 |
 |
|
|
|
||
| |
||
INTRODUCTION
The goal of the BITS (Biological Information Technology and Storage) Program is the creation of new approaches for the solution of complex computational problems. Biological systems constitute a rich source of ideas because they routinely solve extremely complex computational problems ranging from subcellular genetic and metabolic function to tissue level integration to whole animal behavior. Arguably the most successful biological information processing system is the brain; our understanding of it continues to inspire solutions to problems ranging from pattern recognition to motion control. The hypothesis of this proposal is that additional computational insights and strategies are to be gained as a result of the process of understanding signal processing and coding in small networks of neurons that have been deliberately designed for computational function.
A confluence of technologies makes this a particularly attractive time for progress toward NSF/BITS goals. These critical technologies are: (a) microlithography, with which materials can be patterned on a scale commensurate with the dimensions of neurons for reproducible networks; (b) neuronal cell culture, including growth of pure neuronal populations in serum free media; and (c) microelectrode array technology for neuro-electric recording and stimulation; and (d) information and analysis technologies for assessing neuronal computation. These technologies make it possible to design a reproducible, reliable, robust, yet living in vitro neuronal network – a "neuronal network in a dish". Advantages of the system include access to individual cells or small groups of cells electrically, chemically and with light and electron microscopy. More importantly for this proposal, the experimenter may choose the complexity of the system, ranging from a few individual cells to a large network, and still have the tools to analyze its behavior.
Work toward neuronal networks in a dish has proceeded so as to be optimistic for its future as a refined testbed for in silico models of neurobiological function. To succeed, however, the following obstacles must be overcome: experimenters must be able to design networks that are robust, reproducible and reliable; simple, standard circuits must be well-characterized electrically and in terms of input / output activity; and computational analysis tools must be brought to bear to interpret the data. If these basics are achieved, then the circuits will be useful test beds for hypotheses about transmission and modification of information by neuronal circuitry and we will gain new insights into how real neurons implement information learning paradigms The new insights may be translated into solutions to problems of computation and information transfer; perhaps the lessons learned will be useful as in the PI’s directional hearing aid project (see CV); perhaps they will eventually have impact as far reaching as artificial neural networks and analog neural VLSI.
Accordingly this proposal is divided into the following specific aims:
(1) The creation of Reliable Geometric Neuronal Circuits
(2) The creation of Reliable Functional Neuronal Circuits
(3) The understanding of Neuronal Information Processing