OCNC 2013 Program

Gregory Stephens

An introduction to dynamical systems: from neural activity to natural behavior

My lecture will consist of two parts: an introduction to dynamical systems focused in particular on the power of qualitative analysis and a novel, quantitative approach to understanding the motion of C. elegans.

Indeed, while there has been an explosion in our ability to characterize the dynamics of molecules, cells, and circuits, our understanding of behavior on the organism-scale is remarkably less advanced. Here, we use high-resolution video microscopy to show that the space of shapes is low-dimensional, with just four dimensions accounting for 95% of the shape variance.   Projections of worm shape along these four “eigenworms” provide a precise yet substantially complete description of locomotory behavior, capturing both classical motion such as forward crawling, reversals, and Ω-turns and novel behaviors such as “pause” states at particular postures.  We use the eigenworms to construct a stochastic model of the body wave dynamics that predicts transitions between attractors corresponding to abrupt reversals in crawling direction and we show that the noise amplitude decreases systematically with increasing time away from food, resulting in longer bouts of forward crawling and suggesting that worms use noise to adaptive benefit. 

Suggested reading: TBA

Sophie Deneve

Representation of sensory signals and uncertainty in spiking neural networks.

Understanding how human perception is translated into actions and how our experience forms our worldview has been one of the central questions of psychology, cognitive science and neuroscience. In particular it is very hard to understand how our perceptions and strategies persist and/or change in the face of our continuous experience as active agents in an unpredictable and perpetually changing world. Indeed, the tasks humans are facing in “natural” situations (as opposite to simplistic laboratory settings) require the combination of multiple noisy and ambiguous sensory cues, as well as the use of prior knowledge, either innate or acquired from previous experiences. Such incomplete and imperfect sensory cues and priors can only provide probabilistic information (such as object structures that are more likely, or the probability of moving in a particular direction given the observed optic flow).

How the neural substrates perform these probabilistic inference tasks is still an open question. In this lecture, we will explore the hypothesis that rather that representing a single estimate of sensory or motor variables, single neurons or neural populations could be representing, implicitly or explicitly, probabilities or probability distributions. Many behaviorally relevant variables, such as movement direction, are represented by large groups of neurons with wide, overlapping tuning curves, and very noisy responses. Interestingly, this kind of code is ideally suited to perform Bayesian inference. In particular, when several population codes represent statistically related variables, for example the direction of motion of an object on the skin and on the retina, their information can be combined simply by summing inputs from the different populations. 

We will consider how probabilistic inference could be implemented by network of integrate and fire neurons. Such models will have important implications for our views of the neural code and the dynamics of cortical networks. We will show in particular that the high variability of sensory and motor spike trains could in fact not be noise, but a translation, in “neural language”, of sensory and motor uncertainty. Meanwhile, population coding is a natural and necessary consequence of interpreting noisy and ambiguous sensory inputs, and not a redundant code to deal with neural noise. In fact, neural representations at the level of the entire population could be order of magnitudes more precise that they appear to be when extrapolation from single neuron’s recordings. This would also explain the very tight balance between excitation and inhibition observed in cortical neurons, as well as the Poisson statistics of their spike trains.

Suggested reading : TBA

William (Bill) Holmes

Modeling neurons: from theory to practice. Experiences with hippocampal and vestibular neuron models

Dendritic modeling begins with the cable equation and the models of Rall. I will review some of the key solutions to the cable equation and discuss their importance for gaining an intuitive understanding of dendritic function. In particular, there are some simple formulas that allow insight into certain cell characteristics. As useful as these insights are for providing understanding, the assumption of passive membrane is rarely satisfied and this means that single neuron models with voltage- and calcium-dependent conductances in the dendrites and soma, as well as in the axon, are required. I will share some practical ideas for modeling neurons based on my experiences with hippocampal and vestibular neuron models. I will discuss things that “bug” me about single neuron modeling, including the plethora of ion channel types, the use and abuse of Boltzmann fits for modeling conductances, experimental issues with channel kinetics data, unnecessarily complicated time constant expressions and the fact that models that appear to reproduce observed behavior may not always be physiologically realistic.

Suggested reading: TBA

Avrama Blackwell

Molecular Mechanisms of Synaptic Plasticity

Long term synaptic plasticity is long lasting change in the strength of a synaptic connection, and is a proposed mechanism of memory storage. There are many induction paradigms used to induce plasticity, and they vary by brain region, type of plasticity, and molecular mechanisms. Calcium is a crucially important molecule for all types of synaptic plasticity: an elevation in intracellular calcium concentration is critical for induction of both potentiation and depression. Calcium activates many other molecules with a documented role in synaptic plasticity, including calcium calmodulin dependent protein kinase II (CaMKII), protein kinase C, and cAMP dependent protein kinase (PKA). I will describe a few characteristics of hippocampal synaptic plasticity and a few of the signaling pathways implicated, as well as various computational models that have been developed to evaluate whether the identified signaling pathway can explain the sensitivity to stimulation pattern.

Suggested reading: TBA

Michael E. Hasselmo

Memory mechanisms in the entorhinal cortex and hippocampus: Oscillations, grid cells and acetylcholine.

Entorhinal cortex, the hippocampus and the medial septum play an important role in memory for the location and time of events in episodic memory. The physiological properties of these regions help us to understand the mechanisms of episodic memory function. The spatial location of memories may be coded by place cells in the hippocampus that respond when a rat is in a single location (O’Keefe and Burgess, 2005) and grid cells in entorhinal cortex that fire when a rat visits an array of locations in the environment that fall in a hexagonal pattern (Moser and Moser, 2008). Hebbian modification of synapses could form associations between locations and items or events in memory. The effects of acetylcholine could enhance the encoding of new associations between locations and items. I will review the data on the behavioral effects of acetylcholine receptor blockade and models that link the behavioral role to specific cellular effects of acetylcholine (Hasselmo, 2006; Hasselmo and Stern, 2006).

The coding of space could depend on the spacing and size of firing fields that becomes progressively larger for grid cells recorded in more ventral anatomical locations. Models of grid cells show how grid cells could arise from oscillations and resonance in entorhinal cortex (Burgess, Barry and O’Keefe, 2007; Hasselmo, 2008). Loss of oscillations with inactivation of the medial septum is associated with a loss of spatial periodicity of grid cells (Brandon et al., 2011). The spacing of grid cells could depend upon the intrinsic frequency of oscillations and resonance. Whole cell patch data from shows that layer II stellate cells in entorhinal cortex have higher frequencies of resonance and subthreshold membrane potential oscillations in dorsal compared to ventral entorhinal cortex (Giocomo et al., 2007). A new model shows how this resonance could contribute to the spacing of grid cell firing fields.

Brandon, M.P., Bogaard, A.R., Libby, C.P., Connerney, M.A., Gupta, K., Hasselmo, M.E. (2011) Reduction of theta rhythm dissociates grid cell spatial periodicity from directional tuning. Science, 332: 595-599.

Burgess N, Barry C, O'Keefe J. 2007. An oscillatory interference model of grid cell firing. Hippocampus 17(9):801-12.

Giocomo LM, Zilli EA, Fransen E, Hasselmo ME. (2007) Temporal frequency of subthreshold oscillations scales with entorhinal grid cell field spacing. Science, 315:1719-22.

Hasselmo, M.E. (2006) The role of acetylcholine in learning and memory. Curr. Opinion Neurobiol. 16(6): 710-715.

Hasselmo M.E. (2008) Grid cell mechanisms and function: Contributions of entorhinal persistent spiking and phase resetting. Hippocampus. 2008;18(12):1213-29.

Hasselmo, M.E., Stern, C.E. (2006) Mechanisms underlying working memory for novel information. Trends in Cognitive Sciences, 10(11):487-93.

Moser EI, Moser MB (2008) A metric for space. Hippocampus 18: 1142-1156.

O’Keefe J., Burgess N. (2005) Dual phase and rate coding in hippocampal place cells: theoretical significance and relationship to entorhinal grid cells. Hippocampus 15: 853-866.

Suggested reading: TBA

Mitsuo Kawato

From computational model based neuroscience to manipulative neuroscience

My lecture will first summarize computational model based neuroimaging and neurophysilogy studies from our group related to cerebellar internal model and basal ganglia reinforcement learning model. After pointing out pros and cons of this approach, I advocate necessity of causal tools in systems neuroscience. Decoded neurofeedback and decoded connectivity neurofeedback are new experimental tools to realize spatiotemporal neural activities in human brain. Experimental and theoretical issues will be presented.

Suggested reading: TBA

Gaute Einevoll

Modelling and analysis of the local field potential (LFP)

The past decade has witnessed a renewed interest in local field potentials (LFPs), i.e., extracellularly recorded potentials with frequencies up to 500 Hz. This is due to both the advent of multielectrodes allowing for recording of LFPs at tens to hundreds of sites simultaneously and the insight that LFPs offer a unique window into how neurons in cortical populations integrate synaptic inputs. However, owing to its numerous potential neural sources, the LFP is difficult to interprete. Careful mathematical modelling and analysis is thus needed to take full advantage of the opportunities that this signal offers in understanding the signal processing in cortical circuits and, ultimately, the neural basis of perception and cognition. In the lecture I will go through the biophysical origin of LFP, how it can be mathematically modeled, and new methods for analysing the signal.

Suggested reading: TBA

Jonathan Pillow

Neural coding and statistical models for neural spike trains

A central problem in systems neuroscience is to understand the probabilistic relationship between environmental stimuli and neural spike responses. A powerful approach to this problem is to develop explicit statistical models of neural spike trains. Such models allow us to determine how stimulus information is encoded in neural activity and how it might be read out by downstream brain areas.

In this talk, I will provide a general introduction to neural encoding models and review classical methods like reverse-correlation (spike-triggered averaging), Volterra/Weiner kernels, and maximum likelihood estimation. Additional topics will include spike-triggered covariance (STC) analysis and generalized linear models (GLMs) for multi-neuron data. I will discuss several open problems and invite discussion about the use of probabilistic models for understanding the neural code.

Suggested reading: TBA