floyd E. Bloom in Chapter 1 of the new, thousand-page textbook, Fundamental Neuroscience (Zigmond et al., Academic Press, 1999). According to Bloom and colleagues, attendance at the annual Society for Neuroscience meeting has jumped from 1,100 in 1971 to more than 30,000 at the 28th annual meeting in 1998.

    "Even the latest textbooks have not caught up with the computational influence," said Mark Ellisman, leader of the Neuroscience thrust area. The Zigmond text is an example. Chapters are devoted to cellular and molecular neuroscience; nervous system development and structure; sensory, motor, and regulatory systems; and behavioral and cognitive neuroscience.

    "The ability to quantify, compute, and visualize the parameters of these subdisciplines has completely changed their relationship," he said. "Take the area of the structure of the nervous system, treated in textbooks from the point of view of an anatomy long known and settled. Computation has opened up myriad new vistas for neuroanatomy alone, to say nothing of its vast influence in cognitive neuroscience."

    DATA COLLECTION, ANALYSIS, AND MODELING

    FEDERATING BRAIN DATA

    CEREBRAL CORTEX SUMS


    The Continuum of Neurobiological Structure
    Figure 1. The Continuum of Neurobiological Structure
    The "cascade" of scales found in the nervous system starts at the molecular level (upper left) and goes to the largest (complete brain and whole organism at lower right). The Neuroscience thrust links neurobiological data across the scales using advanced computational capabilities to enable new understandings of brain structure and function.



















    DATA COLLECTION, ANALYSIS, AND MODELING

    Ellisman's lab, the National Center for Microscopy and Imaging Research (NCMIR) at UC San Diego, has been a partner of SDSC for nearly a decade, and neuroscientists from the United States and elsewhere have long taken advantage of the computational expertise gathered by Ellisman and the late Stephen Young at NCMIR.

    With support from the NIH National Center for Research Resources, the lab has been pioneering the discipline Ellisman terms "telescience," the use of national- and international-level instrumentation controlled from remote locations. The lab's 400 kV intermediate-high voltage electron microscope (IVEM) is such an instrument. The latest generation of software for remote control of the microscope was demonstrated in August at a NASA Research and Education Network workshop, during which the IVEM in San Diego was controlled from San Jose in northern California. Last March, the lab participated in a trans-Pacific demonstration of telescience, remotely controlling a $50 million, ultra-high-voltage electron microscope, the world's largest and most powerful, at Osaka University.

    When it came to organizing the NPACI thrust area, Ellisman and NPACI leaders sought collaborators who could help in three areas. First and foremost was the area in which such advances as telescience receive their justification--the area of data collection, distribution, and analysis. Neuroscience needs to federate all the data that are being gathered from the new and old imaging instruments of the neurosciences. The IVEM is one example, but significantly more data are coming from various kinds of brain imaging--computed tomography, magnetic resonance, positron emission tomography, to name a few. "Large data caches are being built and maintained around the country, and we sought to combine some of these in an accessible way," Ellisman said. The result is the project in federating brain data.

    The second area of concentration was the advancement of data analysis or data refinement itself by computational means. The brain mapping project is discussed in the next article. Finally, Ellisman sought collaboration in raising the sights of computational neuroscience, where a lively traffic in neural network modeling has been going on for over a decade. "The models are becoming complex enough to represent the electrical and chemical processes going on in brain tissue, well beyond the single-neuron or several-neuron network," Ellisman said. "And the modelers are seeking help in entering the realm of high-performance computation." The neural modeling and simulation project is discussed in the third article in this section.

    Ellisman notes that all of the participants in the thrust area have well-established labs and research groups and that many have grants under a large NIH initiative for study of the human brain. "We have been fortunate to find so many groups able to stretch themselves and collaborate with the computer and computational science communities," he said. "They are playing what I think will prove to be a very large role in revolutionizing their own field."

    Top| Contents | Next

    FEDERATING BRAIN DATA

    Ellisman and Maryann Martone, assistant director of NCMIR, have been joined in the project of federating brain data by a number of leading neuroscience groups. John Miller and Gwen Jacobs, directors of the three-year-old Center for Computational Biology (CCB) at Montana State University, are targeting the common cricket, an insect whose neural system--very much the same from one individual to the next--provides a baseline for the study of fundamental neural capacities. "Crickets have a relatively simple, mappable neural anatomy," Jacobs said. "Yet their systems are capable of sophisticated computations similar to those found in more complex brains."

    In developing a database focused on the structure and function of invertebrate nervous systems, Jacobs, Miller, and other CCB members have also been able to construct a sophisticated interface for posing database queries. In the NPACI project, they are working with data from Ellisman's lab as well as their own and coordinating an investigation of database issues in neuroscience generally. "Our objective is to develop efficient schemas that can scale with the complexity of nervous systems in higher organisms and with the advances in data-intensive computing," Jacobs said. The CCB, the NCMIR lab, and labs at UCLA and Washington University have installed the SDSC Storage Resource Broker system developed in the Data-Intensive Computing Environments thrust area to facilitate the sharing and transmission of data, and they are developing a Neuroscience Data Interchange System.

    A major participant in the data project is the Laboratory of Neuro Imaging at UCLA, led by Arthur W. Toga. Toga, Paul Thompson, and their colleagues have constructed extensive and accurate brain atlases for many organisms--including mice, macaques, and humans--based on high-resolution imagery of microtomographically-sectioned brains. They have concentrated on tools to make the atlases probabilistic, by averaging over many subjects, and deformable to map the brains of new subjects (see p. 4). "These are very large data caches indeed," Toga said. "Our objective is to make them available over the Net within a superstructure or federation of databases with common means of access."

    Also participating are two laboratories at Washington University in Saint Louis. One, the lab of Marc Raichle at the Washington University Medical School, specializes in positron emission tomography (PET) studies. The second is the Center for Imaging Studies, led by David C. Van Essen and Heather A. Drury. Their Java-based Surface Management System (SuMS) is a database and graphical user interface organized for the study of the cerebral cortex in mammals.

    Top| Contents | Next

    CEREBRAL CORTEX SUMS

    "Mammalian cerebral cortex is dauntingly complex from a number of perspectives," Drury said. The surface area is that of an extra-large pizza, she notes, but folded into a complex and irregular set of convolutions. In this configuration, cortical tissue governs hundreds of brain processes, including the all-important processing of visual signals. The high degree of individual variability in cortical geography and functional organization compounds the complexity. The Van Essen lab has developed techniques for surface-based visualization of macaque and human cortexes that deform the cortical covering into intercomparable flat maps, using surface reconstruction programs called CARET and FLATTEN, developed in the lab. This information is accessible through SuMS. The lab is also using surface-based warping methods and the Visible Man atlas (National Library of Medicine) to construct a surface-based atlas of cerebral cortex.

    Another lab involved in the Federating Brain Data project is the Caltech Center for Biological Imaging, directed by Scott Fraser, Anna Rosen Professor of Biology. Fraser and colleague Russell Jacobs also build tools to process, catalogue, and analyze data from two-photon laser microscopy and high-resolution functional magnetic resonance imaging. They are interested in neural development from embryo to adult in a variety of organisms, including fish, frogs, chicks, mice, and mouse lemurs. "We're trying to close the loop between data collection, rendering, and analysis," Fraser said, "and then to carry these tools to the high-performance computational Grid."

    Fraser expressed the general enthusiasm of the participants in the data federation project. "Mark Ellisman looked for people who enjoy collaborations and who are really open to solving problems as a team," he said. "That describes us, and it describes our colleagues at UC San Diego, Montana State, UCLA, and Washington University. That will be essential, not only in solving the technical problems we face as separate research groups, but also in bringing our entire discipline onto a new level of scientific and technical competence."

    The data federation project has already met many milestones for improving connectivity between the collaborating labs and the large data caches are now being populated. "Metadata catalogs are under construction for the data caches," Ellisman said. "Our next hope is to demonstrate the ways in which our nascent computational neuroscience infrastructure permit us to ask--and get answers to--deeper and more far-reaching questions about brains and their structures and functions." --MM *

    Top| Contents | Next