Biology subjects

Mihalas, S.

Publications and source records attributed to Mihalas, S..

51 records · Page 3Linked to original sources

bioRxiv2020

Similar visual perception in GCaMP6 transgenic mice despite differences in learning and motivation

To study mechanisms of perception and cognition, neural measurements must be made during behavior. A goal of the Allen Brain Observatory is to map activity in distinct cortical cell classes during visual processing and behavior. Here we characterize learning and performance of five GCaMP6-expressing transgenic lines trained on a visual change detection task. We used automated training procedures to facilitate comparisons across mice. Training times varied, but most transgenic mice learned the task. Motivation levels also varied across mice. To compare mice in similar motivational states we subdivided sessions into over-, under-, and optimally motivated periods. When motivated, the pattern of perceptual decisions were highly correlated across transgenic lines, although overall d-prime was lower in one line labeling somatostatin inhibitory cells. These results provide important context for using these mice to map neural activity underlying perception and behavior.

neuroscience↗

bioRxiv2019

Quantifying and Modelling Transfer Learning in Mice Between Consecutive Training Stages of a Change Detection Task

Animals are known to be able to rapidly transfer knowledge between tasks with similar structure. We trained a set of mice on a visual change detection task with multiple stages, starting with direct transitions between gratings, adding an intervening gray screen, and subsequently moving to multiple sets of natural images. We observe that, when progressing to new stages, the performance increases very fast. However, when transitioning to a task of higher complexity, the peak performance decreases. This setup facilitates for the first time an experimental platform to study the transfer learning phenomena in mice using visual stimuli. Based on these results and additional neuroscience insight, we propose a cognitive model to explain the quick adaptation observed in mice. It extends a deep Q learning agent with a multi-tiered architecture and the possibility of performing a representation remapping at every level of the hierarchy to prevent the downstream propagation of representation anomalies. This architecture provides the substrate of an adaptation algorithm based on ideas of optimal transport of probability distributions. It matches well key behavioral aspects observed in mice and the experimental constraint that the mice are initially trained using a single task variant before it transitions to the new training phase. The modelling process helped us to gain biological insights: first, the optimal transport mechanism of the representation remapping indicated that a possible reason for the reduced performance when mice move to a new more complex task could be due to limited representation resources which were optimized for the previous task. Second, the multi-tiered architecture was first an engineering constraint and later a biologic insight confirmed by the literature. A final insight came from the computations required to perform the representation remapping. These computations are interesting because they could help to confirm this transfer learning theory by looking for similar neural correlates during the adaptation process.

animal behavior and cognition↗

bioRxiv2019

Widespread presence of direction-reversing neurons in the mouse visual system

Direction selectivity - the preference of motion in one direction over the opposite - is a fundamental property of visual neurons across species. We find that a substantial proportion of direction selective neurons in the mouse visual system reverse their preferred direction of motion in response to drifting gratings at different spatiotemporal parameters. A spatiotemporally asymmetric filter model recapitulates our experimental observations.

neuroscience↗

bioRxiv2019

Experience shapes activity dynamics and stimulus coding of VIP inhibitory and excitatory cells

Cortical circuits are flexible and can change with experience and learning. However, the effects of experience on specific cell types, including distinct inhibitory types, are not well understood. Here we investigated how excitatory and VIP inhibitory cells in layer 2/3 of mouse visual cortex were impacted by visual experience in the context of a behavioral task. Mice learned to perform an image change detection task with a set of eight natural scene images, viewing these images thousands of times. Subsequently, during 2-photon imaging experiments, mice performed the task with these familiar images and three additional sets of novel images. Novel images evoked stronger overall activity in both excitatory and VIP populations, and familiar images were more sparsely coded by excitatory cells. The temporal dynamics of VIP activity differed markedly between novel and familiar images: VIP cells were stimulus-driven by novel images but displayed ramping activity during the inter-stimulus interval for familiar images. Moreover, when a familiar stimulus was omitted from an expected sequence, VIP cells showed extended ramping activity until the subsequent image presentation. This prominent shift in response dynamics suggests that VIP cells may adopt different modes of processing during familiar versus novel conditions.\n\nHIGHLIGHTSO_LIExperience with natural images in a change detection task reduces overall activity of cortical excitatory and VIP inhibitory cells\nC_LIO_LIEncoding of natural images is sharpened with experience in excitatory neurons\nC_LIO_LIVIP cells are stimulus-driven by novel images but show pre-stimulus ramping for familiar images\nC_LIO_LIVIP cells show strong ramping activity during the omission of an expected stimulus\nC_LI

neuroscience↗

bioRxiv2019

Does the neuronal noise in cortex help generalization?

One remarkable feature of neuronal activity in the mammalian cortex is the high level of variability in response to repeated stimuli. First, we used an open dataset, the Allen Brain Observatory, to quantify the distribution of responses to repeated presentations of natural movies. We find that even for their preferred moment in the movie clip, neurons have high variability which cannot be well captured by Gaussian or Poisson distributions. A large fraction of responses are better fit by log-normal or Gaussian mixture models with two components. These distributions are similar to activity distributions during training of deep neural networks using dropout. This poses the interesting hypothesis: is the role of cortical noise to help in generalization during learning?\n\nSecond, to ensure the robustness of our results we analyzed electrophysiological recordings in the same areas of mouse visual cortex, again using repeated natural movie presentations and found similar response distributions. To make sure that the trial-by-trial variations we observe are not due exclusively to the result of changes in state, we constructed a population coupling model, where each neurons activity is coupled to a low-dimension version of the activity of all other simultaneously recorded neurons. The population coupling model can capture global, brain-wide activity fluctuations that are state-dependent. The residuals from this model also show non-Gaussian noise distributions.\n\nThird, we ask a more specific question: is the noise in the cortex more likely to move the representation of the stimulus in-class versus out-of-class? To address this question, we analyzed the responses of neurons across trials from multiple sections of different movie clips. We observe that the noise in the cortex better aligns to in-class variations. We argue that a useful noise for learning generalizations is to move from representations of different exemplars in-class, similar to cortical noise.

neuroscience↗

bioRxiv2019

Systematic Integration of Structural and Functional Data into Multi-Scale Models of Mouse Primary Visual Cortex

AO_SCPLOWBSTRACTC_SCPLOWStructural rules underlying functional properties of cortical circuits are poorly understood. To explore these rules systematically, we integrated information from extensive literature curation and large-scale experimental surveys into a data-driven, biologically realistic model of the mouse primary visual cortex. The model was constructed at two levels of granularity, using either biophysically-detailed or point-neurons, with identical network connectivity. Both variants were compared to each other and to experimental recordings of neural activity during presentation of visual stimuli to awake mice. While constructing and tuning these networks to recapitulate experimental data, we identified a set of rules governing cell-class specific connectivity and synaptic strengths. These structural constraints constitute hypotheses that can be tested experimentally. Despite their distinct single cell abstraction, spatially extended or point-models, both perform similarly at the level of firing rate distributions. All data and models are freely available as a resource for the community.

neuroscience↗

bioRxiv2018

A large-scale, standardized physiological survey reveals higher order coding throughout the mouse visual cortex

To understand how the brain processes sensory information to guide behavior, we must know how stimulus representations are transformed throughout the visual cortex. Here we report an open, large-scale physiological survey of neural activity in the awake mouse visual cortex: the Allen Brain Observatory Visual Coding dataset. This publicly available dataset includes cortical activity from nearly 60,000 neurons collected from 6 visual areas, 4 layers, and 12 transgenic mouse lines from 221 adult mice, in response to a systematic set of visual stimuli. Using this dataset, we reveal functional differences across these dimensions and show that visual cortical responses are sparse but correlated. Surprisingly, responses to different stimuli are largely independent, e.g. whether a neuron responds to natural scenes provides no information about whether it responds to natural movies or to gratings. We show that these phenomena cannot be explained by standard local filter-based models, but are consistent with multi-layer hierarchical computation, as found in deeper layers of standard convolutional neural networks.

neuroscience↗

bioRxiv2018

Synchronization dependent on spatial structures of a mesoscopic whole-brain network

We study how the spatial structure of connectivity shapes synchronization in a system of coupled phase oscillators on a mammalian whole-brain network at the mesoscopic level. Complex structural connectivity of the mammalian brain is believed to underlie the versatility of neural computations. The Allen Mouse Brain Connectivity Atlas constructed from viral tracing experiments together with a new mapping algorithm reveals that the connectivity has a significant spatial dependence: the connection strength decreases with distance between the regions, following a power law. However, there are a number of residuals above the power-law fit, predominantly for long-range connections. We show how these strong connections between distal brain regions promote rapid transitions between highly localized synchronization and more global synchronization as the amount of dispersion in the frequency distribution changes. This may explain the brains ability to switch rapidly between global and modularized computations.

neuroscience↗

bioRxiv2018

Sparse recurrent excitatory connectivity in the cortical microcircuit of the adult mouse and human

Generating a comprehensive description of cortical networks requires a large-scale, systematic approach. To that end, the Allen Institute is engaged in a pipeline project using multipatch electrophysiology, supplemented with 2-photon optogenetics, to characterize connectivity and synaptic signaling between classes of neurons in adult mouse and human cortex. We focus on producing results detailed enough for the generation of computational models and enabling comparison with future studies. Here we report our examination of intralaminar connectivity within each of several classes of excitatory neurons. We find that connections are sparse but present among all excitatory cell types and layers we sampled, with the most sparse connections in layers 5 and 6. Almost all mouse synapses exhibited short-term depression with similar dynamics. Synaptic signaling between a subset of layer 2/3 neurons; however, exhibited facilitation. These results contribute to a body of evidence describing recurrent excitatory connectivity as a conserved feature of cortical microcircuits.

neuroscience↗

bioRxiv2018

The organization of intracortical connections by layer and cell class in the mouse brain

The mammalian cortex is a laminar structure composed of many cell types densely interconnected in complex ways. Recent systematic efforts to map the mouse mesoscale connectome provide comprehensive projection data on interareal connections, but not at the level of specific cell classes or layers within cortical areas. We present here a significant expansion of the Allen Mouse Brain Connectivity Atlas, with [~]1,000 new axonal projection mapping experiments across nearly all isocortical areas in 49 Cre driver lines. Using 13 lines selective for cortical layer-specific projection neuron classes, we identify the differential contribution of each layer/class to the overall intracortical connectivity patterns. We find layer 5 (L5) projection neurons account for essentially all intracortical outputs. L2/3, L4, and L6 neurons contact a subset of the L5 cortical targets. We also describe the most common axon lamination patterns in cortical targets. Most patterns are consistent with previous anatomical rules used to determine hierarchical position between cortical areas (feedforward, feedback), with notable exceptions. While diverse target lamination patterns arise from every source layer/class, L2/3 and L4 neurons are primarily associated with feedforward type projection patterns and L6 with feedback. L5 has both feedforward and feedback projection patterns. Finally, network analyses revealed a modular organization of the intracortical connectome. By labeling interareal and intermodule connections as feedforward or feedback, we present an integrated view of the intracortical connectome as a hierarchical network.

neuroscience↗

bioRxiv2018

High resolution data-driven model of the mouse connectome

Knowledge of mesoscopic brain connectivity is important for understanding inter- and intra-region information processing. Models of structural connectivity are typically constructed and analyzed with the assumption that regions are homogeneous. We instead use the Allen Mouse Brain Connectivity Atlas to construct a model of whole brain connectivity at the scale of 100 {micro}m voxels. The dataset used consists of 366 anterograde tracing experiments in wild type C7BL/6 mice, mapping fluorescently-labeled neuronal projections brain-wide. Inferring spatial connectivity with this dataset remains underdetermined, since the approximately 2 x 105 source voxels outnumber the number of experiments. To address this, we assume that connection patterns and strengths vary smoothly across major brain divisions. We model the connectivity at each voxel as a radial basis kernel-weighted average of the projection patterns of nearby injections. The voxel model outperforms a previous regional model in predicting held-out experiments and compared to a human-curated dataset. This voxel-scale model of the mouse connectome permits researchers to extend their previous analyses of structural connectivity to unprecedented levels of resolution, and allows for comparison with functional imaging and other datasets.

neuroscience↗

bioRxiv2018

Visual physiology of the Layer 4 cortical circuit in silico

Despite advances in experimental techniques and accumulation of large datasets concerning the composition and properties of the cortex, quantitative modeling of cortical circuits under in-vivo-like conditions remains challenging. Here we report and publicly release a biophysically detailed circuit model of layer 4 in the mouse primary visual cortex, receiving thalamo-cortical visual inputs. The 45,000-neuron model was subjected to a battery of visual stimuli, and results were compared to published work and new in vivo experiments. Simulations reproduced a variety of observations, including effects of optogenetic perturbations. Critical to the agreement between responses in silico and in vivo were the rules of functional synaptic connectivity between neurons. Interestingly, after extreme simplification the model still performed satisfactorily on many measurements, although quantitative agreement with experiments suffered. These results emphasize the importance of functional rules of cortical wiring and enable a next generation of data-driven models of in vivo neural activity and computations.\n\nAUTHOR SUMMARYHow can we capture the incredible complexity of brain circuits in quantitative models, and what can such models teach us about mechanisms underlying brain activity? To answer these questions, we set out to build extensive, bio-realistic models of brain circuitry employing systematic datasets on brain structure and function. Here we report the first modeling results of this project, focusing on the layer 4 of the primary visual cortex (V1) of the mouse. Our simulations reproduced a variety of experimental observations in a large battery of visual stimuli. The results elucidated circuit mechanisms determining patters of neuronal activity in layer 4 - in particular, the roles of feedforward thalamic inputs and specific patterns of intracortical connectivity in producing tuning of neuronal responses to the orientation of motion. Simplification of neuronal models led to specific deficiencies in reproducing experimental data, giving insights into how biological details contribute to various aspects of brain activity. To enable future development of more sophisticated models, we make the software code, the model, and simulation results publicly available.

neuroscience↗

bioRxiv2017

Cortical circuits implement optimal context integration

Neurons in the primary visual cortex (V1) predominantly respond to a patch of the visual input, their classical receptive field. These responses are modulated by the visual input in the surround [2]. This reflects the fact that features in natural scenes do not occur in isolation: lines, surfaces are generally continuous, and the surround provides context for the information in the classical receptive field. It is generally assumed that the information in the near surround is transmitted via lateral connections between neurons in the same area [2]. A series of large scale efforts have recently described the relation between lateral connectivity and visual evoked responses and found like-to-like connectivity between excitatory neurons [16, 18]. Additionally, specific cell type connectivity for inhibitory neuron types has been described [11, 31]. Current normative models of cortical function relying on sparsity [27], saliency [4] predict functional inhibition between similarly tuned neurons. What computations are consistent with the observed structure of the lateral connections between the excitatory and diverse types of inhibitory neurons?\n\nWe combined natural scene statistics [24] and mouse V1 neuron responses [7] to compute the lateral connections and computations of individual neurons which optimally integrate information from the classical receptive field with that from the surround by directly implementing Bayes rule. This increases the accuracy of representation of a natural scene under noisy conditions. We show that this network has like-to-like connectivity between excitatory neurons, similar to the observed one [16, 18, 11], and has three types of inhibition: local normalization, surround inhibition and gating of inhibition from the surround - that can be attributed to three classes of inhibitory neurons. We hypothesize that this computation: optimal integration of contextual cues with a gate to ignore context when necessary is a general property of cortical circuits, and the rules constructed for mouse V1 generalize to other areas and species.

neuroscience↗

bioRxiv2017

Generalized Leaky Integrate-And-Fire Models Classify Multiple Neuron Types

In the mammalian neocortex, there is a high diversity of neuronal types. To facilitate construction of system models with multiple cell types, we generate a database of point models associated with the Allen Cell Types Database. We construct a series of generalized integrate-and-fire (GLIF) models of increasing complexity aiming to reproduce the spiking behaviors of the recorded neurons. We test the performance of these GLIF models on data from 771 neurons from 14 transgenic lines, with increasing model performance for more complex models. To answer how complex a model needs to be to reproduce the number of electrophysiological cell types, we perform unsupervised clustering on the parameters extracted from these models. The number of clusters is smaller for individual model types, but when combining all GLIF parameters 18 clusters are obtained, while 11 clusters are obtained using 16 electrophysiological features. Therefore, these low dimensional models have the capacity to characterize the diversity of cell types without the need for a priori defined features.

neuroscience↗

bioRxiv2016

Visual processing mode switching regulated by VIP cells

The responses of neurons in mouse primary visual cortex (V1) to visual stimuli depend on behavioral states. Specifically, surround suppression is reduced during locomotion. Although locomotion-induced vasoactive intestinal polypeptide positive (VIP) interneuron depolarization can account for the reduction of surround suppression, the functions of VIP cell depolarization are not fully understood. Here we utilize a firing rate model and a computational model to elucidate the potential functions of VIP cell depolarization during locomotion. Our analyses suggest 1) that surround suppression sharpens the visual responses in V1 to a stationary scene, 2) that depolarized VIP cells enhance V1 responses to moving objects by reducing self-induced surround suppression and 3) that during locomotion V1 neuron responses to some features of the moving objects can be selectively enhanced. Thus, VIP cells regulate surround suppression to allow pyramidal neurons to optimally encode visual information independent of behavioral state.

neuroscience↗