Journal of Computational Neuroscience

Artificial neural networks were used to explore the auditory function of the Mauthner system, the brainstem circuit in teleost fishes that initiates fast-start escape responses. The artificial neural networks were trained with backpropagation to assign connectivity and receptive fields in an architecture consistent with the known anatomy of the Mauthner system. Our first goal was to develop neurally specific hypotheses for how the Mauthner system discriminates right from left in the onset of a sound. Our model was consistent with the phase model for directional hearing underwater, the prevalent theory for sound source localization by fishes. Our second goal was to demonstrate how the neural mechanisms that permit sound localization according to the phase model can coexist with the mechanisms that permit the Mauthner system to discriminate between stimuli based on amplitude. Our results indicate possible computational roles for elements of the Mauthner system, which has provided us a theoretical context within which to consider past and future experiments on the cellular physiology. Thus, these findings demonstrate the potential significance of this approach in generating experimentally testable hypotheses for small systems of identified cells.

Continuous attractor networks require calibration. Computational models of the head direction (HD) system of the rat usually assume that the connections that maintain HD neuron activity are pre-wired and static. Ongoing activity in these models relies on precise continuous attractor dynamics. It is currently unknown how such connections could be so precisely wired, and how accurate calibration is maintained in the face of ongoing noise and perturbation. Our adaptive attractor model of the HD system that uses symmetric angular head velocity (AHV) cells as a training signal shows that the HD system can learn to support stable firing patterns from poorly-performing, unstable starting conditions. The proposed calibration mechanism suggests a requirement for symmetric AHV cells, the existence of which has previously been unexplained, and predicts that symmetric and asymmetric AHV cells should be distinctly different (in morphology, synaptic targets and/or methods of action on postsynaptic HD cells) due to their distinctly different functions.

Despite their structured receptive fields (RFs) and the strong linear components in their responses, most simple cells in mammalian visual cortex exhibit nonlinear behaviors. Besides the contrast-response function, nonlinearities are evident in various types of failure at superposition tasks, in the disagreement between direction indices computed from drifting and counterphase flickering gratings, in various forms of response suppression (including end- and side-stopping, spatial-frequency-specific inhibition and cross-orientation inhibition), in the advance of phase with increasing contrast, and in phase-insensitive and frequency-doubled responses to counterphase flickering gratings. These behaviors suggest that nonlinearities are involved in the operation of simple cells, but current models fail to explain them. A quantitative model is presented here that purports to describe basic and common principles of operation for all visual cortical cells. Simple cells are described as receiving afferents from multiple subunits that differ in their individual RFs and temporal impulse responses (TIRs). Subunits are independent and perform a spatial integration across their RFs followed by halfwave rectification and temporal convolution with their TIRs. This parallel operation yields a set of temporal functions representing each subunit's contribution to the membrane potential of the host cell, whose final form is given by the weighted sum of all subunits' contributions. By varying the number of subunits and their particular characteristics, different instances of the model are obtained each of which displays a different set of behaviors. Extensive simulation results are presented that illustrate how all of the reported nonlinear behaviors of simple cells arise from these multi-subunit organizations.

Statistical inferences are essentially important in analyzing neural spike trains in computational neuroscience. Current approaches have followed a general inference paradigm where a parametric probability model is often used to characterize the temporal evolution of the underlying stochastic processes. To directly capture the overall variability and distribution in the space of the spike trains, we focus on a data-driven approach where statistics are defined and computed in the function space in which spike trains are viewed as individual points. To this end, we at first develop a parametrized family of metrics that takes into account different warpings in the time domain and generalizes several currently used spike train distances. These new metrics are essentially penalized L p norms, involving appropriate functions of spike trains, with penalties associated with time-warping. The notions of means and variances of spike trains are then defined based on the new metrics when p = 2 (corresponding to the “Euclidean distance”). Using some restrictive conditions, we present an efficient recursive algorithm, termed Matching-Minimization algorithm, to compute the sample mean of a set of spike trains with arbitrary numbers of spikes. The proposed metrics as well as the mean spike trains are demonstrated using simulations as well as an experimental recording from the motor cortex. It is found that all these methods achieve desirable performance and the results support the success of this novel framework.

Correlated neural activities such as synchronizations can significantly alter the characteristics of spike transfer between neural layers. However, it is not clear how this synchronization-dependent spike transfer can be affected by the structure of convergent feedforward wiring. To address this question, we implemented computer simulations of model neural networks: a source and a target layer connected with different types of convergent wiring rules. In the Gaussian-Gaussian (GG) model, both the connection probability and the strength are given as Gaussian distribution as a function of spatial distance. In the Uniform-Constant (UC) and Uniform-Exponential (UE) models, the connection probability density is a uniform constant within a certain range, but the connection strength is set as a constant value or an exponentially decaying function, respectively. Then we examined how the spike transfer function is modulated under these conditions, while static or synchronized input patterns were introduced to simulate different levels of feedforward spike synchronization. We observed that the synchronization-dependent modulation of the transfer function appeared noticeably different for each convergence condition. The modulation of the spike transfer function was largest in the UC model, and smallest in the UE model. Our analysis showed that this difference was induced by the different spike weight distributions that was generated from convergent synapses in each model. Our results suggest that, the structure of the feedforward convergence is a crucial factor for correlation-dependent spike control, thus must be considered important to understand the mechanism of information transfer in the brain.

The suprachiasmatic nucleus (SCN) is known to be the master biological clock in mammals. Despite the periodic mean firing rate, interspike interval (ISI) patterns of SCN neurons are quite complex and irregular. The aim of the present study was to investigate the existence of nonlinear determinism in the complex ISI patterns of SCN neurons. ISI sequences were recorded from 173 neurons in rat hypothalamic slice preparations using a cell-attached patch recording technique. Their correlation dimensions (D2) were estimated, and were then compared with those of the randomly-shuffled surrogate data. We found that only 16 neurons (16/173) exhibited deterministic ISI patterns of spikes. In addition, clustering analysis revealed that SCN neurons could be divided into two subgroups of neurons each having distinct values of coefficient of variation (CV) and skewness (SK). Interestingly, most deterministic SCN neurons (14/16) belonged to the group of irregularly spiking neurons having large CV and SK values. To see if the neuronal coupling mediated by the γ-aminobutyric acid (GABA), the major neurotransmitter in the SCN, contributed to the deterministic nature, we examined the effect of the GABAA receptor antagonist bicuculline on D2 values of 56 SCN neurons. 8 SCN neurons which were originally stochastic became to exhibit deterministic characteristics after the bicuculline application. This result suggests that the deterministic nature of the SCN neurons arises not from GABAergic synaptic interactions, but likely from properties inherent to neurons themselves.

Within biologically constrained models of heading and complex motion processing, localization of the center-of-motion (COM) is typically an implicit property arising from the precise computation of radial motion direction associated with an observer’s forward self-motion. In the work presented here we report psychophysical data from a motion-impaired stroke patient, GZ, whose pattern of visual motion deficits is inconsistent with this view. We show that while GZ is able to discriminate direction in circular motions she is unable to discriminate direction in radial motion patterns. GZ’s inability to discriminate radial motion is in stark contrast with her ability to localize the COM in such stimuli and suggests that recovery of the COM does not necessarily require an explicit representation of radial motion direction. We propose that this dichotomy can be explained by a circular template mechanism that minimizes a global motion error relative to the visual motion input, and we demonstrate that a sparse population of such templates is computationally sufficient to account for human psychophysical performance in general and in particular, explains GZ’s performance. Recent re-analysis of the predicted receptive field structures in several existing heading models provides additional support for this type of circular template mechanism and suggests the human visual system may have available circular motion mechanisms for heading estimation.

We study evoked calcium dynamics in astrocytes, a major cell type in the mammalian brain. Experimental evidence has shown that such dynamics are highly variable between different trials, cells, and cell subcompartments. Here we present a qualitative analysis of a recent mathematical model of astrocyte calcium responses. We show how the major response types are generated in the model as a result of the underlying bifurcation structure. By varying key channel parameters, mimicking blockers used by experimentalists, we manipulate this underlying bifurcation structure and predict how the distributions of responses can change. We find that store-operated calcium channels, plasma membrane bound channels with little activity during calcium transients, have a surprisingly strong effect, underscoring the importance of considering these channels in both experiments and mathematical settings. Variation in the maximum flow in different calcium channels is also shown to determine the range of stable oscillations, as well as set the range of frequencies of the oscillations. Further, by conducting a randomized search through the parameter space and recording the resulting calcium responses, we create a database that can be used by experimentalists to help estimate the underlying channel distribution of their cells.

Integration of multiple sensory cues can improve performance in detection and estimation tasks. There is an open theoretical question of the conditions under which linear or nonlinear cue combination is Bayes-optimal. We demonstrate that a neural population decoded by a population vector requires nonlinear cue combination to approximate Bayesian inference. Specifically, if cues are conditionally independent, multiplicative cue combination is optimal for the population vector. The model was tested on neural and behavioral responses in the barn owl’s sound localization system where space-specific neurons owe their selectivity to multiplicative tuning to sound localization cues interaural phase (IPD) and level (ILD) differences. We found that IPD and ILD cues are approximately conditionally independent. As a result, the multiplicative combination selectivity to IPD and ILD of midbrain space-specific neurons permits a population vector to perform Bayesian cue combination. We further show that this model describes the owl’s localization behavior in azimuth and elevation. This work provides theoretical justification and experimental evidence supporting the optimality of nonlinear cue combination.

A distinction is commonly made between synaptic connections capable of evoking a response (“drivers”) and those that can alter ongoing activity but not initiate it (“modulators”). Here it is proposed that, in cortex, both drivers and modulators are an emergent property of the perceptual inference performed by cortical circuits. Hence, it is proposed that there is a single underlying computational explanation for both forms of synaptic connection. This idea is illustrated using a predictive coding model of cortical perceptual inference. In this model all synaptic inputs are treated identically. However, functionally, certain synaptic inputs drive neural responses while others have a modulatory influence. This model is shown to account for driving and modulatory influences in bottom-up, lateral, and top-down pathways, and is used to simulate a wide range of neurophysiological phenomena including surround suppression, contour integration, gain modulation, spatio-temporal prediction, and attention. The proposed computational model thus provides a single functional explanation for drivers and modulators and a unified account of a diverse range of neurophysiological data.