Integration and Differentiation: Changing Scientific Publication and Communication with Models.
Dr. James M. Bower
Professor of Computational Neuroscience
University of Texas San Antonio
University of Texas Health Science Center San Antonio
The current form of scientific journal publication has remained largely unchanged since it was invented in 1665 as a means to describe (and control) the publication of scientific information and theory. For most of the ensuing 345 years theory in particular has generally been built around relatively small numbers of often closed form equations of the kind made famous by physics and also readily publishable in short-from scientific articles. While this form of description may capture the essentials of planetary motion, it is increasingly unlikely to suffice as a first or perhaps even final description of complex systems like those represented by biology in general or the brain in particular. With the growth in supercomputing and software systems for modeling, many brain models have already grown and expanded to a level of complexity that makes them nearly impossible to describe, replicate, or even understand using standard scientific journal formats. In this talk, Dr. Bower will describe a new form of publication based on models themselves, rather than text describing models. Being constructed as part of the GENESIS 3.0 project, the G3-publication system is designed not only to publish full complete functioning models, but also to foster collaboration and communication as well as education within the neuroscience community. Dr. Bower will illustrate the functionality and capability of this new system using the multi-compartmental realistic cerebellar Purkinje cell model developed in his laboratory using GENESIS 1.0 more than 15 years ago (De Schutter and Bower, Proceedings of the National Academy of Sciences, US, 91: 4736-4740 1994). This model has become one of the first in the history of neuroscience to prorogate to multiple laboratories and modeling systems, becoming, in effect, one of the first “community” models in neurobiology. This model was also the first large-scale realistic model to run on a parallel supercomputer, as it was implemented on the Intel Touchstone Delta, the prototype for the Intel Paragon. Dr. Bower will illustrate how models are described and compared using the G3-publication system, as well as procedures for tracing model lineage and assigning scientific credit and attribution. He will also discuss the possibilities for semi-automated review and annotation of submitted models providing several examples of errors in his own published models that would have been detected by this system. In conclusion, Dr. Bower will consider how model-based publication is likely to change neuroscience itself, providing for the first time, a solid shared theoretical neuron-level foundation for understanding brain function. The work described in this talk is supported by the NINCDS of the U.S. National Institutes of Health.