Final Report Summary - CONEURON (Drawing neuronal circuits without seeing them)
To describe and understand the complex, high dimensional neuronal data arising from state of the art recording techniques, we need to develop a new mathematical framework that is algorithmically efficient and at the same time incorporates our knowledge about neuronal electrophysiology. The CoNeuron project combines computational neuroscience with the analysis of neuronal and behavioral data to (a) describe high dimensional neuronal data, (b) understand how neuronal networks can generate non-trivial behaviors, (c) predict neuronal interactions from the observed activity patterns, and (d) correlate activity patterns to psychological phenomena. Given the broad scope of the problems faced and the variety of methods required, this project is clearly multidisciplinary, involving techniques from physics, mathematics and neuroscience.
• Project objectives
Aim 1. Development of mathematical tools for describing the complex activity patterns in realistic neuronal networks.
Aim 2. Estimation of connectivity in realistic neuronal networks using efficient algorithms
Aim 3. Analysis of neuronal data and estimation of connectivity in neuronal circuits
Aim 4. Study of information transmission in neuronal networks
• Description of the work performed and main results
1 “Poisson-like spiking with probabilistic synapses” Rubén Moreno-Bote (PLoS Comp Biol., 2014). Neuronal activity in cortex is variable both spontaneously and during stimulation, and has the remarkable property that it is Poisson-like over broad ranges of firing rates covering from virtually zero to hundreds of spikes per second. The mechanisms that lead to cortical-like spiking variability over such broad range are currently unknown. We have shown that neuronal networks endowed with probabilistic synaptic transmission, a well-documented source of variability in cortex, robustly generate Poisson-like variability over several orders of magnitude in their firing rate without fine-tuning of the network parameters. This mechanism predicts that the size of membrane potential fluctuations is approximately constant with firing rate and that the variance of the synaptic conductances is proportional to the mean conductances. We finally have shown that probabilistic synapses allow networks to sample their states in a contrast-invariant manner, providing cortical circuits with a natural way to perform probabilistic inference in ambiguous conditions. This project addresses Aim 1.
2 “Multiplicative and additive modulation of neuronal tuning with population activity affects encoded information” Iñigo Arandia-Romero, Seiji Tanabe, Jan Drugowitsch, Adam Kohn, Alexandre Pouget and Rubén Moreno-Bote (Submitted, 2015). Numerous studies have shown that neuronal responses are modulated by stimulus properties, and also by the state of the local network. However, little is known about how activity fluctuations of neuronal populations modulate the sensory tuning of cells and affect their encoded information. We found that fluctuations in ongoing and stimulus-evoked population activity in primate visual cortex modulate the tuning of individual neurons in a multiplicative and additive manner. Neurons with strong multiplicative effects tended to have little additive modulation, and vice versa. The information encoded by multiplicatively-modulated neurons increased with greater population activity, while that of additively-modulated neurons decreased. These effects offset each other, so that population activity had little effect on total information. Our results thus suggest that intrinsic activity fluctuations act as a ‘traffic light’ that determines which subset of neurons are most informative. The most salient results are: (1) tuning of neurons in monkey primary visual cortex is modulated by fluctuations in population activity, (2) neurons with strong multiplicative modulation have weak additive modulation, and vice versa, (3) encoded information increases with population activity for multiplicatively-modulated neurons, and decreases for additively-modulated neurons, and (4) population activity thus determines which neurons are most informative, without affecting total information. This project addresses Aims 2 and 3.
3 “Information limiting correlations” Rubén Moreno-Bote, Jeffrey Beck, Ingmar Kanischeider, Xaq Pitkow, Peter Latham and Alexandre Pouget (Nature Neuroscience, 2014). Neuronal responses are typically variable in the sense that the number and timing of the spikes in response to the same stimulus varies from trial to trial. This variability can greatly reduce the precision of the neural code, because several values of the encoded stimulus might be compatible with the observation of a given spike count. This problem can be reduced through averaging when the code involves the activity of a large population of neurons but the efficiency of this averaging process depends greatly on the pattern of correlations across neuron. We show that for realistic correlations, small correlations do not necessarily imply large information content and information is not limited by correlations between neurons with similar preferred stimuli. Instead, information is limited by correlations proportional to the product of the derivative of the sensory tuning curves. This project addresses Aim 4.
• Expected final results and impact
A critical problem in neuroscience is to uncover the way neuronal circuits perform computations that allow capabilities like perception, decision making and attention. Unraveling the fine details of how neurons are connected with their neighboring neurons represents a challenge today and it is a necessary step to understand how the brain works. Our work will allow us to move towards this goal. The work undertaken in the project is of particular relevance at this time and can has an important impact on society because we will help to advance in the interpretation of the complex, high dimensional neuronal data, and provide a conceptual framework that will be critical for the next generation of questions and problems that will be faced in neuroscience. In particular, we will move towards a better understanding of how small neuronal circuits process and transform information.