PLOS Computational Biology

Markov models are widely used to describe the gating kinetics of voltage-gated ion channels, but increasing model complexity can introduce parameters that are difficult to constrain from macroscopic measurements. In this study, I used global variance-based Sobol sensitivity analysis to examine how model topology, stimulation protocol, and parameter uncertainty shape the accessibility of kinetic parameters in voltage-gated ion channel models. I analyzed progressively more complex Markov schemes, beginning with a two-state closed-open model, extending to a three-state linear closed-closed-open model, a cyclic three-state model with a direct closed-open transition, and a four-state closed-closed-open-inactivated model. Sensitivity was quantified for the open probability (P) under voltage-step and sinusoidal protocols. In linear models, parameter influence was concentrated in transitions directly connected to the open state, whereas distal closed-closed transitions contributed little to output variance. This weak influence was not rescued by higher-order interaction effects. Multi-frequency sinusoidal stimulation (20-200 Hz) preserved the same sensitivity hierarchy observed under step protocols, indicating that dynamic stimulation does not overcome the structural limitations of serial topologies. In contrast, introducing a cyclic pathway fundamentally redistributed sensitivity, showing that distal-parameter weakness is a consequence of serial arrangement rather than a universal property of Markov gating. Adding inactivation shifted dominant variance control to the open-inactivated transition during sustained depolarization, yet distal closed-closed transitions remained weak. Finally, constraining variability at the dominant activation edge shifted variance upstream, demonstrating that low Sobol indices reflect the relative flexibility of competing bottlenecks within a given ensemble rather than the intrinsic irrelevance of a transition. Together, these results define topology-dependent limits on parameter accessibility in Markov models of voltage-gated ion channels and provide practical guidance for selecting model complexity, designing informative protocols, and favoring cyclic over extended linear topologies when constructing robust channel models.

A typical neuron receives thousands of inputs and is able to adapt the strength of its synapses to store new information and meet ongoing computational demands. The synaptic response to plasticity induction is stochastic and spatially structured but is traditionally described by deterministic models representing the "average" dynamics. Growing experimental evidence indicates that not only the stimulation protocol determines the plasticity outcome but that the initial synaptic sizes, their fluctuations, and the spatial competition for the plasticity-relevant proteins play a decisive role. This probabilistic perspective makes it hard to predict the fate of a given synapse and requires a conceptual shift from a single synapse view to a probabilistic multi-spine competitive process where the plasticity needs and the available resources are considered together. Here, we propose a data-driven modeling framework able to predict collective plasticity outcomes along a dendrite based on the initial size, the number, and the spatial distance between simultaneously stimulated synapses. Our data analysis reveals a log-normal distribution of protein numbers for many plasticity-mediating proteins and shows that this log-normal protein allocation constrains and controls the collective plasticity outcome across multiple stimulated and non-stimulated synapses while preserving a global size distribution. Our findings highlight how local stochastic processes and global protein allocation rules give rise to synaptic plasticity outcomes, offering a new framework to understand and predict dendritic computation.

Compartmental infectious disease models are used to calculate disease transmission, estimate underlying rates, forecast future burden, and compare benefits across intervention scenarios. These models aggregate individuals into compartments, often stratified by characteristics to represent groups that might be intervention targets or otherwise of particular concern. Ideally, model calculation could occur at the most demanding resolution for the overall analysis, but this may be infeasible due to availability of computational resources or empirical data. Instead, detailed population age-structure might be consolidated into broad categories such as children, working-age adults, and seniors. Researchers must then discretise key epidemic parameters, like the infection-fatality ratio, for these lower resolution groups. After estimating outcomes for those crude groups, follow on analyses, such as calculating years of life lost (YLLs), may need to distribute or weight those low-resolution outcomes back to the high resolution. The specific calculation for these aggregation and disaggregation steps can substantially influence outcomes. To assist researchers with these tasks, we developed paramix, an R package which simplifies the transformations between high and low resolution. We demonstrate applying paramix to a common discretisation analysis: using age structured models for health economic calculations comparing YLLs. We compare how estimates vary between paramix and several alternatives for an archetypal model, including comparison to a high resolution benchmark. We consistently found that paramix yielded the most similar estimates to the high-resolution model, for the same computational burden of low-resolution models. In our illustrative analysis, the non-paramix methods estimated up to twice as many YLLs averted as the paramix approach, which would likely lead to a similarly large impact on incremental cost-effectiveness ratios used in economic evaluations. Author summaryResearchers use infectious disease models to understand trends in disease spread, including predicting future infections under different interventions. Constraints like data availability and numerical complexity drive researchers to group individuals into broad categories; for example, all working age adults might be represented as a single set of model compartments. Key epidemic parameters can vary widely across such groups. Additionally, model outcomes calculated using these broad categories often need to be disaggregated to a high resolution, for example a precise age at death for calculating years life lost, a key measure when estimating the cost-effectiveness of interventions. To satisfy these needs, we present a software package, paramix, which provides tools to move between high and low resolution data. In this paper, we demonstrate the capabilities of paramix by comparing various methods of calculating deaths and years of life lost across broad age groups. For an analysis of an archetypal model, we find paramix best matches a high-resolution model, while the alternatives are substantially different.

Constitutively secreted by endothelial cells, soluble FLT1 (sFLT1 or sVEGFR1) binds and sequesters extracellular vascular endothelial growth factors (VEGF), thereby reducing VEGF binding to VEGF receptor tyrosine kinases and their downstream signaling. In doing so, sFLT1 plays an important role in vascular development and in the patterning of new blood vessels in angiogenesis. Here, we develop multiple mechanistic models of sFLT1 secretion and identify a minimal mechanistic model that recapitulates key qualitative and quantitative features of temporal experimental datasets of sFLT1 secretion from multiple studies. We show that the experimental data on sFLT1 secretion is best represented by a delay differential equation (DDE) system including a maturation term, reflecting the time required between synthesis and secretion. Using optimization to identify appropriate values for the key mechanistic parameters in the model, we show that two model parameters (extracellular degradation rate constant and maturation time) are very strongly constrained by the experimental data, and that the remaining parameters are related by two strongly constrained constants. Thus, only one degree of freedom remains, and measurements of the intracellular levels of sFLT1 would fix the remaining parameters. Comparison between simulation predictions and additional experimental data of the outcomes of chemical inhibitors and genetic perturbations suggest that intermediate values of the secretion rate constant best match the simulation with experiments, which would completely constrain the model. However, some of the inhibitors tested produce results that cannot be reproduced by the model simulations, suggesting that additional mechanisms not included here are required to explain those inhibitors. Overall, the model reproduces most available experimental data and suggests targets for further quantitative investigation of the sFLT1 system. Author SummaryProteins that are typically found outside cells are initially made inside cells, and later secreted into extracellular space. Many of these secreted proteins have important functions outside the cell that are well-studied; however, usually much less is known about the pre-secretion life of these molecules. Many computational models only represent the extracellular versions of secreted proteins, reducing all production and secretion steps into a single modeled process. Here, we develop a mechanistic model of the production and secretion of a specific secreted protein, sFLT1, which inhibits blood vessel growth by acting as an extracellular sponge for another set of secreted proteins, the vascular endothelial growth factors. We compare several models to existing experimentally-measured sFLT1 data, and we show that the data are most simply explained by including a delay between intracellular sFLT1 production and sFLT1 transport or degradation. This is consistent with the biology of the cells secretory pathway, where immature proteins are gradually processed into mature forms over minutes to hours. Our approach could be incorporated into improved models for any pathway involving secreted proteins, including sFLT1-regulated models of blood vessel biology.

Retinal ganglion cells, the output neurons of the vertebrate retina, often display nonlinear summation of visual signals over their receptive fields. This creates sensitivity to spatial contrast, letting the cells respond to spatially structured visual stimuli even when no net change in overall illumination of the receptive field occurs. Yet, computational models of ganglion cell responses are often based on linear receptive fields, and typical nonlinear extensions, which separate receptive fields into nonlinearly combined subunits, are often cumbersome to fit to experimental data. Previous work has suggested to model spatial-contrast sensitivity in responses to flashed images by combining signals from the mean and variance of light intensity inside the receptive field. Here, we extend and adjust this spatial contrast model for application to spatiotemporal stimulation and explore its performance on spiking responses that we recorded from ganglion cells of marmosets under artificial and naturalistic movies. We show how the model can be fitted to experimental data and that it outperforms common models with linear spatial integration to different degrees for different types of ganglion cells. Finally, we use the model framework to infer the cells spatial scale of nonlinear spatial integration. Our work shows that the spatial contrast model can capture aspects of nonlinear spatial integration in the primate retina with only few free parameters. The model can be used to assess the cells functional properties under natural stimulation and provides a simple-to-obtain benchmark for comparison with more detailed nonlinear encoding models. Author SummaryOur visual experience depends on the retinas remarkable ability to detect light patterns and contrast in the world around us. Retinal ganglion cells, the output neurons of the retina, modulate their activity based on signals within small, specific regions of the visual scene, called their receptive fields. But many cells do not only encode overall brightness, summed linearly across the receptive field, but are also sensitive to local spatial contrast, that is, variations in brightness within the receptive field. Computational models that account for this nonlinear spatial integration exist, but require large amounts of data and are challenging to fit. We therefore developed the spatial contrast model, which takes a simple measure of light-intensity variations as an input, and tested it on measured responses of primate retinal ganglion cells to both artificial and naturalistic movies. The model substantially outperformed standard models with linear receptive fields, despite having only one additional tunable parameter. Furthermore, we used the model to investigate the spatial scale at which the cells integrate spatial contrast and found striking consistency across cell types. The spatial contrast model thus offers a practical tool for capturing retinal stimulus encoding and a simple-to-obtain benchmark for modeling nonlinear spatial integration.

Understanding how the brain integrates multisensory information during detection and decision-making remains an active area of research. While many inferences have been drawn about behavioural outcomes, key questions persist regarding both the nature of environmental cues and the internal mechanisms of integration. These complexities make multisensory integration particularly well suited to investigate through mathematical modelling. In this study, we present three models of audio-visual integration within a biologically motivated mean-field recurrent framework. These models extend a non-linear system of differential equations originally developed for unisensory decision-making. The OR and SUM models represent opposing ends of the integration spectrum: the former simulates independent unisensory processing using a winner-take-all (WTA) strategy, while the latter implements a linear summation model for full integration. A third model--the REPEAT Model--incorporates switch and repeat costs observed in multisensory tasks. We simulate 121 participants with varying unisensory evidence accumulation rates, capturing behavioural diversity from modality dominance to balanced integration. Model outputs (reaction time and accuracy) were compared with empirical results from audio-visual detection tasks. We further fit the outputs to a drift diffusion model, allowing comparison between simulated and theoretically optimal multisensory drift rates. The OR and SUM Models reproduced established unisensory response patterns. Drift diffusion analysis revealed suboptimal integration in the OR Model and optimal integration in the SUM Model. However, the SUM Model also produced supra-optimal responses under certain conditions, inconsistent with behavioural data. The REPEAT Model successfully captured the role of priming in sensory repetition effects, distinguishing it from true multisensory integration. Overall, these models highlight how biologically grounded mathematical frameworks can shed light on the mechanisms underlying multisensory integration, particularly the nuanced contributions of modality repetition and integration efficiency.

Building a simple model that precisely and functionally characterizes a neuron is a challenging and important task to select the best concise and computationally efficient model. However, this type of work has only been done for subthreshold properties of neurons. Here, we take a different perspective and suggest a method to obtain point-neuron models from morphologically-detailed models with dendrites. To do this, we focus on the functional characterization of the neuron response under in vivo conditions, and compute the transfer function of the detailed model. The parameters of this transfer function, in terms of mean voltage, voltage standard deviation and correlation time, can be used to compute the "best" point-neuron model that generates a transfer function very close to that of the morphologically-detailed model. We illustrate this approach for two very different neuronal morphologies, one from Drosophila larvae and one from mammals. In conclusion, this approach provides a tool to generate point-neuron models from detailed models, based on a functional characterization of the neuron response. Significance StatementThis study provides a new computational method to reduce morphological models into point-neuron models. To do so, we calculate the transfer function parameters, ie the voltage standard deviation, the mean voltage and the correlation time, of the morphological model and fit a point neuron-model onto this data. Here, we successfully apply this approach for two very different neuron morphologies, a drosophila neuron and a rat motoneuron.

Recent data on structure of dyads in cardiac myocytes indicate variable clustering of RyR calcium release channels. The question arises as to how geometric factors of RyR arrangement translate to their role in formation of calcium release events (CRE). Since this question is not experimentally testable in situ, we performed in silico experiments on a large set of calcium release site (CRS) models. The models covered the range of RyR spatial distributions observed in dyads, and included gating of RyRs with open probability dependent on Ca2+ and Mg2+ concentration. The RyR single-channel calcium current, varied in the range of previously reported values, was set constant in the course of CRE simulations. Other known features of dyads were omitted in the model formulation for clarity. CRE simulations initiated by a single random opening of one of the RyRs in a CRS produced spark-like responses with characteristics that varied with RyR vicinity, a newly defined parameter quantifying spatial distribution of RyRs in the CRSs, and with the RyR single-channel calcium current. The CRE characteristics followed the law of mass action with respect to a CRS state variable, defined as a weighed product of RyR vicinity and RyR single-channel calcium current. The results explained the structure-function relations among determinants of cardiac dyads on synergy principles and thus allowed to evolve the concept of CRS as a dynamic unit of cardiac dyad.

Effective public health strategies for preventing infectious diseases require a deep understanding of transmission dynamics. However, epidemiological data alone is often insufficient for guiding public health action, as transmission patterns emerge from complex human-environment interactions. Additionally, collecting empirical data is frequently limited by cost and logistical challenges, especially in low-resource settings. Consequently, health policy increasingly relies on inference models that account for spatial and temporal dimensions, confounding factors, and uncertainty. In this work, we introduce a novel hierarchical mechanistic Bayesian model based on ecological diffusion to improve epidemiological inference for spreading pathogens. This model is specifically designed to reveal underlying spatiotemporal dynamics from empirical data, explicitly accounting for sampling bias and uncertainty. To demonstrate its effectiveness, we apply the model to simulated data of a drug-resistant pathogen and formally compare its performance to state-of-the-art epidemiological regression models. Our findings show that the model accurately identifies transmission patterns and is particularly effective at pinpointing transmission hotspots and predicting spread pathways--critical capabilities for controlling drug-resistant pathogens. While reliable monitoring remains challenging in low-resource environments, this model addresses demographic sampling bias and spatiotemporal complexities, though concentrated sampling in presumed high-risk areas can sometimes distort transmission insights. The proposed framework can inform targeted public health interventions in the early stages of resistance emergence, offering valuable guidance for resource-constrained settings.

Predicting the species composition of microbial communities is a problem that has proven surprisingly difficult, including in fully controlled microbial systems in a lab setting. Few organisms are able to establish themselves in an community with no others, even if nutrients are provided. At the same time some species are observed to be unable to stably coexist with certain other species. We here present a model attempting to reduce the dynamics of community assembly to its most basic components, while preserving its salient features. The model deals not with abundances or densities of bacterial populations or resources, but only with their presence or absence. Similarly, only three types of discrete relationships between species are considered: species excluding each other, and mutualistic dependence on and production of nutrients in the form of exometabolites. Despite these simplifications, the model system still exhibits rich dynamics, including extinction cascades and emerging stability against invasion. We derive conditions for these to occur and compare our results with existing knowledge of microbial communities. Our results provide a novel approach for the theoretical study of microbial communities and their stability. SignificanceUnderstanding the processes governing the community composition of microbial ecosystems is an as-yet unsolved problem. In this article, we propose a novel rule-based model for the assembly and stability of communities of microbial species. Despite its simplicity the model exhibits rich dynamics, including critical points and extinction cascades. It presupposes no knowledge of specific population parameters or exact resource or species abundances, but focuses on pairwise interactions between species and their metabolites. The framework may be useful for understanding how real microbial communities arise and what causes them to break down.

The retinas phenomenological function is often considered to be well-understood: individual retinal ganglion cells are sensitive to a projection of the light stimulus movie onto a classical center-surround linear filter. Recent models elaborating on this basic framework by adding a second linear filter or spike histories, have been quite successful at predicting ganglion cell spikes for spatially uniform random stimuli, and for random stimuli varying spatially with low resolution. Fitting models for stimuli with more finely grained spatial variations becomes difficult because of the very high dimensionality of such stimuli. We present a method of reducing the dimensionality of a fine one dimensional random stimulus by using wavelets, allowing for several clean predictive linear filters to be found for each cell. For salamander retinal ganglion cells, we find in addition to the spike triggered average, 3 identifiable types of linear filters which modulate the firing of most cells. While some cells can be modeled fairly accurately, many cells are poorly captured, even with as many as 4 filters. The new linear filters we find shed some light on the nonlinearities in the retinas integration of temporal and fine spatial information.

The quantitative determination of the effects of antimicrobials is essential for our understanding of pharmacodynamics and for their rational clinical application. Common pharmacodynamic measures of antimicrobial efficacy, such as the MIC and the pharmacodynamic function, fail to capture the observed dependence of efficacy on the bacterial population size -- a phenomenon called inoculum effect. We assessed the relationship between bacterial inoculum size and pharmacodynamic relationship and determined the consequences of the inoculum effect on bacterial population dynamics with a mathematical multi-hit model that explicitly describes the interaction between antimicrobial molecules with their targets on the bacterial cells. Our model showed that the inoculum effect can arise from the binding dynamics of antimicrobial molecules to bacterial targets alone. A pharmacodynamic function extended by the inoculum effect on its parameters was able to predict the long-term population dynamics of simple scenrios well. More complex scenarios, however, were only captured with by the mechanistically more explicit multi-hit model. In simulations with competing antimicrobial-susceptible and -resistant bacteria, neglecting the inoculum effect led to an overestimation of the competitive ability of the resistant strain. Our work underpins the importance of including the inoculum effect into quantitative pharmacodynamic frameworks, and provides approaches to accomplish that.

Microbial metabolic interactions impact ecosystems, human health and biotechnological processes profoundly. However, their determination remains elusive, invoking an urgent need for predictive models that seamlessly integrate metabolic details with ecological and evolutionary principles which shape the interactions within microbial communities. Inspired by the evolutionary game theory, we formulated a bi-level optimization framework termed NECom for the prediction of Nash equilibria of microbial community metabolic models with significantly enhanced accuracy. NECom is free of a long hidden forced altruism setup in previous static algorithm while allowing for sensing and responding between microbial members that is missing in dynamic methods. We successfully predicted several classical games in the context of metabolic interactions that were falsely or incompletely predicted by existing methods, including prisoners dilemma, snowdrift game and mutualism. The results provided insights into why mutualism is favorable despite seemingly costly cross-feeding metabolites, and demonstrated the potential to predict heterogeneous phenotypes among the same species. NECom was then applied to a reported algae-yeast co-culture system that shares typical cross-feeding features of lichen, a model system of mutualism. More than 1200 growth conditions were simulated, of which 488 conditions correspond to 3221 experimental data points. Without fitting any ad-hoc parameters, an overall 63.5% and 81.7% reduction in root-mean-square error in predicted growth rates for the two species respectively was achieved when compared with the standard flux balance analysis. The simulation results further show that growth-limiting crossfeeding metabolites can be pinpointed by shadow price analysis to explain the predicted frequency-dependent growth pattern, offering insights into how stabilizing microbial interactions control microbial populations.

Sensory input is often ambiguous, leading to uncertain interpretations of the external world. Estimates of perceptual uncertainty might be useful in guiding behavior, but it remains unclear whether humans explicitly represent uncertainty in naturalistic settings, and whether any such representations are normatively correct. Progress has been hindered by the absence of stimulus-computable models that estimate uncertainty. We developed a class of task-optimized models that generate probability distributions over perceptual estimates. To assess whether human uncertainty representations align with the models, we compared human confidence judgments, which might indirectly reflect uncertainty representations, to confidence judgments extracted from the models uncertainty. In both sound localization and pitch perception, human confidence varied systematically, being lower for stimuli that produced more variable estimates across trials. Human confidence tracked model confidence across conditions, suggesting that human uncertainty representations accurately reflect the actual uncertainty of perceptual estimation. The modeling framework is extensible to other perceptual domains.

GABAB receptors (GABABRs) are an important building block in neural activity. Despite their widely hypothesized role in many basic neuronal functions and mental disorder symptomatology, there is a lack of biophysically and biochemically detailed models of these receptors and the way they mediate neuronal inhibition. Here, we developed a computational model for the activation of GABABRs and its effects on the activation of G protein-coupled inwardly rectifying potassium (GIRK) channels as well as inhibition of voltage-gated Ca2+ channels. To ensure the generality of our modelling framework, we fit our model to electrophysiological data including patch-clamp and intracellular recordings that described both pre- and postsynaptic effects of the receptor activation. We validated our model using data on postsynaptic effects of GABABRs on layer V pyramidal cell firing activity ex vivo and in vivo and confirmed the strong impact of dendritic GIRK channel activation on the neuron output. Finally, we reproduced and dissected the effects of a knockout of RGS7 (a G protein signalling protein) on CA1 pyramidal cell electrophysiological properties, which shows the potential of our model in generating insights on genetic manipulations of the GABABR system and related genetic variants. Our model thus provides a flexible tool for biochemically and biophysically detailed simulations of different aspects of GABABR activation that can reveal both foundational principles of neuronal dynamics and brain disorder-associated traits and treatment options.

The cerebral microvasculature forms a dense network of interconnected blood vessels where flow is modulated partly by astrocytes. Increased neuronal activity stimulates astrocytes to release vasoactive substances at the endfeet, altering the diameters of connected vessels. Our study simulated the coupling between blood flow variations and vessel diameter changes driven by astrocytic activity in the rat somatosensory cortex. We developed a framework with three key components: coupling between vasculature and synthesized astrocytic morphologies, a fluid dynamics model to compute flow in each vascular segment, and a stochastic process replicating the effect of astrocytic endfeet on vessel radii. The model was validated against experimental flow values from literature across cortical depths. We found that local vasodilation from astrocyte activity increased blood flow, especially in capillaries, exhibiting a layer-specific response in deeper cortical layers. Additionally, the highest blood flow variability occurred in capillaries, emphasizing their role in cerebral perfusion regulation. We discovered that astrocytic activity impacts blood flow dynamics in a localized, clustered manner, with most vascular segments influenced by two to three neighboring endfeet. These insights enhance our understanding of neurovascular coupling and guide future research on blood flow-related diseases.

Learned information and experiences are thought to be stored in synapses, composed of building block molecules whose number typically correlates with synaptic strength. Activity-dependent plasticity mechanisms, such as Hebbian learning, regulate these building blocks, promoting synaptic growth to encode acquired knowledge. However, this process can destabilize cortical networks through overexcitation, leading to runaway dynamics. To prevent such instabilities the brain uses compensatory mechanisms like synaptic scaling. Existing models rely on rapid timescales, contradicting experimental observations that synaptic scaling occurs slowly. Here, we introduce aggregate scaling, a simple framework to study synapse-mediated homeostasis based on the availability and competitive redistribution of synaptic building blocks. Our model enforces stability by integrating rapid regulation of the total synaptic strength and firing rate homeostasis over much slower, realistic timescales. It preserves key neuronal properties, such as firing activity around a homeostatic set-point, long-tailed distributions of synaptic weights, and response to brief stimulation. Author summaryLearning and memory rely on changes in the strength of synapses, connections between neurons. When these changes go unchecked, they can lead to abnormal brain activity. Homeostatic mechanisms such as synaptic scaling seem to contribute to the brains solution to this problem. However, most existing models of synaptic scaling require this process to be much faster than what is observed in real neurons, raising doubts about their biological accuracy. In our study, we present aggregate scaling, an alternative framework where synapses share and compete for limited molecular resources needed to maintain their strength. This competition is paired with a homeostatic regulation of the abundance of synaptic rescources. We show that this allows neurons and networks to remain stable while still supporting learning. Unlike previous models, our approach operates on biologically realistic timescales and reproduces key experimental findings, including stable neural activity and natural patterns of synaptic strength. Overall our model provides insights into how the brain maintains balance during learning, emphasizing the importance of synaptic resource logistics.

Spontaneous network bursts, the intervals of intense network-wide activity interleaved with longer periods of sparse activity, are a hallmark phenomenon observed in cortical networks at postnatal developmental stages. Generation, propagation and termination of network bursts depend on a combination of synaptic, cellular and network mechanisms; however, the interplay between these mechanisms is not fully understood. We study this interplay in silico, using a new data-driven framework for generating spiking neuronal networks fitted to the microelectrode array recordings. We recorded the network bursting activity from rat postnatal cortical networks under several pharmacological conditions. In each condition, the function of specific excitatory and inhibitory synaptic receptors was reduced in order to examine their impact on global network dynamics. The obtained data was used to develop two complementary model fitting protocols for automatic model generation. These protocols allowed us to disentangle systematically the modeled cellular and synaptic mechanisms that affect the observed network bursts. We confirmed that the change in excitatory and inhibitory synaptic transmission in silico, consistent with pharmacological conditions, can account for the changes in network bursts relative to the control data. Reproducing the exact recorded network bursts statistics required adapting both the synaptic transmission and the cellular excitability separately for each pharmacological condition. Our results bring new understanding of the complex interplay between cellular, synaptic and network mechanisms supporting the burst dynamics. While here we focused on analysis of in vitro data, our approach can be applied ex vivo and in vivo given that the appropriate experimental data is available. New & NoteworthyWe studied the role of synaptic mechanisms in shaping the neural population activity by proposing a new method to combine experimental data and data-driven computational modeling based on spiking neuronal networks. We analyze a dataset recorded from postnatal rat cortical cultures in vitro under the pharmacological influence of excitatory and inhibitory synaptic receptor antagonists. Our computational model identifies neurobiological mechanisms necessary to reproduce the changes in population activity seen across pharmacological conditions.

Orientation hypercolumns in the visual cortex are delimited by the repeating pinwheel patterns of orientation selective neurons. We design a generative model for visual cortex maps that reproduces such orientation hypercolumns as well as ocular dominance maps while preserving retinotopy. The model uses a neural placement method based on t-distributed stochastic neighbour embedding (t-SNE) to create maps that order common features in the connectivity matrix of the circuit. We find that, in our model, hypercolumns generally appear with fixed cell numbers independently of the overall network size. These results would suggest that existing differences in absolute pinwheel densities are a consequence of variations in neuronal density. Indeed, available measurements in the visual cortex indicate that pinwheels consist of a constant number of [~]30, 000 neurons. Our model is able to reproduce a large number of characteristic properties known for visual cortex maps. We provide the corresponding software in our MAPStoolbox for Matlab. In briefWe present a generative model that predicts visual map structures in the brain and a large number of their characteristic properties; a neural placement method for any given connectivity matrix. HighlightsO_LIGenerative model with retinotopy, orientation preference and ocular dominance. C_LIO_LIPrediction of constant neuronal numbers per orientation hypercolumn. C_LIO_LICurated data shows constant [~]30, 000 neurons per pinwheel across species. C_LIO_LISimple explanation for constant pinwheel and orientation hypercolumn ratios. C_LIO_LIPrecise prediction of [~]80% nearest neighbour singularities with opposing polarity. C_LIO_LIModel asymptotically approaches realistic normalised pinwheel densities. C_LIO_LISmall brains with < [~]300 potential pinwheels exhibit salt-and-pepper maps. C_LIO_LIDifferent map phenotypes can exist even for similar connectivity. C_LI

Multistable perception - spontaneous switches of perception when viewing a stimulus compatible with several distinct interpretations - is often characterized by the distribution of durations of individual dominance phases. For continuous viewing conditions, these distributions look remarkably similar for various multistable displays and are typically described using Gamma distribution. Moreover, durations of individual dominance phases show a subtle but consistent dependence on prior perceptual experience with longer dominance phases tending to increase the duration of the following ones, whereas the shorter dominance leads to similarly shorter durations. One way to generate similar switching behavior in a model is by using a combination of cross-inhibition, self-adaptation, and neural noise with multiple useful models being built on this principle. Here, we take a closer look at the history-dependent changes in the distribution of durations of dominance phases. Specifically, we used Gamma distribution and allowed both its parameters - shape and scale - to be linearly dependent on the prior perceptual experience at two timescales. We fit a hierarchical Bayesian model to five datasets that included binocular rivalry, Necker cube, and kinetic-depth effects displays, as well as data on binocular rivalry in children and on binocular rivalry with modulated contrast. For all datasets, we found a consistent change of the distribution shape with higher levels of perceptual history, which can be viewed as a proxy for perceptual adaptation, leading to a more normal-like shape of the Gamma distribution. When comparing real observers to matched simulated dominance phases generated by a spiking neural model of bistability, we found that although it matched the positive history-dependent shift in the shape parameter, it also predicted a negative change of scale parameter that did not match empirical data. We argue that our novel analysis method, the implementation is available freely at the online repository, provides additional constraints for computational models of multistability.