William H. Nesse

Research Interests

Biophysical information representation for signal detection

Sensory neurons represent information about the physical world around us in a temporal pattern of spikes. The space of all spike sequence patterns {t_i} and associated probabilities of occurrence given a time-varying stimulus input can be thought of as a code. Each sensory neuron has access to a tiny fraction of our sensory world, but this code is information rich. For large changes in stimulus intensity, the change in {t_i} can be easily apparent, in which case, only a crude method of coding is necessary, represented by the average spike rate. However, in many instances crucial for survival, it is important to detect very small differences in stimulus intensity. Certainly, evolutionary pressures have evolved brains to extract as much information as possible from spike trains. Particularly, sensory afferents—the first stage of sensory processing—must represent as much information as possible. Any information lost at this stage cannot be recaptured at later stages. Information can be lost because the sensory apparatus was not built to detect very small or very large changes in input, or there is intrinsic noise that pollutes the sensory channel. In signal processing lingo, this polution is often termed “garbage in, garbage out.”

The spike code of sensory afferents {t_i} is determined by the interaction of the stimulus time-course, but is also shaped by intrinsic physiological dynamics of the neuron. How these intrinsic dynamics can be used to extract stimulus information is a rich area of research. My research studies how these intrinsic neural dynamics are used to represent sensory information biophysically as a set of molecular states of the cell. My research focuses on how these molecular states, quite exquisitely, provide a information-theoretically efficient stimulus representation, and how these states are transmitted effectively to “downstream” neurons for further processing. Temporal correlations in {t_i} are a common feature of spike trains, particularly sensory afferents (see Farkhooi et al, 2009, Phys Rev E). In a recently published paper (Nesse et al, 2010, PNAS) I study how these temporal correlations are a result of these molecular activations. Using analytical and computational methods how neurons can employ these dynamical patterns in the spike train to create a statistically efficient mode of stimulus representation in the biophysical states of neurons.

Information Encoding and Short-Timescale Synaptic Plasticity

I have studied the role of ultra-fast synaptic depression of P-Unit afferents fibers for information encoding to the electrolateral line lobe (ELL) of the weakly electric fish. This synaptic depression is instantaneous, depending only on the previous interspike interval. The standard understanding of short term synaptic depression (STD) is that it depends on past mean activity over longer timescales than a single interspike interval. Hence, the novel form of STD that we have observed we call “fast” STD: FSTD (see Khanbabie and Nesse et al, 2010, J. Neurophysiol.)

Emergent Network Phenomena and Random Fluctuations

Neuronal networks are a prime example of a system consisting of many similar sub components that are coupled together to form a larger system. Emergent network phenomena are novel behaviors of the larger networked system that are not exhibited by the isolated sub systems but are a result of complex interactions between the components. Furthermore, random fluctuations or noise in neuron behavior is commonly observed. How noise influences and shapes emergent network dynamics is a central theme in two areas of my research.

(1)  In the brain stem of mammals there exists a cluster of cells, termed the pre Botzinger Complex (preBotC), responsible for the coordinated rhythmic activation of the diaphragm and rib muscles responsible for breathing. The cells in the preBotC exhibit rhythmic bursts of activity that command the inspiratory phase of the breath, and serves as the “pacemaker” for breathing. Cells in the PreBotC exhibit synchronized bursts of action potentials that together form a population-level oscillation with periods on the order from seconds to minutes. These rhythms persist in the absence of inhibition. A number of studies have focused on how intrinsic currents in a minority population of intrinsically rhythmic bursting, so called “pacemaker” cells, could mediate population rhythmicity. More recently, however, there is evidence that pacemaker bursting cells may not be necessary for the production of the population rhythm, and it has been hypothesized that the rhythm is an emergent network property mediated by recurrent excitation. In a recent paper, Myself and collaborators (Nesse, Borisyuk, and Bressloff, 2008) have studied an excitatory all-to-all coupled network of N spiking neurons with synaptically filtered background noise and slow activity-dependent hyperpolarization (AHP) currents, where individual isolated model cells do not exhibit bursting behavior. However, the network system exhibits noise-induced burst oscillations over a range of values of the noise strength (variance) and level of cell excitability. Since both of these quantities depend on the rate of background synaptic inputs, we have shown how noise can provide a mechanism for increasing the robustness of rhythmic bursting and the range of burst frequencies. By exploiting a separation of time scales we have also shown how the system dynamics can be reduced to low-dimensional mean field equations in the large-N limit. Analysis of the bifurcation structure of the mean field equations provides insights into the dynamical mechanisms for initiating and terminating the bursts.

cropped-nwk-nr.jpg

In collaboration with Christopher Del Negro (College of William and Mary), we have extended this mean field model to incorporate multiple time-scale AHP currents in order to provide a possible explanation for the experimentally observed peak in oscillation irregularity (maximal incoherence) for intermediate levels of excitability. Many neurons exhibit multiple activity-dependent adaptation currents that span many different time scales, from fast spike repolarization (5 msec), to potassium after hyperpolarization currents (200 msec), to modulatory channel phosphorylization (seconds), to channel insertion and deletion (minutes or longer). We find that if two timescales of adaptation are present with sufficiently large separation, there exists a peak in the oscillation irregularity for intermediate excitability (set by input current) with more regularity for high and low excitability. Such a mechanism is distinct from standard forms of coherence resonance and is a result a balanced state between the multiple frequency-limiting time-scale components of oscillation mechanism (see Nesse, Del Negro, and Bressloff, 2008).

(2) The above research centered on emergent properties resulting from very large networks where population averaging played a central role in the analysis. Noise can play a distinct and important role in small networks as well, found in invertebrates. In collaboration with Dr. Gregory A. Clark (U of Utah Bioengineering) we have investigated the model system of the Hermissenda eye, a marine nudibranch found in sub tidal estuaries seen below.

01064

The eye contains five photoreceptors which spike repetitively in response to light. We find that intrinsic noise in this biological network serves an important role of destroying emergent phase locking spike patterns between the oscillators, that “lock” one rhythmic neuron to another. Such pathological patterns can destroy the efficient encoding of light stimuli in the eye. Using experimental and theoretical techniques, we have discovered that the stochastic oscillatory photoreceptors, when coupled together, transform their firing statistics from regular Gaussian-like statistics that contain memory of past cycles, to memoryless Poisson-like firing that serves both to encode fast transient stimuli, and to destroy pathological phase locking (see Nesse and Clark, 2010,  Biol. Cybernetics).

Advertisements

Leave a Comment »

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at WordPress.com.

%d bloggers like this: