Columbia Workshop on Brain Circuits, Memory and Computation
Thursday and Friday, March 21-22, 2019
Davis Auditorium, CEPSR
Columbia University, New York, NY 10027
The goal of the workshop is to bring together researchers interested in developing executable models of neural computation/processing of the brain of model organisms. Of interest are models of computation that consist of elementary units of processing using brain circuits and memory elements. Elementary units of computation/processing include population encoding/decoding circuits with biophysically-grounded neuron models, non-linear dendritic processors for motion detection/direction selectivity, spike processing and pattern recognition neural circuits, movement control and decision-making circuits, etc. Memory units include models of spatio-temporal memory circuits, circuit models for memory access and storage, etc. A major aim of the workshop is to explore the integration of various sensory and control circuits in higher brain centers.
A Fruit Fly Brain Hackathon is being conducted in conjunction with the workshop. Workshop participants are welcome to attend the hackathon.
Organizer and Program Chair
Aurel A. Lazar, Department of Electrical Engineering, Columbia University.
Registration is free but all participants have to register. Thank you!
Lodging and Directions to Venue
Please follow this link for lodging details and directions to the hotel and venue.
The 2019 Columbia Workshop on Brain Circuits, Memory and Computation is supported by the
Department of Electrical Engineering, Columbia University
Center for Computing Systems for Data-Driven Science, Data Science Institute, Columbia University
School of Engineering and Applied Science, Columbia University
Anton Arkhipov, Allen Institute of Brain Science.
Data-Driven Modeling of the Cortex Based on a Systematic Experimental Platform
The Mindscope project at the Allen Institute for Brain Science aims to elucidate mechanisms underlying cortical function in the mouse, focusing on the visual system. This involves concerted efforts of multiple teams characterizing cell types, connectivity, and neuronal activity in behaving animals. An integral part of these efforts is the construction of models of the cortical tissue and cortical computations. We will discuss our current progress on this front. First, we built a computational pipeline to produce models of individual neurons based on slice electrophysiology and morphology reconstructions. Second, we used these models as building blocks for large simulations of cortical activity in response to visual stimuli. A highly realistic 230,000-neuron model of the mouse cortical area V1, receiving thalamocortical visual inputs, has been constructed at two levels of resolution: one using the biophysically detailed, compartmental neuron models and the other using point-neuron models. Third, we perform systematic comparisons of simulated responses to in vivo experiments and investigate the structure-function relationships in the models to make mechanistic predictions for experimental testing. To enable this work, we developed the software suite called Brain Modeling ToolKit (BMTK) and a modeling file format called SONATA. These tools, the models, and simulation results are all being made freely available to the community via the Allen Institute Modeling Portal.
Richard Benton, Center for Integrative Genomics, University of Lausanne.
Olfactory Evolution in Drosophilids: Receptors, Neurons and Behaviours
My group is interested in understanding the structure, function and evolution of neural circuits. We exploit the olfactory system of Drosophila melanogaster as a model, which is well-described, experimentally accessible and dynamically evolving. Furthermore, genomic and growing genetic access to closely related, but ecologically diverse, drosophilids provides an unparalleled foundation for comparative studies of their olfactory circuits. I will present our recent work on the evolution of the olfactory pathways of Drosophila sechellia, an island endemic that displays extreme specialisation for the Morinda citrifolia “noni” fruit as a food source and breeding site. We have started to define the molecular basis by which D. sechellia’s olfactory receptors are re-tuned to the odours of its host and – through development of novel neurogenetic tools in this species – how these different sensory pathways contribute to host-seeking behaviours.
Benjamin L. de Bivort, Department of Organismic and Evolutionary Biology, Harvard University.
The Neural Circuit Basis of Behavioral Individuality
Individuals animals vary in their behaviors even when their genetics and environment are held constant. The mechanisms underlying this variation is still largely uncharacterized, though we have made some progress in understanding genetic and circuit variants that lead a population of animals to exhibit high or low variability in behavior. These are key insights, but fall short of predicting the specific behavioral biases of individual animals. We term the causal biological features that determine individual behavioral biases "loci of individuality," and we have begun to search for them in the circuits that mediate sensory-evoked and spontaneous behaviors. We have found neural circuit elements, whose morphological properties predict behavioral biases. Specifically, the volume of axonal output arbors of central complex neurons that project to the Lateral Accessory Lobe correlates with changes in locomotor behavior in specific sensory contexts. We hypothesize that individual wiring variation in these neurons has a large effect on behavior because they lie at a bottleneck in the sensorimotor circuit, where stochastic fluctuations have an outsized effect on circuit outputs. Thus, we have found that individual variation in the structure of small numbers of neurons, in topologically critical circuit positions, predict individual behavioral biases in a sensory-context specific fashion.
Kristin Branson, HHMI Janelia Research Campus, Ashburn, VA.
Gwyneth Card, HHMI Janelia Research Campus, Ashburn, VA.
Towards a Brain Architecture for Visual Behavior Selection
Selecting the right behavior at the right time is critical for animal survival. Animals rely on their senses to deliver information about the environment to sensory processing areas in the brain that extract relevant features and form the perceptual representations that guide behavior. We aim to uncover the organization of this feature space and the neural mechanisms by which these cues are translated into dynamic motor activity.
Our current focus is visually-driven behaviors of the fly. In particular, those driven by visual looming cues produced by an approaching predator or an imminent collision. The same looming stimulus can evoke a wide range of different behaviors, including a rapid escape jump, a slower, more stable takeoff sequence, or a landing response. We use whole-cell patch clamp physiology in behaving flies, calcium imaging, high-throughput/high-resolution behavioral assays, and genetic tools to examine the transformation of information from sensory to motor. I will discuss our recent work investigating the representation of ethologically-relevant visual features in the fly optic glomeruli and the mechanisms by which descending neurons read out this feature information to produce an appropriate behavioral choice.
Kevin M. Franks, Department of Neurobiology, Duke University.
Recurrent Circuitry Stabilizes Cortical Odor Representations Despite Degraded Sensory Inputs
Animals must recognize familiar objects even when incoming sensory input may be noisy, degraded or incomplete. Theoretical studies have shown that this process, often called “pattern completion”, can be implemented by recurrent cortical circuits within an autoassociative network. However, direct experimental evidence for this process has been lacking. Here, we recorded odor-evoked activity from simultaneous populations of neurons in mouse olfactory bulb and olfactory cortex before and after inducing anesthesia. We found that cortical odor representations remained stable across brain states even though upstream representations, in olfactory bulb, were markedly degraded under anesthesia. Odor representations were more robust in cortical pyramidal cells, which receive recurrent connections, than in semilunar cells, which do not. Furthermore, cortical odor representations became as state-dependent as those in olfactory bulb after blocking recurrent connections between pyramidal cells. Thus, we provide direct evidence for the crucial role of recurrent cortical circuitry in stabilizing sensory representations driven by degraded sensory input.
Paul A. Garrity, Department of Biology, Brandeis University.
Thermosensing in the Fly: from Genes to Cells to Behavior
Temperature is a universal physical variable that affects all aspects of physiology. Among other vital functions, animals rely on their thermosensory systems to maintain appropriate body temperatures and avoid thermal extremes. Thermosensing in Drosophila melanogaster depends on multiple classes of thermosensors, which rely on diverse classes of molecular thermoreceptors, exhibit distinct thermal sensitivities and have distinct behavioral functions. How these thermosensory neurons encode temperature information and how the information they provide supports behavior remain largely open questions. I will present recent work in which we have identified a critical role for phasic thermosensors (sensory neurons which primarily respond to heating and cooling rather than hot or cold temperatures) in innocuous thermosensation in Drosophila. At the molecular level, we have identified the molecular receptors that confer thermosensitivity upon these neurons and found that these molecules are critically important for the morphogenesis of the elaborate sensory compartment involved in sensing temperature changes. At the behavioral level, we find that these phasic thermosensors are essential for mediating behavioral responses across a wide range of temperatures, consistent with their ability to respond robustly to small temperature changes at very different ambient temperatures. Together these findings begin to provide a consistent picture of thermosensory processing in the fly brain and raise questions concerning how phasic inputs might combine with other sensory inputs to drive behavior.
Stephen F. Goodwin, Department of Physiology, Anatomy and Genetics, University of Oxford.
Tim Jarsky, Allen Institute of Brain Science.
Karla Kaun, Department of Neuroscience, Brown University.
Circuits that Encode and Predict Alcohol Associated Preference
The ability to associate a rewarding stimulus with a sensory cue from the environment is critical for an animal’s ability to find food and mates. Mapping the circuits in which these associations are formed, then initiate an output response provides a scaffold for understanding how experiences can influence decisions. Drugs of abuse such as alcohol can induce lasting changes in these circuits to induce maladaptive behavioral decisions. Here we describe a how memories for the intoxicating properties of alcohol are acquired and expressed through different mushroom body circuits. Acquisition of odor-alcohol memories induce lasting molecular changes which affect plasticity within circuits important for memory expression. Expression of these memories requires a remarkably complex multi-level circuit whereby dopamine directly, and indirectly via the mushroom body, modulates the activity of glutamatergic and cholinergic output neurons. Moreover, trans-synaptic tracing the outputs of these neurons suggests convergent and divergent networks, providing an elaborate framework for integrating external content and internal state to coordinate an appropriate output response. Together this work provides a snapshot of how alcohol can affect the dynamic molecular and circuit mechanisms required for behavioral decisions.
Joint work with Kristin M. Scaplen1, Mustafa Talay2, Emily Petruccelli3, Nicolas Ledru4, Gilad Barnea1, 1Department of Neuroscience, Brown University, Providence, RI. 2Department of Molecular and Cell Biology, Harvard University, Cambridge, MA. 3Department of Biology, Southern Illinois University, Edwardsville, IL. 4MSTP Program, Washington University, St Louis, MO.
Funding support: NIAAA R01AA024434, NIGMS P20GM103645, and the Smith Family Award for Excellence in Biomedical Research.
Gero A. Miesenboeck, Centre for Neural Circuits and Behaviour, University of Oxford.
The Somnostat: Mechanisms for Balancing Sleep Need and Sleep
Sleep is vital and universal, but its biological function remains unknown. We seek to understand why we need to sleep by studying how the brain responds to sleep loss. Our studies in Drosophila have pinpointed neurons whose sleep-inducing activity switches on as sleep deficits accrue, revealed how this activity switch works, and furnished a molecular interpretation of sleep pressure, its accumulation, and its discharge.
Venkatesh N. Murthy, Department of Molecular and Cellular Biology, Harvard University.
Decoding and Demixing Smells
Fluctuating mixtures of odorants, often transported in fluid environments, are detected by an array of chemical sensors and parsed by neural circuits to recognize odor objects that can inform behavioral decisions. Whether and how mice segment odor mixtures into individual components (“demix”) remains unclear. We have found that mice can be trained to recognize individual odorants embedded in unpredictable and variable background mixtures with high degree of success . Despite nonlinear interactions and variability in the representations of odor mixtures by odorant receptors , a simple linear feedforward decoding is sufficient to explain the performance of mice in this task . Current experiments are aimed at understanding how the mouse brain represents information about odor mixtures to aid odor object identification and categorization.
Stephan Saalfeld, HHMI Janelia Research Campus, Ashburn, VA.
Louis Scheffer, HHMI Janelia Research Campus, Ashburn, VA.
Completing the Fly Model?
We are on the verge of having a full connectome of all neurons and their chemical synapses in Drosophila. But while necessary, this is not sufficient to model the fly's nervous system. Additional information includes gap junction varieties and locations, identities of neurotransmitters, receptor types and locations, neuromodulators and hormones (with sources and receptors), the role of glial cells, time evolution rules for synapses, and more. This talk will outline these missing pieces, look at methods by which the data might be obtained, and summarize existing efforts (where they exist) to get this needed information.
Srinivas C. Turaga, HHMI Janelia Research Campus, Ashburn, VA.
Connecting the Structure and Function of Neural Circuits
In this talk, I will describe how we developed deep learning based computational tools to solve two problems in neuroscience: inferring the activity of a neural network from measurements of its structural connectivity, and inferring the connectivity of a network of neurons from measurements and perturbation of neural activity.
- Can we infer neural connectivity from noisy measurement and perturbation of neural activity? Population neural activity measurement by calcium imaging can be combined with cellular resolution optogenetic activity perturbations to enable the mapping of neural connectivity in vivo. This requires accurate inference of perturbed and unperturbed neural activity from calcium imaging measurements, which are noisy and indirect. We built on recent advances in variational autoencoders to develop a new fully Bayesian approach to jointly inferring spiking activity and neural connectivity from in vivo all-optical perturbation experiments. Our model produces excellent spike inferences and predicts connectivity for mouse primary visual cortex which is consistent with known measurements.
- Are measurements of the structural connectivity of a biological neural network sufficient to predict its function? We constructed a simplified model of the first two stages of the fruit fly visual system, the lamina and medulla. The result is a deep hexagonal lattice convolutional neural network which discovered well-known orientation and direction selectivity properties in T4 neurons and their inputs. Our work demonstrates how knowledge of precise neural connectivity, combined with knowledge of the function of the circuit, can enable in silico predictions of the functional properties of individual neurons in a circuit, leading to an understanding of circuit function from structure.
More information about BCMC 2019 can be found here.