Timing is a critical element of most behaviors and of many cognitive tasks, yet we have little understanding of how neural circuits estimate, remember and control time intervals or generate temporal sequences of activity. A new class of network models will be studied that show great promise for uncovering the dynamic mechanisms, operating at the neural circuit level, supporting sequence generation and timing computations. These models will be built from a generic network structure through the addition of tuned feedback loops. This allows the same network to perform many different tasks and resembles the reconfiguration of existing circuits that occurs when a new task is learned. Constructing realistic models that perform complex tasks raises the possibility of making deep connections between the dynamic mechanisms operating in model and real neuronal networks. Three steps are required to fulfill this promise, and these are the three specific aims of the proposal. First, a study will be undertaken to determine what motor and cognitive tasks network models are capable of performing and to relate their level of performance to that of animals and humans performance analogous tasks. Second, the models will be made realistic enough to make definitive and predicted statements about experimental data. Third, an approach will be developed so that the network models can function as a tool to uncover the dynamic mechanisms operating in real neural circuits. For this purpose a collaboration has been established with experimental colleagues acquiring multi-electrode recordings from monkeys performing delayed-reaching tasks. Successfully accomplishing these three aims will significantly advance understanding of how timing arises from neural circuit dynamics and lead to hypotheses about circuit malfunctions that cause defects in timing estimation and movement initiation and control.
The proposed research will use network models and parallel analyses of models and experimental data to reveal the dynamic mechanisms by which motor sequences are initiated and carried out and other timing-related tasks are performed. Parkinson's disease is characterized by difficulty in initiating voluntary movements and by abnormalities of timing estimation. This work thus has the potential to illuminate our understanding of motor and cognitive tasks involving timing in both healthy and diseased states.
|Kennedy, Ann; Wayne, Greg; Kaifosh, Patrick et al. (2014) A temporal basis for predicting the sensory consequences of motor commands in an electric fish. Nat Neurosci 17:416-22|
|Le Masson, Gwendal; Przedborski, Serge; Abbott, L F (2014) A computational model of motor neuron degeneration. Neuron 83:975-88|
|Fink, Andrew J P; Croce, Katherine R; Huang, Z Josh et al. (2014) Presynaptic inhibition of spinal sensory feedback ensures smooth movement. Nature 509:43-8|
|Wayne, Greg; Abbott, L F (2014) Hierarchical control using networks trained with higher-level forward models. Neural Comput 26:2163-93|
|Kato, Saul; Xu, Yifan; Cho, Christine E et al. (2014) Temporal responses of C. elegans chemosensory neurons are preserved in behavioral dynamics. Neuron 81:616-28|
|Sussillo, David; Barak, Omri (2013) Opening the black box: low-dimensional dynamics in high-dimensional recurrent neural networks. Neural Comput 25:626-49|
|Barak, Omri; Sussillo, David; Romo, Ranulfo et al. (2013) From fixed points to chaos: three models of delayed discrimination. Prog Neurobiol 103:214-22|
|Babadi, Baktash; Abbott, L F (2013) Pairwise analysis can account for network structures arising from spike-timing dependent plasticity. PLoS Comput Biol 9:e1002906|
|Schaffer, Evan S; Ostojic, Srdjan; Abbott, L F (2013) A complex-valued firing-rate model that approximates the dynamics of spiking networks. PLoS Comput Biol 9:e1003301|