Some small networks of neurons are remarkable for their ability to execute multiple functions. It has been a challenge to understand what the rules are for integrating feedback into features such as phase relationships among the firing patterns of active neurons, and how a small network can 'switch' from one characteristic behavior to another. Although research is clarifying the cellular and molecular mechanisms of learning, it has been more difficult to understand how changes in the properties of individual neurons change the activity of a whole neural circuit, and in turn alter an animal's overall behavior. This project is a collaboration using computational, theoretical and experimental approaches to analyze the feeding behavior of a marine mollusk, the sea hare Aplysia. This animal ingests food with rhythmic rasping and sucking motions of a jawless buccal mass, run by a network of about 130 motor neurons and interneurons; if potential food is sensed by its physical properties as inedible, the pattern of muscle activity changes from ingestion to food rejection. The overall goal is to determine how small changes in the properties of individual nerve cells create the large changes in feeding behavior that are observed after learning. Specific Aim 1 is to construct a kinetic mathematical model of the buccal mass (finite-element method), its neural control (continuous-time recurrent neural network (CTRNN), with Hodgkin-Huxley models for motor and sensory neurons), and inedible food, and to conduct experimental studies to improve the understanding of each of these components of the model. The focus of this modeling is to reproduce the changes in motor pattern observed during repeated encounters with inedible food. Specific Aim 2 is to develop a numerically optimal controller for the new kinetic model, and use it to predict the effects of small changes in timing, phasing and intensity of neural input on the behavior generated by the buccal mass. The focus here is on how the biomechanics of the periphery influences the design properties of the neural controller. The models developed in Specific Aims 1 and 2 will be used to analyze the contributions of individual neurons to the shifting coalitions that stabilize the rhythmic behavior, and to predict the importance of local changes in synaptic strengths or intrinsic properties to the overall dynamics of the neural circuit both in isolation and when it is connected to the biomechanical model. Under Specific Aim 3, experimental studies will be designed to test these predictions from simulation studies. These experimental studies will record neural activity in intact animals as they learn that food is inedible, and in reduced preparations (that show feeding-like movements) after perturbations of the activity of specific nerve cells.
This work will have an impact beyond computational neuroscience and behavioral neuroscience, to invertebrate physiology, engineering, robotics and control systems. First, it is likely to generate principles for understanding the effects of localized changes in neural activity on an animal's overall behavior. Second, it may suggest design principles for devices that can persistently pursue a specific goal despite distracting inputs, and at the same time be remarkably flexible and change behavior if an appropriate stimulus occurs in the correct context. . Third, these principles are likely to serve as the basis for novel biologically-inspired robotic and control devices. In addition, students and collaborators will be involved together in cross-disciplinary approaches and techniques that will enhance training for the next generation of scientists.
