Biology subjects

Miller, K. D.

Publications and source records attributed to Miller, K. D..

At least 19 recordsLinked to original sources

bioRxiv2026

Linear NeuroPaint: An Efficient Baseline for Cross-Session Neural Alignment and Inpainting

Large-scale Neuropixels recordings sample many brain areas, but no single experimental session records every area of interest. NeuroPaint, a transformer-based autoencoder, integrates such recordings by aligning area-specific latent dynamics across sessions and "inpainting" activity in unrecorded areas. Could a linear model, which would be computationally efficient and potentially more interpretable, suffice to achieve this? To address this question, we introduce Linear NeuroPaint, a linear variant of NeuroPaint, and evaluate it on two multi-area Neuropixels datasets. Linear NeuroPaint achieves strong predictive performance on recorded areas, however, it is less accurate than nonlinear models at cross-area inpainting, despite extensive hyperparameter optimization. Moreover, its parameters are not readily interpretable: incomplete cross-session alignment leads parameter-defined neuron clusters to exhibit heterogeneous response profiles. Together, these results show that linear transformations are insufficient for cross-session alignment and cross-area prediction. Nevertheless, Linear NeuroPaint provides a computationally efficient baseline for benchmarking multi-session neural models.

neuroscience↗

bioRxiv2026

Inter-regional interactions uncouple within-region inhibition stabilization and paradoxical responses

Paradoxical responses, in which excitatory perturbations of inhibitory neurons reduce their firing rates, are widely regarded as a hallmark of inhibition-stabilized networks (ISNs) and have been observed in multiple brain regions. However, because brain regions are interconnected by long-range projections, it remains unclear whether a paradoxical response observed in a given region reflects inhibition stabilization within that region or instead arises from distributed interactions among regions. Here, using analytically tractable multi-region population models and numerical simulations, we show that inter-regional connections can dissociate local inhibition stabilization from paradoxical responses. The condition for a paradoxical response in a given region is jointly determined by recurrent excitation within that region, feedback mediated through other regions, and the dynamics of those other regions. Consequently, inter-regional coupling can generate a paradoxical response in a region that is not locally inhibition-stabilized or abolish it in a region that is. Thus, in an interconnected network, a paradoxical response is neither necessary nor sufficient for local inhibition stabilization. These findings demonstrate that local perturbation responses cannot generally be interpreted solely in terms of local circuit dynamics and highlight the importance of accounting for inter-regional connections when inferring the local circuit properties of individual brain regions.

neuroscience↗

bioRxiv2026

A nonlinear inhibition pathway underlying cortical responses to tuned holographic optogenetic perturbations

Optogenetics enables causal manipulation of cortical activity. Perturbation responses can be counterintuitive due to network interactions, making theory essential for predicting them. Existing approaches often rely on linear approximations, which fail for many biologically relevant perturbations. Here we develop a nonlinear theory of responses to holographic perturbations in cell-type-specific recurrent networks with structured connectivity. We fit a nonlinear model to mouse V1 data, which shows cotuned-ensemble suppression: perturbing spatially clustered neurons with similar preferred orientations yields markedly stronger short-range suppression than perturbing untuned ensembles. We show that cotuned-ensemble suppression arises from a feature-tuned, nonlinear inhibition pathway implicating somatostatin-positive (SST) interneurons. The theory predicts that cotuned ensembles suppress parvalbumin-positive (PV) neurons but facilitate SST neurons, and links the degree of cotuned-ensemble suppression or facilitation to the variance of the SST response. This framework identifies mechanisms by which nonlinear inhibition sculpts cortical dynamics and establishes a predictive basis for targeted optogenetic interventions.

neuroscience↗

bioRxiv2026

Characteristics and dynamical signatures of recurrent cortical circuits during context-dependent processing

Context profoundly shapes neural responses and behavior. During context-dependent sensory processing, recurrent connections shape the integration of feedforward sensory input and feed-back input from downstream brain regions. How do different cell types, interacting through spatially structured recurrent lateral connections, give rise to context-dependent processing and circuit stability, and what dynamical signatures reveal their individual roles? To answer these questions, we employ data-driven approaches to construct spatially extended stabilized supralinear network models that capture the responses of diverse cell types in the mouse primary visual cortex during context-dependent processing. Analysis of well-fitting models reveals that the dominant inhibitory cell type affecting excitatory neurons is not fixed but dynamically varies with stimulus and space. While PV-mediated stabilization is indispensable across all models and stimulus conditions, SST-mediated stabilization is also required, and likely in a stimulus-dependent manner. Interestingly, even when a specific inhibitory cell type is required for circuit stabilization, a uniform perturbation of it does not necessarily produce a paradoxical change in its mean activity. Instead, assessing cell-type-specific circuit stabilization requires patterned perturbations, where paradoxical effects manifest along specific activity modes. Finally, we show that recurrent connections and input-output nonlinearities are essential for integrating feedforward and feedback inputs to reproduce the observed spatial response profiles. Recurrent excitatory connections, in particular, are required to account for responses to small stimuli, where external inputs are relatively weak. Taken together, our work reveals the crucial role of ubiquitous biological components in context-dependent processing and delineates the characteristics and dynamical signatures of these circuits.

neuroscience↗

bioRxiv2025

Natural Scene Coding Consistency in Genetically-Defined Cell Populations

Understanding how genetically-defined cell populations encode visual information remains a fundamental challenge in systems neuroscience. While extensive research has characterized individual cell responses to simple stimuli such as static gratings, the population-level coding principles that govern naturalistic visual processing across cell types remain largely unexplored. We analyzed population responses from 43,018 neurons across 12 genetically-defined cell types in 243 mice from the Allen Brain Observatory, comparing representational geometry between natural scenes and static gratings. We found that inhibitory cell populations (VIP, SST, PV) cluster distinctly in representational space almost independent of anatomical location when responding to natural scenes but not static gratings, suggesting preserved cell-type specific computational functions specific for natural scenes. To assess coding capacity of a population of neurons, we developed Inter-Individual Representational Similarity (IIRS), which measures consistency of neural representations across different individuals in response to an ensemble of stimuli. All inhibitory populations showed significantly higher IIRS for natural scenes compared to static gratings, indicating consistent encoding of naturalistic visual features across individuals comparable to excitatory populations (Cux2, Rorb, Rbp4). Parallel analysis of neural networks trained on natural images (ImageNet) with different random initializations revealed similar patterns: models showed higher cross-initialization similarity for naturalistic stimuli compared to static gratings, suggesting that cross-individual consistency emerges when experimental stimuli engage the features that neural circuits are adapted to extract. These findings establish IIRS as a metric for identifying coding capacity in cell populations and reveal that inhibitory cell populations encode consistent aspects of natural scenes across individuals, indicating these circuits may have evolved specialized tuning for naturalistic visual environments.

neuroscience↗

bioRxiv2025

Generative inference unifies feedback processing for learning and perception in natural and artificial vision

We understand how neurons respond selectively to patterns in visual input to support object recognition; however, how these circuits support perceptual grouping, illusory percepts, and imagination is not understood. These perceptual experiences are thought to require combining what we have learned about the world with incoming sensory signals, yet the neural mechanism for this integration remains unclear. Here we show that networks tuned for object recognition implicitly learn the distribution of their input, which can be accessed through feedback connections that tune synaptic weights. We introduce Generative Inference, a computational framework in which feedback pathways that adjust connection weights during learning are repurposed during perception to combine learned knowledge with sensory input, fulfilling flexible inference goals such as increasing confidence. Generative Inference enables networks tuned solely for recognition to spontaneously produce perceptual grouping, illusory contours, shape completion, and pattern formation resembling imagination, while preserving their recognition abilities. The framework reproduces neural signatures observed across perceptual experiments: delayed responses in feedback-receiving layers of early visual cortex that disappear when feedback connections are disrupted. We show that, under stated assumptions, gradients of classification error approximate directions that are informative about the data distribution, establishing a theoretical connection between recognition and generation. Together, these findings show that pattern recognition and pattern generation rely on a shared computational substrate through dual use of feedback pathways. This principle explains how neural systems recognize familiar objects reliably while remaining flexible enough to interpret incomplete or ambiguous information, and suggests that reusing learning signals for perception may be a general feature of both biological brains and artificial networks.

neuroscience↗

bioRxiv2025

Theory of interaction between untuned modulatory inputs and tuned sensory inputs

How does the brain integrate sensory inputs with non-feature-tuned signals, such as those arising from behavioral state changes or neuromodulation? Here, we show that the dynamics of disordered E/I networks with structured, feature-dependent connectivity can be well characterized by an effective model describing interactions between the responses of cells who prefer the current sensory stimulus ("matched" cells) and the responses of cells firing at the baseline. This effective network exhibits strong feedback from the baseline onto the matched responses but weak reverse projections. Thus, an untuned stimulus not only directly drives matched cells, but also indirectly drives them via modulation of the baseline. We demonstrate through a linear response analysis that the baseline effect on the matched response is suppressive if the network is strongly coupled and feedback-inhibition dominated. In particular, in this regime, feature-dependent networks produce "rate reshuffling", wherein untuned optogenetic excitation yields large changes in the individual responses of matched cells without significantly changing their overall firing rate distribution, as the optogenetically-induced baseline response suppresses the matched response. Finally, if multiple sensory stimuli are presented, yielding sublinear response summation ("normalization"), the influence of the baseline on the matched responses is weakened. Thus, an untuned (e.g., optogenetic) stimulus is less suppressive to multiple stimuli than to a single stimulus, making normalization effectively weaker in the presence of an untuned stimulus. Our framework provides the first theory of the interaction of untuned modulatory and tuned sensory inputs, reconciles prior experiments, and provides testable predictions about tuned-untuned interactions in cortical processing. Significance StatementSensory cortex receives both tuned sensory inputs, and untuned signals e.g. from global stimulus changes, behavioral state changes, or neuromodulation. We lack a theory of how these inputs are integrated. We demonstrate reduction of circuit models to a model of interactions between baseline and sensory-stimulus-matched responses, and develop an exact linear response analysis of activity perturbation by untuned inputs. We show baseline-to-matched coupling is strong, while matched-to-baseline is weak, and the former is suppressive given sufficiently strong and inhibition-dominated connectivity. We extend the theory to multiple tuned stimuli. The theory offers a mechanistic explanation of previous surprising observations ("rate reshuffling"), yields new predictions, and provides a general framework for understanding the impact of modulatory influences on sensory processing.

neuroscience↗

bioRxiv2025

The geometry of the neural state space of decisions

How do populations of neurons collectively encode and process information during cognitive tasks? We analyze high-yield population recordings from the macaque lateral intraparietal area (LIP) during a reaction-time random-dot-motion direction-discrimination task. We find that the trajectories of neural population activity patterns during single decisions lie within a curved two-dimensional manifold. The reaction time of trajectories systematically varies along one dimension, such that slow and fast decisions trace distinct activity patterns. Trajectories transition from a deliberation stage, in which they are noisy and remain similar between the choices, to a commitment stage, in which they are far less noisy and diverge sharply for the different choices. The deliberation phase is pronounced for slower decisions and gradually diminishes as reaction time decreases. A mechanistic circuit model provides an explanation for the observed properties, and suggests the transition between stages represents a transition from more sensory-driven to more circuit-driven dynamics. It yields two striking predictions we verify in the data. First, whether neurons are more choice selective for slow or fast trials varies systematically with the retinotopic location of their response fields. Second, the slower the trial, the more saccades undershoot the choice target. The results highlight the roles of distributed and dynamic activity patterns and intrinsic circuit dynamics in the population implementation of a cognitive task.

neuroscience↗

bioRxiv2024

Exact linear theory of perturbation response in a space- and feature-dependent cortical circuitmodel

What are the principles that govern the responses of cortical networks to their inputs and the emergence of these responses from recurrent connectivity? Recent experiments have probed these questions by measuring cortical responses to two-photon optogenetic perturbations of single cells in the mouse primary visual cortex. A robust theoretical framework is needed to determine the implications of these responses for cortical recurrence. Here we propose a novel analytical approach: a formulation of the dependence of cell-type-specific connectivity on spatial distance that yields an exact solution for the linear perturbation response of a model with multiple cell types and space- and feature-dependent connectivity. Importantly and unlike previous approaches, the solution is valid in regimes of strong as well as weak intra-cortical coupling. Analysis reveals the structure of connectivity implied by various features of single-cell perturbation responses, such as the surprisingly narrow spatial radius of nearby excitation beyond which inhibition dominates, the number of transitions between mean excitation and inhibition thereafter, and the dependence of these responses on feature preferences. Comparison of these results to existing optogenetic perturbation data yields constraints on cell-type-specific connection strengths and their tuning dependence. Finally, we provide experimental predictions regarding the response of inhibitory neurons to single-cell perturbations and the modulation of perturbation response by neuronal gain. Significance StatementThe cerebral cortex is strongly recurrently connected with complex wiring rules. This circuitry can now be probed by studying responses to optogenetic perturbations of one or small numbers of cells. However, we currently lack a general theory connecting these responses to underlying circuitry. Here we develop a novel, exactly solvable theory to determine responses to small perturbations from the underlying connectivity. Analysis of these equations reveals simple rules that govern perturbation response patterns. Comparison with experimental data yields new constraints on the connectivity parameters. The theory yields predictions for the responses of unmeasured cell types and in new experimental conditions.

neuroscience↗

bioRxiv2024

Adaptation shapes the representational geometry in mouse V1 to efficiently encode the environment

Sensory adaptation dynamically changes neural responses as a function of previous stimuli, profoundly impacting perception. The response changes induced by adaptation have been characterized in detail in individual neurons and at the population level after averaging across trials. However, it is not clear how adaptation modifies the aspects of the representations that relate more directly to the ability to perceive stimuli, such as their geometry and the noise structure in individual trials. To address this question, we recorded from a population of neurons in the mouse visual cortex and presented one stimulus (an oriented grating) more frequently than the others. We then analyzed these data in terms of representational geometry and studied the ability of a linear decoder to discriminate between similar visual stimuli based on the single-trial population responses. Surprisingly, the discriminability of stimuli near the adaptor increased, even though the responses of individual neurons to these stimuli decreased. This improvement arose mainly from how adaptation reshaped the geometry of the mean population responses, with changes in the noise structure playing a minor role. Locomotion modulated the population responses along a direction largely independent of adaptation, allowing the adaptation-induced increase in discriminability to persist during locomotion. Finally, artificial neural networks trained to reconstruct the visual stimulus under metabolic constraints reproduced the increase in discriminability and the decrease in responses. The paradoxical effects of adaptation are thus consistent with the efficient coding framework, allowing the brain to improve the representation of frequent stimuli while limiting the associated metabolic cost.

neuroscience↗

bioRxiv2024

ZNFX1 is a Novel Master Regulator in Epigenetically-induced Pathogen Mimicry and Inflammasome Signaling in Cancer

DNA methyltransferase and poly(ADP-ribose) polymerase inhibitors (DNMTis, PARPis) induce a stimulator of interferon (IFN) genes (STING)-dependent pathogen mimicry response (PMR) in ovarian (OC) and other cancers. We now show that combining DNMTis and PARPis upregulates expression of a little-studied nucleic-acid sensor, NFX1-type zinc finger-containing 1 protein (ZNFX1). We demonstrate that ZNFX1 is a novel master regulator for PMR induction in mitochondria, serving as a gateway for STING-dependent PMR. In patient OC databases, high ZNFX1 expression levels correlate with advanced stage disease. ZNFX1 expression alone significantly correlates with an increase in overall survival in a phase 3 trial for therapy-resistant OC patients receiving bevacizumab in combination with chemotherapy. In correlative RNA-seq data, inflammasome signaling through ZNFX1 correlates with abnormal vasculogenesis. ZNFX1 controls PMR signaling through the mitochondria and may serve as a biomarker to facilitate offering personalized therapy in OC patients, highlighting the strong translational significance of our findings. Significance statementDNA methyltransferase and poly(ADP-ribose) polymerase inhibitors upregulate expression of a novel nucleic-acid sensor, ZNFX1 that serves as a mitochondrial gateway to STING-dependent interferon/inflammasome signaling with tumor suppressor properties in ovarian cancer.

cancer biology↗

bioRxiv2023

Circuit-motivated generalized affine models characterize stimulus-dependent visual cortical shared variability

Correlated variability in the visual cortex is modulated by stimulus properties. The stimulus dependence of correlated variability impacts stimulus coding and is indicative of circuit structure. An affine model combining a factor proportional to mean stimulus response and an additive offset has been proposed to explain how correlated variability in primary visual cortex (V1) depends on stimulus orientations. However, whether the affine model could be extended to explain modulations by other stimulus variables or variability shared between two brain areas is unknown. Motivated by a simple neural circuit mechanism, we modified the affine model to better explain the contrast-dependence of neural variability shared within either primary or secondary visual cortex (V1 or V2) as well as the orientation-dependence of neural variability shared between V1 and V2. Our results bridge neural circuit mechanisms and statistical models, and provide a parsimonious explanation for the stimulus-dependence of correlated variability within and between visual areas.

neuroscience↗

bioRxiv2023

The stabilized supralinear network accounts for the contrast dependence of visual cortical gamma oscillations

When stimulated, neural populations in the visual cortex exhibit fast rhythmic activity with frequencies in the gamma band (30-80 Hz). The gamma rhythm manifests as a broad resonance peak in the powerspectrum of recorded local field potentials, which exhibits various stimulus dependencies. In particular, in macaque primary visual cortex (V1), the gamma peak frequency increases with increasing stimulus contrast. Moreover, this contrast dependence is local: when contrast varies smoothly over visual space, the gamma peak frequency in each cortical column is controlled by the local contrast in that columns receptive field. No parsimonious mechanistic explanation for these contrast dependencies of V1 gamma oscillations has been proposed. The stabilized supralinear network (SSN) is a mechanistic model of cortical circuits that has accounted for a range of visual cortical response nonlinearities and contextual modulations, as well as their contrast dependence. Here, we begin by showing that a reduced SSN model without retinotopy robustly captures the contrast dependence of gamma peak frequency, and provides a mechanistic explanation for this effect based on the observed non-saturating and supralinear input-output function of V1 neurons. Given this result, the local dependence on contrast can trivially be captured in a retinotopic SSN which however lacks horizontal synaptic connections between its cortical columns. However, long-range horizontal connections in V1 are in fact strong, and underlie contextual modulation effects such as surround suppression. We thus explored whether a retinotopically organized SSN model of V1 with strong excitatory horizontal connections can exhibit both surround suppression and the local contrast dependence of gamma peak frequency. We found that retinotopic SSNs can account for both effects, but only when the horizontal excitatory projections are composed of two components with different patterns of spatial fall-off with distance: a short-range component that only targets the source column, combined with a long-range component that targets columns neighboring the source column. We thus make a specific qualitative prediction for the spatial structure of horizontal connections in macaque V1, consistent with the columnar structure of cortex.

neuroscience↗

bioRxiv2022

Mechanisms underlying reshuffling of visual responses by optogenetic stimulation in mice and monkeys

The ability to optogenetically perturb neural circuits opens an unprecedented window into mechanisms governing circuit function. We analyzed and theoretically modeled neuronal responses to visual and optogenetic inputs in mouse and monkey V1. In both species, optogenetic stimulation of excitatory neurons strongly modulated the activity of single neurons, yet had weak or no effects on the distribution of firing rates across the population. Thus, the optogenetic inputs reshuffled firing rates across the network. Key statistics of mouse and monkey responses lay on a continuum, with mice/monkeys occupying the low/high rate regions, respectively. We show that neuronal reshuffling emerges generically in randomly connected excitatory/inhibitory networks, provided the coupling strength (combination of recurrent coupling and external input) is sufficient that powerful inhibitory feedback cancels the mean optogenetic input. A more realistic model, distinguishing tuned visual vs. untuned optogenetic input in a structured network, reduces the coupling strength needed to explain reshuffling.

neuroscience↗

bioRxiv2022

Targeted cortical stimulation reveals principles of cortical contextual interactions

Cross-orientation suppression is a classic form of contextual normalization in visual cortex. There is substantial evidence that feed-forward processes contribute to cross-orientation suppression, but binocular studies suggest there may also be a contribution from cortical circuits. We used a combined imaging/optogenetic stimulation system to provide patterned optogenetic stimulation to cells in orientation columns that either matched or were orthogonal to the preferred visual orientation of neurons recorded with electrodes. When visual or optogenetic stimulation of columns preferring one orientation was combined with optogenetic stimulation of columns preferring the orthogonal orientation, we observed less suppression than when orthogonal stimulation was provided visually, suggesting that cortical circuits do not provide a large fraction of visual cross-orientation suppression. Integration of visual and optogenetic signals was linear when neurons exhibited low firing rates and became sublinear when neurons exhibited higher firing rates. We probed the nature of sublinearities in cortex by examining the influence of optogenetic stimulation of cortical interneurons. We observed a range of responses, including evidence for paradoxical responses in which interneuron stimulation caused a decrease in inhibitory firing rate, presumably due to the withdrawal of recurrent excitation. These results are compatible with cortical circuits that exhibit strong recurrent excitation with stabilizing inhibition that provides normalization, albeit normalization that is too weak across columns to account for cross-orientation suppression.

neuroscience↗

bioRxiv2022

Unifying model for three forms of contextual modulation including feedback input from higher visual areas

Sensory systems must combine local features with context to infer meaning. Accordingly, context profoundly influences neural responses. We developed a unified circuit model demonstrating how feedforward and feedback inputs are integrated to produce three forms of contextual effects in mouse primary visual cortex (V1). First, reanalyzing existing data, we discovered that increasing stimulus size only weakly increases the area of V1 neural response, conflicting with previous models of surround suppression (SS). Second, through modeling, we found that, in Layer 2/3, (1) SS and its contrast dependence are largely inherited from Layer 4; (2) Inverse responses (IR) - size-tuned responses to a gray "hole" in a full-field grating - are driven by feedback connections provided they are sufficiently wide; (3) Cross-orientation surround facilitation is induced by the summation of feedback input driving IR with the feedforward-driven classical center response. The model accounts for many previous findings and makes multiple testable predictions. HighlightsO_LIOne model explains three different types of contextual modulation C_LIO_LIThe widths of spatial response patterns grow much more slowly than stimulus size. C_LIO_LIInverse responses depend on the geometry of feedback response fields and projections C_LIO_LISummation of classical and inverse response accounts for surround facilitation. C_LI

neuroscience↗

bioRxiv2021

Learning of biased representations in LIP through interactions between recurrent connectivity and Hebbian plasticity

When monkeys learn to group visual stimuli into arbitrary categories, lateral intraparietal area (LIP) neurons become category-selective. Surprisingly, the representations of learned categories are overwhelmingly biased: nearly all LIP neurons in a given animal prefer the same category over other behaviorally equivalent categories. We propose a model where such biased representations develop through the interplay between Hebbian plasticity and the recurrent connectivity of LIP. In this model, two separable processes of positive feedback unfold in parallel: in one, category selectivity emerges from competition between prefrontal inputs; in the other, bias develops due to lateral interactions among LIP neurons. This model reproduces the levels of category selectivity and bias observed under a variety of conditions, as well as the redevelopment of bias after monkeys learn redefined categories. It predicts that LIP receptive fields would spatially cluster by preferred category, which we experimentally confirm. In summary, our model reveals a mechanism by which LIP learns abstract representations and assigns meaning to sensory inputs.

neuroscience↗

bioRxiv2021

Antagonistic inhibitory subnetworks control cooperation and competition across cortical space

The cortical microcircuit can dynamically adjust to dramatic changes in the strength, scale, and complexity of its input. In the primary visual cortex (V1), pyramidal cells (PCs) integrate widely across space when signals are weak, but narrowly when signals are strong, a phenomenon known as contrast-dependent surround suppression. Theoretical work has proposed that local interneurons could mediate a shift from cooperation to competition of PCs across cortical space, underlying this computation. We combined calcium imaging and electrophysiology to constrain a stabilized supralinear network model that explains how the four principal cell types in layer 2/3 (L2/3) of mouse V1- somatostatin (SST), parvalbumin (PV), and vasoactive intestinal peptide (VIP) interneurons, and PCs- transform inputs from layer 4 (L4) PCs to encode drifting gratings of varying size and contrast. Using bidirectional optogenetic perturbations, we confirmed key predictions of the model. Our data and modeling showed that recurrent amplification drives a transition from a positive PC[->]VIP{dashv}SST{dashv}PC feedback loop at small size and low contrast to a negative PC[->]SST{dashv}PC feedback loop at large size and high contrast to contribute to this flexible computation. This may represent a widespread mechanism for gating competition across cortical space to optimally meet task demands.

neuroscience↗