CNS*2013 Workshop on
Methods of System Identification for Studying Information Processing in Sensory Systems
Wednesday, July 17, 2013
A functional characterization of an unknown system typically begins by making observations about the response of that system to input signals. The knowledge obtained from such observations can then be used to derive a quantitative model of the system in a process called system identification. The goal of system identification is to use a given input/output data set to derive a function that maps an arbitrary system input into an appropriate output.
In neurobiology, system identification has been applied to a variety of sensory systems, ranging from insects to vertebrates. Depending on the level of abstraction, the identified neural models vary from detailed mechanistic models to purely phenomenological models.
The workshop will provide a state of the art forum for discussing methods of system identification applied to the visual, auditory, olfactory and somatosensory systems in insects and vertebrates.
The lack of a deeper understanding of how sensory systems encode stimulus information has hindered the progress in understanding sensory signal processing in higher brain centers. Evaluations of various systems identification methods and a comparative analysis across insects and vertebrates may reveal common neural encoding principles and future research directions.
The workshop is targeted towards systems, computational and theoretical neuroscientists with interest in the representation and processing of stimuli in sensory systems in insects and vertebrates.
- Vasilis Z. Marmarelis (2004). Nonlinear Dynamic Modeling of Physiological Systems. Wiley-IEEE Press, Hoboken, NJ, 2004.
- Wu, M., David, S., & Gallant, J. (2006). Complete Functional Characterization of Sensory Neurons by System Identification. Annual Review of Neuroscience, 29, 477–505.
- Ljung, L. (2010). Perspectives on System Identification , Annual Reviews in Control, 34 (2010), 1-12.
- Aurel A. Lazar, Department of Electrical Engineering, Columbia University.
- Mikko I. Juusola , Department of Biomedical Science, University of Sheffield.
Wednesday, 9:00 AM - 5:50 PM, July 17, 2013
Morning Session I (9:00 AM - 10:20 AM)
Chair: Mikko I. Juusola
9:00 AM - 9:40 AM
Neural Circuits for Fly Visual Course Control
Alex Borst , Max Planck Institute for Neurobiology, Martinsried.
Visual navigation has been studied extensively in flies, both in tethered as well as in freely flying animals. As neural control elements, the tangential cells of the lobula plate seem to play a key role: they are sensitive to visual motion, have large receptive fields, and, with their spatial distribution of preferred directions, match the optic flow as elicited during certain types of flight maneuvers. However, several key questions have remained unanswered for long: 1. What is the neural circuit presynaptic to the tangential cells responsible for extracting the local direction of motion? 2. Do the lobula plate tangential cells indeed control turning responses of the fly? 3. Is there a separate visual course control system allowing the fly to detect and track individual objects?
By combining whole-cell patch recording and behavioral studies with silencing , optogenetic stimulation and optical recording from genetically targeted candidate neurons in Drosophila, the following progress has been made towards answering these questions: 1. Using apparent motion stimuli on flies with lamina neurons L1 or L2 silenced, we find that L1 and L2 feed into separate motion pathways specialized for the detection of ON (L1) and OFF (L2) signals. Optical recording from T4 and T5 cells reveals that these cells carry the directionally selective output of ON (T4) and OFF (T5) motion detectors. 2. Optogenetic stimulation of lobula plate HS-cells via a switchable variant of channelrhodopsin2 elicits head and body turns of Drosophila in tethered flight, thus demonstrating that HS-cells indeed are involved in flight control. 3. Behavioral studies of flies with T4/T5 cells blocked show that these flies, although being completely blind to motion of gratings and small objects, are still able to fixate objects under closed-loop conditions. Fixation behavior is based on a system, implemented in parallel to the motion pathway, which uses local temporal brightness changes to detect the position of an object.
9:40 AM - 10:20 AM
Sensorimotor Circuit Dysfunction as the Origin of a Motor Neuron Disease
Brian D. McCabe, Department of Pathology and Cell Biology, Columbia University.
Spinal Muscular Atrophy (SMA) is a neurodegenerative motor system disease caused by mutations in Survival Motor Neuron (SMN). It is the most common inherited cause of childhood mortality and currently there is no treatment. To study SMA we employed Drosophila SMN mutants which have defective muscle growth, locomotion and motor neuron function similar to human disease phenotypes. Surprisingly, we found that none of these defects were corrected by transgenic restoration of SMN in either muscles or motor neurons. Instead, we discovered that SMN must be restored in both proprioceptive sensory and central interneurons in the motor circuit in order to rescue disease. We established that SMN-deficient motor circuit neurons had defective neurotransmission and that this was sufficient to non-autonomously perturb motor neuron function. Furthermore, we found that genetic inhibition of K+ channels or administration of an FDA approved small molecule K+ channel inhibitor, 4AP, restored motor circuit activity and provided benefit in the Drosophila SMA model. Based on this data, 4AP is now in human clinical trials for SMA.
To establish the molecular mechanisms that lead to selective motor circuit dysfunction in SMA, we investigated SMN-dependent gene expression. From a genome-wide functional analysis of Drosophila genes regulated by SMN, we identified a novel evolutionarily conserved protein, Stasimon, required for sensory-motor circuit function in both Drosophila and vertebrates. Our results illuminated a cohesive chain of molecular events linking SMN-depletion to motor circuit dysfunction, establishing a mechanistic framework to understand the selective vulnerability of neurons in SMA and demonstrating that neurodegenerative disease can be induced by the dysfunction of neuronal circuits.
Morning Break 10:20 AM - 10:50 AM
Morning Session II (10:50 AM - 12:10 PM)
Chair: Vivek Jarayaman
10:50 AM - 11:30 AM
System Identification in Olfactory Receptor Neurons
Jean-Pierre Rospars, INRA Versailles.
Olfactory receptor neurons (ORNs) have been the subject of intensive studies both in vivo, in vitro and in silico. Using molecular genetics, electrophysiology, calcium imaging, behavioural and modelling methods, the various biochemical and electrical processes involved in transduction have been investigated, mainly in insects and vertebrates. Although much progress has been made on these mechanisms, many uncertainties remain.
We present three examples for which modelling and system identification have provided useful insight on olfactory reception and transduction processes. These examples give an overview of the present state of knowledge, describe processes at various levels (molecular, cellular, multicellular), illustrate diverse methods that can be utilized, and their limitations.
The first example focuses on odorant-receptor interactions. It shows how the spiking responses of rat ORNs to odorant mixtures can throw light on internal processes within receptor proteins.
The second example examines post-receptor processes, especially in the pheromone-sensitive ORNs of male moths that can detect female-released sexual pheromone with exquisite sensitivity. We gave special attention to the generation of the receptor potential, which is the first integrated response of the ORN and the most precisely measured. Using many known facts, both qualitative and quantitative, in this and other ORNs, we developed an integrated biochemical and electrical model of the whole reaction network, including the various activating, inactivating and regulating feedback mechanisms that are needed to return the whole system to its resting state. It accounts quantitatively for the kinetics and dynamic range of the receptor potential, describes the kinetics of the main components (receptors, enzymes, channels), suggests a specific role for the various channels at different stimulus intensities, and gives insight on calcium regulation for example.
The third example describes the nonsynaptic (ephaptic) interaction between insect ORNs housed in the same sensillum. A simple electrical model of the sensillum predicts that the transient activation of one ORN can inhibit the response of a neighbouring ORN. This prediction, which has been experimentally confirmed recently, shows that integration of sensory information may not be restricted to second or higher order neural layers.
11:30 AM - 12:10 PM
Functional Identification of Neural Circuits with Spiking Inputs
Aurel A. Lazar, Department of Electrical Engineering, Columbia University.
We present a new approach for a complete functional identification of biophysical spike-processing neuronal circuits. The circuits considered accept multidimensional spike trains as their input and are comprised of a multitude of temporal receptive fields and of conductance-based models of action potential generation. We investigate circuits of various architectures, including circuits with lateral connectivity and feedback and present algorithms for identifying circuit parameters directly from spike times produced by neurons. The algorithms obviate the need to repeat experiments in order to compute the neurons’ rate of response, rendering our methodology of interest to both experimental and theoretical neuroscientists.
References:  Aurel A. Lazar and Yevgeniy B. Slutskiy, Functional Identification of Neural Circuits with Spiking Inputs, Neural Computation, in press (2013).  Bionet Group, Channel Identification Machines Toolbox.
Lunch 12:10 PM - 2:00 PM
Afternoon Session I (2:00 PM - 3:20 PM)
Chair: Aurel A. Lazar
2:00 PM - 2:40 PM
Motor-Modulated Visual Maps in the Fly Brain
Vivek Jayaraman, Janelia Farm, Ashburn, VA.
Many animals, including insects, use visual landmarks for orientation and navigation. In Drosophila melanogaster, behavioral genetics studies have identified the central complex as being required for innate attraction to particular visual features, and for short- and long-term memory for visual patterns. Studies in several insects suggest that the region is also important for motor coordination. Here we present an analysis of physiological recordings from this region in Drosophila. We focused on neurons implicated in orientation and place memory in the fly: ring neurons of the ellipsoid body, a sub-region of the central complex.
We show that each ring neuron sends dendrites to a single microglomerulus in the lateral triangle (LTr), a multi-glomerular brain region that is a major source of input to the ellipsoid body. We studied the responses of complete populations of ring neuron classes using two-photon calcium imaging in head-fixed flies that were flying or walking on an air-supported ball in an LED arena. LTr microglomeruli respond strongly to visual input, showing retinotopically organized receptive fields (RF). RFs are stereotyped enough to allow microglomeruli to be functionally identifiable across flies. LTr responses to visual stimuli are diminished during flight, but are not significantly modulated during walking. A simple linear model based on a population code of LTr responses, recorded during closed-loop virtual reality flight behavior, is sufficient to compute the fly’s heading relative to visual features in its surroundings. We suggest that ring neurons may provide the visual pattern information necessary for a variety of orienting and navigation behaviors in the fly. Our results provide the first evidence for retinotopic maps in higher brain structures in Drosophila, and set the stage for mechanistic studies of sensorimotor integration underlying visually-guided decision-making in this genetic model organism.
2:40 PM - 3:20 PM
Dynamics of Thalamocortical Circuits Underlying Sensorimotor Integration
Karim G. Oweiss, Department of Electrical and Computer Engineering and Department of Neuroscience, Michigan State University.
The highly sophisticated interaction between the somatosensory and motor systems underlies our ability to seamlessly control our movements. Loss of this interaction results in a devastating quality of life. We have proposed to restore somatosensory feedback in motor impaired subjects by stimulating the somatosensory pathway to deliver information about limb state. The thalamus plays an indispensible role in relaying information from the periphery to the cerebral cortex through multiple transformations of afferent inputs. To deliver somatosensory feedback that can be naturally perceived by the subject, it is necessary to first use system identification techniques to quantify the extent to which artificial stimulation can successfully recruit existing thalamocortical circuits. It is also necessary to do so during passive as well as active sensing, where in the latter case motor cortex influence on somatosensory cortex becomes significant.
I will present a general framework for using system identification techniques to characterize the properties of the thalamocortical system during sensorimotor integration. I will present data demonstrating how thalamic neurons influence the dynamics of primary somatosensory cortex (S1) neurons in the rat during whisker deflection. I will further demonstrate how S1 response dynamics can be manipulated by thalamic optogenetic stimulation in the absence and presence of congruent primary motor cortex (M1) stimulation. Results suggest that S1 responses to thalamic stimulation have fast dynamics and are frequency-dependent. In contrast, S1 responses to M1 stimulation exhibit much slower dynamics. These results suggest that M1 circuits may be programmed to modulate S1 circuits over long time scales to facilitate sensory expectation of evolving motor commands. Thalamic input, on the other hand, may work to engage or disengage S1 circuits to sudden perturbations from the outside world that may interfere with movement goals during task execution. These findings suggest the need to more systematically characterize how information is integrated between the two areas, and pave the way to improve the design of bi-directional sensorimotor prosthesis where microstimulation of the afferent pathway is optimized based on the underlying system properties.
Afternoon Break 3:20 PM - 3:50 PM
Afternoon Session II (3:50 PM - 5:50 PM)
Chair: Brian D. McCabe
3:50 PM - 4:30 PM
Why White-Noise Stimulation Fails to Test Information Capacity of Neurons?
Mikko I. Juusola, Department of Biomedical Science, University of Sheffield.
White-noise analysis is used extensively in neurosciences. White-noise stimulation modulates around a mean, has a flat spectrum, independent values at each moment and Gaussian distribution. These properties are thought beneficial for quantifying response dynamics and signaling performance of neurons. Because Gaussian white-noise (GWN) utilizes input frequencies evenly, it is often considered a prerequisite for estimating the information capacity of neural output, based on Shannon’s ideas of maximum transmission in communication channels. It then follows that the higher the signal-to-noise ratio and the broader the response bandwidth to GNW stimulation, the higher the neuron’s information capacity.
However, sensory neurons, such as photo-, olfactory- and mechanoreceptors, adapt continuously, generating noisy responses of more complex amplitude and phase correlations to stimulus changes in natural environment. These neurons have transduction reactions compartmentalized in membrane elaborations to work as sampling units, such as cilia or microvilli, and use adaptive stochastic sampling to encode discrete information. Their structures and coding strategies are thought to be matched to their natural inputs to represent the sensory world efficiently. In contrast, GWN stimulation, which misses these correlations, is rather unnatural to neurons and, thus, its encoding likely less efficient.
Here, we analyze this problem by using intracellular recordings and stochastic model simulations. We show that, while GWN adapts a fly photoreceptor, much stimulus energy dissipates on frequencies too fast to see and many of its microvilli (sampling units) become refractory. Lowering GWN bandwidth improves microvilli utilization, integrating larger responses. But because GWN lacks the correlations of natural scenes, which drive microvilli more efficiently, responses to GWN contain smaller sample rate changes than responses to natural stimuli and, consequently, less information. We further show how refractory microvilli sacrifice sensitivity to enhance contrast resolution, and how adding microvilli or quickening their refractoriness improves encoding of GWN; yet, this performance is always submaximal. Hence, GWN stimulation underutilizes information capacity of neurons, in which structures and functions promote encoding of invariable features in natural environment.
4:30 PM - 5:10 PM
A Systems Approach to Sensorimotor Control Design in Blowflies
Holger G. Krapp, Department of Bioengineering, Imperial College.
Flies have been successfully used as animal models to study fundamental principles of optomotor control over many decades. The result is a comprehensive data body that enables us to take on new challenges in terms of understanding what the neuronal mechanisms are that define the design of multisensory behavioural control from a well-informed starting point. In my talk I will present a systems approach applied to study stabilization reflexes in blowflies. It combines electrophysiological, behavioural, anatomical and computational methods aiming to elucidate some general principles of biological control systems and the constraints under which they operate. I will also argue that – to understand sensorimotor design in general – we not only require knowledge of the functional properties of the sensor systems involved but also on the dynamic properties of the motor systems they control.
5:10 PM - 5:50 PM
Spike Synchrony and Spike-LFP Locking in Monkey Primary Visual Cortex during Free Viewing of Natural Scenes
Sonja Gruen, Institute of Neuroscience and Medicine & Institute for Advanced Simulation, Juelich Research Centre and JARA.
During natural vision, primates perform frequent saccadic eye movements, allowing only a narrow time window for processing the visual information at each location. Individual neurons may contribute only with a few spikes to the visual processing during each fixation, suggesting precise spike timing as a relevant mechanism for information processing. We found in V1 of monkeys freely viewing natural images, that fixation-related spike synchronization occurs at the early phase of the rate response after fixation-onset (Maldonado et al, 2008), suggesting a specific role of the first response spikes in V1. In Ito et al (2011) we showed that there are strong local field potential (LFP) modulations locked to the onset of saccades, which continue into the successive fixation periods. Visually induced spikes, in particular the first spikes after the onset of a fixation, are locked to a specific epoch of the LFP modulation. We suggested that the modulation of neural excitability, which is reflected by the saccade-related LFP changes, serves as a corollary signal enabling precise timing of spikes in V1 and thereby providing a mechanism for spike synchronization.