Terrence J. Sejnowski
e-mail: tsejnowski@ucsd.edu |
 |
The research in this laboratory ranges from experimental studies
of the biophysical mechanisms underlying neural computation to
large-scale cortical models of visual processing and sensorimotor
coordination. The ultimate goal of this research is to provide
linking principles from neural mechanisms to behavior. Computational
neuroscience is a relatively recent approach to understanding
how nervous systems represent, process, store, and act upon information
that is latent in the environment or is
expressed genetically through developmental mechanisms.
The temporal processing of information in a neural system depends on both the intrinsic properties of the individual neurons composing the system and on the synaptic interactions between them. In the cerebral cortex, for example, neurons differ in their morphology, response properties, and connectivity with other neurons. Experimental and modeling techniques are used to explore the biophysical basis for the firing patterns observed in cortical neurons and the influence of neuromodulators.
Neocortical neurons display a wide variety
of dendritic morphologies, ranging from compact arbors to highly
elaborate branching patterns. Electrical recordings from these
neurons have revealed a correspondingly diverse range of intrinsic
firing patterns, including non-adapting, adapting, and bursting
types. This heterogeneity of electrical responsivity has generally
been attributed to variability in the types and densities of ionic
channels, but compartmental models of reconstructed cortical neurons
display the entire spectrum of observed firing patterns by varying
only the dendritic geometry. These results suggest a causal relationship
for the observed correlations between dendritic structure and
firing properties and emphasize the importance of active dendritic
conductances in neuronal function.
When fluctuating current is injected into a cortical cell, the repetition of the irregular spike train is less than a millisecond. This is consistent with the hypothesis that cortical neurons could encode information in the timing of spikes. Neuromodulators such as acetylcholine, however, are known to reduce the adaptation of spike firing in cortical pyramidal neurons by blocking potassium currents responsible for the afterhyperpolarization, thereby altering the spike timing. In slice experiments, application of a cholinergic agonist did not change spike timing by more than that observed under control conditions. Additional spikes were generated because the neuron responded to smaller fluctuations that previously had not elicited a spike.
In addition to these cellular studies, advances have also been made in analyzing the extracellular field potentials generated by large populations of neurons. A new statistical technique has been used to probe the effects of attention and memory on the temporal sequence of neural activations that occur in response to briefly presented stimuli.
Most learning algorithms based on detecting correlations between presynaptic and postsynaptic activity are sensitive primarily to second-order statistics of the input. A new unsupervised learning algorithm was recently introduced by this laboratory that is sensitive to higher-order statistics. This new algorithm, called Independent Component Analysis (ICA), can be used for blindly separating a set of linearly-mixed signals to recover the original, statistically independent sources.
Functional magnetic resonance imaging
allows localization of human brain activity based on changes in
blood flow on a time scale of seconds. Multichannel electric recordings
from the scalp provide higher temporal resolution, but are not
easily identified with sources of activity in the brain. ICA has
been used to separate event-related brain responses into spatially
stationary and temporally independent subcomponents. ICA has been
applied to a variety of event-related potential recordings, including
an auditory detection task, an attentional task, and a memory
task. Previously identified response components were decomposed
into subcomponents and many entirely new components were found
that were differentially modulated by spatial and featural attention,
novelty, and learning context. This new technique promises to
complement the spatial resolution offered by functional
imaging techniques with high resolution timing information.
Laughlin, S. B., and Sejnowski, T. J., (2003). Communication in neuronal networks, Science 301: 1870-1874.
Fellous, J.-M., Tiesinga, P. H.E., Thomas, P. J., and Sejnowski, T. J., (2004). Discovering spike patterns in neuronal responses, Journal of Neuroscience 24(12): 2989-3001.
Eagleman, D. M., Jacobson, J. E., and Sejnowski, T. J., (2004). Perceived luminance depends on temporal context, Nature 428: 854-856.
Coggan, J. S., Bartol, T. M., Esquenazi, E., Stiles, J. R., Lamont, S., Martone, M. E., Berg, D. K., Ellisman, M. H., and Sejnowski, T. J., (2005). Evidence for ectopic neurotransmission at a neuronal synapse, Science 39: 446-451.
Frohlich, F., Bazhenov, M., Timofeev, I., Steriade, M., and Sejnowski, T. J., (2006). Slow state transitions of sustained neural oscillations by activity-dependent modulation of intrinsic excitability, Journal of Neuroscience 26: 6153 6162.
Dr. Sejnowski received his Ph.D. in physics from Princeton University. He was a postdoctoral fellow at Princeton University and the Harvard Medical School. He served on the faculty of Johns Hopkins University and was a Wiersma Visiting Professor of Neurobiology and a Sherman Fairchild Distinguished Scholar at Caltech
