Tools and Resources

Neuroscience

Modeling and simulation of neocortical micro- and mesocircuitry (Part II, Physiology and experimentation)

Blue Brain Project, École polytechnique fédérale de Lausanne (EPFL), Campus Biotech, Switzerland
Department of Basic Neurosciences, University of Geneva, Switzerland
CHU Sainte-Justine Research Center, Canada
Department of Neurosciences, Faculty of Medicine, Université de Montréal, Canada
Neural Circuits Laboratory, Newcastle University, United Kingdom

Jan 20, 2026

https://doi.org/10.7554/eLife.99693.3

Open access
Copyright information

Figures
Tables
Additional files

8 figures, 2 tables and 9 additional files

Figures

Figure 1

Download asset Open asset

Overview of the physiology and simulation workflow.

1. Anatomical model: summary of the anatomical nbS1 model described in the companion paper. 2. Neuron physiology: neurons were modeled as multicompartment models with ion channel densities optimized using previously established methods and data from somatic and dendritic recordings of membrane potentials in vitro. 3. Synaptic physiology: Models of synapses were built using previously established methods and data from paired recordings in vitro. 4. Compensation for missing synapses: Excitatory synapses originating from outside nbS1 were compensated with noisy somatic conductance injection, parameterized by a novel algorithm. 5. In vivo-like activity: We calibrated an in silico activity regime compatible with in vivo spontaneous and stimulus-evoked activity. 6. In silico experimentation: Five laboratory experiments were recreated. Two were used for calibration and three of them were extended beyond their original scope. 7. Open Source: simulation software and a seven-column subvolume of the model are available on Zenodo (see ‘Data availability statement’). Data generalizations: three data generalisation strategies were employed to obtain the required data. Left: mouse to rat; middle: adult to juvenile (P14) rat; right: hindlimb (S1HL) and barrel field (S1BF) subregions to the whole nbS1. Throughout the figure, the corresponding purple icons show where these strategies were used.

Figure 2

Download asset Open asset

Modeling and validation of neuron physiology.

(A1) Example excitatory neuron morphologies. See companion paper for exemplar morphologies of all morphological types. (A2) Example morphologies placed within the atlas-based volume. Axons shown in gray. (B) Optimized conductance densities for two exemplary e-types. (B1) cADpyr e-type on L5_TPC:A m-type; (B2) bSTUT and cSTUT e-types on L5_NBC m-type. (The L5_NBC m-type is combined with more e-types than the two shown, see panel D.) Morphologies were visualized with NeuroMorphoVis (Abdellah et al., 2018). Neurite diameters are enlarged (×3) for visibility. Soma and dendrites in black, axon in red. (C) e-types used in the model. As used in Markram et al., 2015 and similar to the Petilla terminology Ascoli et al., 2008. (D) me-type composition. Heatmap shows the proportion of e-types for each m-type. Each me-type is assigned to one of the three I subpopulations. Assignments depending not only on m-type are highlighted by a green box. Particularly, for BTC and DBC m-types, the cACint e-type belongs to the Sst +subpopulation. (E) Validation of dendritic physiology of all L5 TTPCs. Panel reproduced from Reva et al., 2023. (E1) Validation of back-propagating action potential (bAP) amplitude (i.e., the dependence of bAP amplitude on distance from the soma) for basal (green) and apical (blue) dendrites. Reference data (in red) comes from Stuart and Sakmann, 1994; Larkum et al., 2001 (apical) and Nevian et al., 2007 (basal). Lines show exponential fits for the in silico (green and blue) and in vitro (red) data. Color bar indicates dendritic diameter. (E2) Validation of EPSP attenuation. Reference data (in red) comes from Berger et al., 2001 (apical) and Nevian et al., 2007 (basal). Lines and colorbar same as in (E1).

Figure 3

Download asset Open asset

Modeling and validation of synaptic physiology.

(A1) Exemplary pair of L5 TTPCs (visualized with NeuroMorphoVis; Abdellah et al., 2018). Presynaptic cell in gray, postsynaptic cell in red, synapses between them in purple. Neurite diameters are enlarged (x3) for visibility and axons were cut to fit into the figure. Pre- and postsynaptic voltage traces on the top right. (A2) Exemplary postsynaptic traces with different STP profiles. (A3) Assignment of STP profiles to viable pathways. (Pathways were considered viable if there were at least 10 connections in all eight subregions of the model.) (B1) Validation of first PSP amplitudes (see also table in Supplementary file 5). Dashed gray line represents perfect correlation between experimental and model values. Error bars show 1 standard deviation (also for C1, **D, F**). (B2) Predicted PSP amplitudes of all viable pathways in the circuit. Postsynaptic cells were held at –70 mV using an in silico voltage-clamp. Means were calculated over 100 pairs of neurons with 35 repetitions each. (**C1, C2**) same as (B1) and (B2), but showing the CV of the first PSP amplitude (corresponding table is in Supplementary file 6). (D) Validation of mPSC frequencies (see also table in Supplementary file 7). (E) Location of synapses from VPM fibers (purple) and POm fibers (red) on 38 neurons (dark gray) in a 5 µm radius column (visualized with BioExplorer). (F) Validation of thalamocortical EPSP amplitudes as in (B1). The four pathways used for the validation are marked with a black rectangle on H1 to its right. (G) EPSP latencies (time from presynaptic spike to the rise to 5% of peak EPSP amplitude in the postsynaptic trace). (H1) Left: mean VPM evoked EPSP amplitudes on postsynaptic cell types (over 50 pairs). Right: comparison of normalized in silico amplitudes (normalized by L4 E as in Sermet et al., 2019) to in vitro reference data from Sermet et al., 2019. Heatmap shows model minus reference values, thus positive values indicate a higher normalized EPSP amplitude in our model than in the reference experimental dataset. (H2) same as (H1) but for POm (normalized by L5 E as in Sermet et al., 2019).

Figure 4 with 7 supplements

Download asset Open asset

Spontaneous activity: calibration, in vivo-like dynamics, linking structure to function.

(A) Seven-column subvolume. (B) $O U_{μ}$ and $O U_{σ}$ of somatic conductance injection (B1) parameterized by population (B2) determine FRs (B3). Target FRs equal to in vivo firing rates multiplied by $P_{F R}$ . Inset: table summarizing meta-parameters. (C) Target $P_{F R}$ (x-axis) plotted against the observed $P_{F R}$ (y-axis) for each E (red) and I (blue) population after calibration. The number of sides of the shape indicates the layer (triangle used for L2/3). (D) Euclidean distance between target and observed $P_{F R}$ values (over populations) decreases over iterations for two meta-parameter combinations (upper). Values after final iteration shown for all meta-parameter combinations (lower; dashed line shows termination condition). (E) (left) Firing rates, (center) mean conductance injection, and (right) $O U_{μ}$ by population for the non-bursting simulations (one line per simulation; E and I values separated). (F) Estimated mean number of missing number synapses per neuron by population vs. the mean conductance injection, averaged over all combinations, for each population. Markers as in (C). (G1) Max normalized histograms for two meta-parameter combinations (5 ms bin size, 1σ Gaussian smoothing). (G2) Effect of meta-parameters on correlation between E and I histograms (5 ms bin size, 1σ Gaussian smoothing; central column). (G3) Fourier analysis of spontaneous activity for the 59 non-bursting simulations. (H) Activity in flat space over seven consecutive 100 ms windows. Activity smoothed (Gaussian kernel, $σ = 1 p i x e l$ ). (I) Left: correlation r-value between the histograms of layer-wise E and I populations for the 59 non-bursting simulations. Right: $R_{C / U}$ by population for the 59 non-bursting simulations. (J) Top: the mean *node participation* of E and I populations in each layer for dimensions 1 (left) and 6 (right) vs. $R_{C / U}$ for the parameter combination $[C a^{2 +}]_{o} = 1.05$ mM, $R_{O U} = 0.4$ , and $P_{F R} = 0.3$ . Markers as in (C). Line shows linear fit. Bottom: correlation r-values between mean node participation and $R_{C / U}$ when neurons with highest node participation are not included in the calculation of mean node participation (via a sliding threshold).

Figure 4—figure supplement 1

Download asset Open asset

Spontaneous activity calibration: estimating $χ$ and $ϕ$ .

(A) Estimation of $χ$ and $ϕ$ for L5 E (illustration). $Ψ_{C a^{2 +}, R_{O U}} : C_{F R} \mapsto O U_{μ}$ is estimated by dividing it into the two functions $χ$ and $ϕ$ , which are determined separately. Left: $χ : (R_{O U}, U_{F R}) \mapsto O U_{μ}$ . The color of each point shows the mean unconnected FR in a single column for L5E for different combinations of $O U_{μ}$ (x-axis) and $O U_{σ}$ (y-axis). For a given value of $R_{O U}$ interpolation is used over the datapoints to estimate the predicted FRs along a line of gradient $R_{O U}$ starting at the origin. As FRs increase monotonically with increasing $O U_{μ}$ and $O U_{σ}$ , the interpolation can be used to estimate the $O U_{μ}$ and $O U_{σ}$ combination that will produce a given target unconnected FR. Right: $ϕ_{C a^{2 +}, R_{O U}} : U_{F R} \mapsto C_{F R}$ . In each iteration, 10 simulations of the seven hexagon subvolume are run for a range of target $P_{F R}$ values. For each population (i.e., L5 E) an exponential is fit to unconnected vs. connected firing rates, and used to estimate unconnected firing rates which will achieve target connected firing rates on the next iteration. (B) Visualization of the meta-parameter space. The algorithm is run for a single combination of Ca²⁺ and $R_{O U}$ (blue), and is then generalized for other combinations (red). (C) Data for estimation of $χ$ for each population. Population FRs from unconnected simulations of a single column. Simulations vary by $O U_{μ}$ and $O U_{σ}$ . Color shows FR. Black circles show region where FRs are greater than 0 and less than the population’s in vivo reference FR. The dependence of unconnected FR on $O U_{μ}$ and $O U_{σ}$ is heterogeneous across populations. (D) Final estimation of $ϕ_{1.1, 0.4}$ for each neuron group. Exponential fit of unconnected vs. connected FRs for each population after fifth iteration. The final fits are heterogeneous across populations.

Figure 4—figure supplement 2

Download asset Open asset

Mean dendritic length and missing synapses by population.

(A) Mean dendritic length by population. Error bars show ± SD. (B) Mean estimated number of missing synapses by population. Existing and missing synapses are stacked. Error bars show ± SD.

Figure 4—figure supplement 3

Download asset Open asset

Spontaneous activity firing rate distributions.

Firing rate distributions of different subpopulations colored by m-type for two meta-parameter combinations, each over 50 seconds of simulation. Neurons which spike zero or one times are excluded. Titles of each subplot show the percentage of neurons spiking at least once, and the mean firing rate of neurons of neurons spiking at least once.

Figure 4—figure supplement 4

Download asset Open asset

Firing rate statistics.

(A) Histograms of the inverse of inter-spike intervals (ISI) (a proxy of non-zero firing rate) for three different meta parameter combinations of $[C a^{2 +}]_{o}$ , $R_{O U}$ and $P_{F R}$ for the connected (red) and unconnected (blue) circuits. (B) For each dimension, the markers show the weighted average of the inverse ISI for all neurons, where the weights are given by node participation in the given dimension. Shaded regions show standard error of the mean. Color scheme as in (A). Inset, ratio of connected vs. unconnected means, i.e., ratio of the red vs. blue curves. Note that the minor dependence on the unconnected firing rate can be explained by the correlation between dimension and the layer of a neuron, combined with different firing rates in different layers. (C) Same as (B) but additionally showing dimension 6.

Figure 4—figure supplement 5

Download asset Open asset

Over-/under-expression of neurons that maximize node participation.

Negative logarithm of the probability of the over-/under-expressions (see ‘Methods’), in red, respectively blue, for all mtypes, calculated for the top 5% of neurons that maximize node participation in each dimension.

Figure 4—video 1

Download asset

posterframe for video — Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for the central column of the seven-column subvolume for each of the 60 meta-parameter combinations.

Meta-parameter values are shown in brackets in the following order: Ca²⁺, $P_{F R}$ , $R_{O U}$ .

Figure 4—video 2

Download asset

Figure 5 with 5 supplements

Download asset Open asset

Layer-wise population responses to single whisker deflection and active whisker touch stimuli closely match in vivo millisecond dynamics and response amplitudes.

(A1) Stimuli are simulated by activating a proportion ( $F_{P}$ ) of fibers projecting to the central hexagon of the seven-column subvolume. (A2) Each fiber is assigned a spike time from a PSTH of VPM activity recorded in vivo for either a single whisker deflection or active whisker touch paradigm. (B) Spiking activity and histograms (baseline and max normalized) for each layer-wise E and I population for the parameter combination $[C a^{2 +}]_{o} = 1.05$ mM, $R_{O U} = 0.4$ , $P_{F R} = 0.3$ , and $F_{P} = 10 %$ for a 2.5-second section of the 10 whisker deflection test protocol. (C) Trial-averaged PSTHs (baseline and max normalized) in response to the single whisker deflection stimulus for all of the parameter combinations which passed the initial criteria, and the in vivo references. (D) Spatiotemporal evolution of the trial-averaged stimulus-response in flat space for the same parameter combination in (B). (E) Left: latencies of different layer-wise E and I populations for all criteria passing combinations in response to single whisker deflection stimuli. Right: latencies for inhibitory interneuron subpopulations in response to active whisker touch stimuli. Corresponding in vivo references are shown. (F) Heatmap showing the effect of the meta-parameters on the difference of $R_{E}$ for the entire central column from the corresponding in vivo reference. Parameter combinations which failed the criteria tests are masked. (G) Left: ratio $R_{E}$ between maximum evoked response and spontaneous baseline activity by population. One line per simulation. Simulation line color shows the difference of $R_{E}$ for the entire central column from the corresponding in vivo reference (color values shown in G). Thick black line shows median of in silico values. White shows in vivo reference. Right: normalized ratios of each simulation by dividing by $R_{E}$ for the entire central column.

Figure 5—figure supplement 1

Download asset Open asset

Whisker deflection responses.

(A) PSTHs in response to the single whisker deflection stimulus for all parameter combinations. PSTHs of simulations which passed the initial criteria are colored green, whilst those which did not are colored gray. (B) Layer-wise PV+ and Sst+ subpopulation PSTHs for simulations which passed the criteria. (C) Analysis of evoked responses of L5 E under two simulation cases. The first case uses the mean of the VPM to L4 E and L6 E synaptic conductance for the VPM to L5 E conductance, whilst the second uses the maximum of the two values (VPM to L6 E). (C1) For the two simulation cases, L5 E PSTHs for the criteria passing (green) and failing (gray) simulations. For the mean case, some of the simulations that do not burst or do not have a secondary rebound, fail the criteria because of low signal to noise ratio. (C2) Ratios between evoked and spontaneous activity of L5 E population for criteria passing simulations. There are fewer criteria passing simulations for the mean case, and those that do are much lower than the in vivo reference (40.2). (D1) Proportion of spiking cells by population in the central column over 100 ms following stimulus onset. Each line represents a different criteria passing simulation. (D2) Heatmap showing the effect of the meta-parameters on the proportion of spiking cells over the entire central column. (E) Heatmap showing the effect of the meta-parameters on the proportion of spiking cells which spike only once over 100 ms following stimulus onset.

Figure 5—figure supplement 2

Download asset Open asset

Active whisker touch responses.

Same as Figure 5 but using an active whisker touch VPM stimulus and in vivo reference for the active whisker touch paradigm (see Methods).

Figure 5—video 1

Download asset

Figure 5—video 2

Download asset

Figure 5—video 3

Download asset

Figure 6 with 6 supplements

Download asset Open asset

Reproducing and extending detailed experiments exploring the canonical model and encoding of contrast, rate-coded and synchronous information.

First experiment: reproducing the effect of optogenetic inactivation of L4 pyramidal cells on L2/3 stimulus-responses and predicting the role of other pathways through circuit lesions. (A) Schematics of whisker kinetics and VPM fiber rates during 500-ms-long whisker hold stimulus. Fraction of VPM fibers coding for each kinetic feature are taken from Petersen et al., 2008. (B) Mimicking the effect of activation of the Halo inhibitory opsin in silico. Injected somatic hyperpolarizing current mimicking opsin activation (top), and the resulting somatic voltage trace from a combination of injected conductance, current, and synaptic PSPs from the network (bottom). (C) Comparison of average traces from selected L2/3 PCs (see ‘Methods’) in control (black) and optogenetically inhibited (green) conditions. (D) Same as (C), but instead of mimicking the optogenetic inhibition of L4 PCs, only the connections to L2/3 PCs are ‘cut’ (compare inset with the one in C). The right part depicts connections systematically cut from PCs in all layers, while the left shows L4 only for a better visual comparison with the conditions of Varani et al., 2022 in (C). Second experiment: contrast tuning modulation by optogenetic activation of PV+ interneurons. (E) Spatial distribution of firing rates of VPM fibers projecting to a single simulated column at different points in time, corresponding to linear drifting gratings with a temporal frequency of $f_{t e m p} = 2$ Hz and a spatial frequency of $f_{s p a t} = 0.001$ cycles/extμm. (F) Firing rate signal of a single VPM fiber corresponding to a random sequence of drifting grating patterns with different contrast levels $C$ , as indicated by the legend. All optogenetic stimuli targeting PV+ (or Sst+) interneurons were completely overlapping with the grating stimulus intervals, as indicated by the blue bars. (G1) Spiking activity and PSTH of an exemplary PV+ interneuron over 10 repetitions (trials) in response to a grating stimulus at maximum contrast ( $C = 1.0$ ) without (control) and with medium strength (150%) optogenetic stimulation. (H1) Contrast tuning modulation of an exemplary PV+ interneuron by different levels of PV+ optogenetic stimulation, ranging from 0% (control) to 300%. The individual markers denote actual data points (mean ± SD over 10 repetitions, normalized to baseline response at maximum contrast) while curves illustrate sigmoidal fits to these data points. (G2/H2) Same as (C1/D1), but for exemplary PCs. (H2) inset: Illustration of sigmoidal parameters $m$ , $n$ , $R_{m a x}$ , and $c_{50}$ . Third experiment: encoding of synchronous and rate-coded signals by interneuron subtypes. (I) A binary signal is encoded either through changes in the rate or synchronicity of optogenetic pulses applied to a set of PCs. We define a direct correspondence between single optogenetic pulses and single spikes. Spiking activity is measured in PV+, Sst+, and 5HT3aR+ cells, and the mutual information between the binned activity of individual cells and the original binary signal is measured. (J) Visualization of the rate (upper) and synchronous (lower) coded stimulus experiments stimulating 1000 PCs, showing the binary signal (top), input neuron spike trains for 40 neurons (middle), and responses of the three L2/3 neuron types (bottom). (K) Upper: results for the rate coded stimulus experiment. Left: mutual information between spiking activity and the binary signal (one point for each cell that spiked at least once). Only activity in the 50 ms following the change of the binary signal is considered. Cells with mutual information not significantly different from that of a shuffled control are colored gray. Center: same as left but considering all time bins. Right: the effect of the number of stimulus neurons on the mean single cell mutual information for each subpopulation. Lower: the same as upper but for the synchronous stimulus type.

Figure 6—figure supplement 1

Download asset Open asset

In silico experimental methods.

(A) Synaptic pathway lesions. After selecting a pre- (green) and postsynaptic (blue) population of neurons, connections between them in the specified direction are removed (red arrows). (B) Optogenetic manipulations. (B1) Decay with depth of the strength of a stimulus applied to mimic the effect of optogenetic manipulations with 470 nm (blue) and 595 nm (yellow) light. Indicated for exemplary intensities (values of $I_{0}$ ). (B2) Targeted neuronal populations in two in silico optogenetic experiments. Density of affected cells (combination of expression level and light depth) indicated by the corresponding color of light. Unaffected populations in gray. Left: excitation of inhibitory populations. Right: inhibition of PCs in L4. (C) Stimulation with thalamic inputs. (C1) Each thalamic fiber in the model is assigned a position in a flat coordinate system. Based on the coordinates, type of fiber (VPM vs. POm), or randomly, time series of sensory stimulation are assigned to the fibers. (C2) Sensory stimuli are transformed by a transfer function capturing pre-thalamic processing. The transfer function can simply be identity. The transformed signal is then used with a spiking model to generate stochastic spike trains that serve as inputs into a simulation.

Figure 6—figure supplement 2

Download asset Open asset

Additional validations: reproducing the experiments of Varani et al., 2022.

(A) Raster plots of the microcircuit’s activity and population firing rates below. (A1) Control conditions, (A2) 95% of L4 PCs inhibited (by direct somatic current injection depicted in (B) above, see ‘Methods’). (B) Voltage traces of all L2/3 (top) and all L4 (bottom) PCs. Panels show spiking traces (top), and subthreshold traces (bottom). (B1) and (B2) depict the same conditions as (C) above.

Figure 6—figure supplement 3

Download asset Open asset

Calibration of the drifting grating stimulus based on peak firing rates.

The drifting grating stimulus was calibrated by running a parameter scan of 45 simulations to determine optimal values for $F_{P}$ , $R_{b k}$ , and $R_{p e a k}$ . The maximum (top) and average (middle) of the first and second peak firing rates $r_{1}$ and $r_{2}$ (connected points) over the whole population of PCs, together with the average of the normalized peak difference (Michelson contrast) ${\hat{r}}_{d i f f}$ (bottom) are summarized. Different colors indicate different contrast levels $C \in [0.06, 0.12, 0.5, 1.0]$ . The rate threshold $r_{t h} = 30$ Hz at maximum contrast, the calibration target values that were taken into account for selecting the optimal parameter set, as well as the optimal parameter set ( $F_{P} = 1.0$ , $R_{b k} = 0.2$ Hz, and $R_{p e a k} = 10.0$ Hz), are highlighted.

Figure 6—figure supplement 4

Download asset Open asset

Contrast tuning by layer and paradoxical effect of optogenetic PV+ stimulation.

(A) Normalized contrast tuning functions of layer-wise PV+ (upper; blue) and PC (lower; red) subpopulations for different levels of optogenetic PV+ stimulation ranging from 0% to 300%, as indicated by the legend. (B) Same data as in (A), but plotted dependent on the optogenetic PV+ stimulation strength (x-axis) for the different contrasts, as indicated by the legend. At high contrasts, the optogenetic stimulation has a paradoxical effect on L6 PV+ neurons, i.e., reducing their activity (black arrows).

Figure 6—figure supplement 5

Download asset Open asset

Sigmoidal parameter fits to contrast tuning functions.

(A1) Scatter plots showing the fitted sigmoidal parameter values (see Figure 6H2 inset; ‘Methods’) in control condition vs. PV+ optogenetic stimulation at lowest (50%) and highest (300%) level respectively for all 259 robustly tuned PV+ interneurons. Diamond markers indicate mean values over neurons, including all optogenetic stimulation levels from 50% to 300%. Colors as in Figure 6H legend. (A2) Same as (A1), but for all 228 robustly tuned PCs.

Figure 6—figure supplement 6

Download asset Open asset

Modeling the effects of optogenetic stimulation of interneurons.

(A1–A3) Divisive scaling (Div), subtractive shifting (Sub), and saturation additive (SA) model fits (dashed lines) to responses of an exemplary PV+ interneuron (solid lines) at different levels of optogenetic PV+ stimulation, ranging from 0% (baseline) to 300%. (B1) Goodness of Div/Sub/SA model fits ( $μ \pm S E M$ of $r^{2}$ score) for different depolarization levels over all 259 robustly tuned PV +interneurons, showing that direct photostimulation effects on interneurons can be best described by the saturation additive model. (C1) Polar plot of $S$ (saturation) and $A$ (additive) parameter values of the SA model fits to all 259 robustly tuned PV+ interneurons at different levels of optogenetic PV+ stimulation. Diamond markers indicate mean values (in polar coordinates) over neurons, colored arc lines the circular SD of the polar angles. (B2) Same as (B1), but for conductance-based model fits describing indirect photostimulation effects on all 228 robustly tuned PCs, assuming a saturating additive model description of interneurons. (C2) Same as (C1), but for the conductance-based model fits to all robustly tuned PCs.

Figure 7 with 1 supplement

Download asset Open asset

The effect of inhibitory targeting trends observed in electron microscopy on spontaneous and stimulus-evoked activity.

(A1) Panel illustrating interneuron synaptic targeting features and caption taken from Schneider-Mizell et al., 2025 (under CC-BY 4.0 license). (i) For determining inhibitory cell subclasses, connectivity properties were used such as an axon (green) making synapses (green dots) the perisomatic region of a target pyramidal cell (purple). (ii) Dendritic compartment definitions for excitatory neurons. (iii) Cartoon of a multisynaptic connection (left) and the synapses within the multisynaptic connection considered ‘clumped’ along the presynaptic axon (right). (A2) Exemplar neurons showing the synapse locations from different inhibitory m-type groupings in the original and SM connectome, respectively: distal targeting (DistTC: martinotti, bipolar, double bouquet cells), Perisomatic targeting (PeriTC: large basket, nest basket cells), Sparse targeting (SparTC: descending axon, neurogliaform, horizontal axon cells), Inhibitory targeting (InhTC: bitufted, small basket cells). (A3) Heatmaps showing the mean fraction of efferent synapses from the inhibitory m-type groupings onto inhibitory neurons, somas, proximal dendrites, apical dendrites, distal dendrites, or that are part of clumped or multisynaptic connections. Heatmaps show mean fractions (left to right) for the original, Schneider-Mizell et al., 2025 characterization of the MICrONS dataset (MICrONS Consortium, 2025), SM-connectome, and the change between the original and SM-connectome. (B1) Proportional change in mean conductance injection between original and SM-connectome by population. (B2) Distribution of ratios between connected and disconnected firing rates ( $R_{C / U}$ ) by population over non-bursting meta-parameter combinations for original and SM-connectome. (B3) Comparison of correlation (r-values) between E and I histograms (5 ms bin size, 1σ Gaussian smoothing; central column) between original and SM-connectome (each point represents a specific meta-parameter combination). (C) Scatter plot showing the change in the mean number of afferent inhibitory synapses onto each E (red) and I (blue) population after calibration versus the proportional change in mean conductance injection. The number of sides of the shape indicates the layer (triangle used for L2/3). (D1) Proportional change in the ratio between the peak of the evoked response and the pre-stimulus baseline activity level ( $R_{E}$ ) between the original and SM-connectome. Plot shows distribution over meta-parameter combinations which passed the initial assessment of in vivo similarity to the in vivo data based on latencies and decays (see ‘Methods’). (D2) Change in the latencies of populations peak responses between original and SM-connectome (distribution over same meta-parameter combinations shown in D1). (E) Mean membrane potential responses for layer-wise subpopulations for the meta-parameter combination $[C a^{2 +}]_{o} = 1.05$ mM, $R_{O U} = 0.4$ , $P_{F R} = 0.3$ , and $F_{P} = 10 %$ . Responses averaged over 20 stimulus repetitions with an inter-trial-interval of 500 ms.

Figure 7—figure supplement 1

Download asset Open asset

Re-calibration of SM-connectome.

(A) The SM-connectome took six iterations to re-calibrate for the full 60 meta-parameter combinations. The mean conductance injections found for the original connectome for the 60 meta-parameter combinations were used for the first iteration of the re-optimization and lead to decreases in firing rates for all inhibitory populations except L1 I (A1). In (A1) and (A2), blue and yellow points refer to Ca²⁺ 1.05 and 1.1, respectively. After six iterations, the firing rates converged to the target firing rates (**A2–A5**) and there was minimal difference in population firing rates over the 60 meta-parameter combinations between the original and SM-connectome (A4) .(B) Correlation r-values between the E and I populations, as for the original connectome in the main text. (C) Unconnected vs. connected MFRs for the different populations after the sixth iterations. Unlike for the original connectome, firing rates decrease following connection of the network.

Figure 8 with 7 supplements

Download asset Open asset

Full nbS1 simulations: independent functional units, millisecond-precise stimulus-responses, selective propagation of stimulus-evoked activity through mid-range connectivity, linking structure to function.

(A) Spiking activity and max normalized histograms for layer-wise E and I populations for the meta-parameter combination $[C a^{2 +}]_{o}$ : 1.05 mM, $R_{O U}$ : 0.4, $P_{F R}$ : 0.15, and the $0^{t h}$ (upper) and $39^{t h}$ (lower) hexagonal subvolumes (shown in B). (B1) Mean FRs across the model. Columns 0 and 39 (subsequently used) are highlighted. (B2) Spiking correlations of E/I populations in 240 hexagonal subvolumes (diameter: ≈460 µm). (C) Target vs. observed $P_{F R}$ values for the four non-bursting simulations. The number of sides of each marker corresponds to the layer (triangle represents L2/3). Red: E, blue: I. (D) Correlations between E and I populations by layer for the four non-bursting simulations for the two hexagons (top and bottom). Color of line from light to dark represents $P_{F R}$ . Middle: distributions of spiking correlations of layer-specific populations in columns (520 $μ m$ ) of the nbS1 model (black dots; mean values: gray). Compared to the central column when the seven-column subvolume is simulated in isolation (green). (E) Trial-averaged activity in flatspace following single whisker deflections. Lower: subset of time windows shown with only the top 2% of bins for each time window after baseline activity level subtracted. (F) Trial-averaged PSTHs (baseline and max normalized) for column 0. For $F_{P} = 15 %$ and $F_{P} = 25 %$ , respectively, the evoked responses passed and failed the latency and decay-based criteria tests for similarity to in vivo. Response for $F_{P} = 20 %$ passed the tests for all populations except L5 E, which showed a more sustained response. (G) Trial-averaged response of column 39. (H) Correlations of mid-range inputs. Upper: distribution of spiking correlations of inputs into 240 subvolumes. Calculated at reduced spatial resolution, based on connection counts and correlations between the subvolumes (see ‘Methods’). Subvolumes are sorted by the E/I correlation of neurons within them. Lower: mean of the correlations of inputs into 240 subvolumes vs. internal spiking correlation for all subvolumes during spontaneous activity. Black line: linear fit. calculated based on connection counts and correlations between 50 $μ m$ hexagonal subvolumes (see ‘Methods’). Data for evoked activity is shown in Figure 8—figure supplement 1. (I) Upper: spiking correlations of subvolumes against their distances in spontaneous (red) and evoked (orange: $F_{p} = 0.15$ %, blue: $F_{p} = 0.2$ %, green: $F_{p} = 0.25$ ) activity. For increased spatial resolution, smaller (58 $μ m$ ) subvolumes were used, hence correlation values are not comparable to values in (B). Lower: distribution of soma distances of neuron pairs connected by local (orange) or mid-range (blue) connectivity. Green: distances of all pairs, independent of connectivity. (J) Upper: classification of subvolumes based on low or high internal correlation and membership in mid-range rich club Data for evoked activity shown in Figure 8—figure supplement 1. Lower: E/I correlation of subvolumes for members of the rich club (red) and non-members (blue). (K) FRs of subvolumes following a stimulus with $F_{P} = 0.15$ % (top) and $F_{P} = 0.2$ % (bottom). Mean (lines) and SEM (shaded area) over 10 repetitions shown for subvolumes directly innervated by the VPM stimulus (black), strongly innervated by directly innervated subvolumes (with over 10⁶ synapses, red), or by medium strength (over 10⁵ synapses, yellow) or weak (over 10⁴ synapses) indirect innervation. Dashed lines indicate locations of peaks.

Figure 8—figure supplement 1

Download asset Open asset

Full nbS1—supplementary.

(A) Distance dependence of spiking correlations. Digitized from Reyes-Puerta et al., 2015a. (B) Evoked responses of a single hexagonal subvolume over 10 single whisker deflections. (C) Trial-averaged activity of the full circuit for $F_{P} = 15 %$ following the single whisker deflection stimulus (top). Middle and bottom: trial-averaged activity of the full circuit for $F_{P} = 15 %$ following the single whisker deflection stimulus for $F_{P} = 15 %$ and $F_{P} = 20 %$ , showing only the top 2% of bins for each time window after the baseline activity level has been subtracted. (D) Distribution of spiking correlations of inputs into 240 subvolumes. Calculated at reduced spatial resolution, based on connection counts and correlations between the subvolumes (see Supp. Methods). Subvolumes are sorted by the E/I correlation of neurons within them. For evoked activity with $F_{P} = 20 %$ . (E) Mean of the correlations of inputs into 240 subvolumes (as in D) against internal spiking correlation for all subvolumes during spontaneous activity. Black line: linear fit. Calculated based on connection counts and correlations between 50 $μ m$ hexagonal subvolumes (see Supp. Methods).

Figure 8—video 1

Download asset

Figure 8—video 2

Download asset

Figure 8—video 3

Download asset

Figure 8—video 4

Download asset

Figure 8—video 5

Download asset

Figure 8—video 6

Download asset

Tables

Table 1

Previously published modeling techniques and data to parameterize them that were combined in this work to model the physiology of nbS1 and conduct in silico experimentation.

Data
Stage	Topic	Reference
Neuron physiology	E-type composition	Markram et al., 2015; Markram et al., 2004
	Electrophysiological recordings	Markram et al., 2015
		Larkum et al., 2001
		Nevian et al., 2007
	Ion-channel models	Reva et al., 2023
Synaptic physiology	Synaptic pathway physiology	Markram et al., 2015
		Barros-Zulaica et al., 2019
		Gupta et al., 2000
		13 additional sources in Supplementary files 5–7
In vivo-like activity	In vivo spont. firing rates	Reyes-Puerta et al., 2015b
		De Kock et al., 2007
	In vivo evoked responses	Reyes-Puerta et al., 2015b
		Yu et al., 2019
	In vivo thalamic spiking	Diamond et al., 1992
		Yu et al., 2019
In silico experimentation	L4-L2/3 pathway	Varani et al., 2022
	Visual contrast	Shapiro et al., 2022
	Coding in inhibitory subpopulations	Prince et al., 2021
Modeling methods
Neuron physiology	E-model building	Van Geit et al., 2016
		Reva et al., 2023
Synaptic physiology	Synapse model building	Ecker et al., 2020
	Model of multi-vesicular release	Barros-Zulaica et al., 2019
In vivo-like activity	Missing input compensation	*New, original methods based on
		Destexhe et al., 2001
In silico experimentation		*New, original methods

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Additional information
Software	Electrical model optimization	https://github.com/BlueBrain/singlecell-emodel-suite; Reva et al., 2023; Van Geit et al., 2024
Software	Simulation	This paper	10.5281/zenodo.8075202
Software	Model loading and interaction	This paper	10.5281/zenodo.8026852
Software	Model and simulation analysis	This paper	10.5281/zenodo.8016989
Software	Specific simulation configurations	This paper	10.5281/zenodo.7930276
Software	Connectome Manipulation	https://github.com/BlueBrain/connectome-manipulator; Pokorny et al., 2025; Wolf et al., 2024

Additional files

Supplementary file 1 Table of predictions made by the study.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp1-v1.pdf
Download elife-99693-supp1-v1.pdf
Supplementary file 2 Synaptic parameters of excitatory pathways. Average class parameters are marked in bold and are used predictively (in lack of reference in vitro data) for the remaining pathways belonging to the same class. Physical dimensions are as follows: peak conductance $\hat{g}$ : nS, depression and facilitation time constants $D$ , $F$ , and the EPSC $τ_{d e c a y}$ : ms, the Hill coefficient of the nonlinear $[C a^{2 +}]_{o}$ dependent scaling of release probability $U_{H i l l}$ : mM, the release probability $U_{S E}$ , the average number of vesicles in the release-ready pool $N_{R R P}$ , and the NMDA/AMPA ratio ${\hat{g}}_{r a t i o}$ are dimensionless.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp2-v1.pdf
Download elife-99693-supp2-v1.pdf
Supplementary file 3 Synaptic parameters of inhibitory pathways. Average class parameters are marked in bold and are used predictively (in lack of reference in vitro data) for the remaining pathways belonging to the same class. Physical dimensions are as follows: peak conductance $\hat{g}$ : nS, depression and facilitation time constants $D$ , $F$ , and the IPSC $τ_{d e c a y}$ : ms, the Hill coefficient of the nonlinear $[C a^{2 +}]_{o}$ dependent scaling of release probability $U_{H i l l}$ : mM, the release probability $U_{S E}$ , the average number of vesicles in the release-ready pool $N_{R R P}$ , and the $G A B A_{B} / G A B A_{A}$ ratio ${\hat{g}}_{r a t i o}$ are dimensionless.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp3-v1.pdf
Download elife-99693-supp3-v1.pdf
Supplementary file 4 Synaptic parameters of thalamocortical synaptic pathways. Values taken from the internal connectivity (Supplementary file 2) are marked in bold. Physical dimensions are as follows: peak conductance $\hat{g}$ : nS, depression and facilitation time constants $D$ , $F$ , and the EPSC $τ_{d e c a y}$ : ms, the Hill coefficient of the nonlinear $[C a^{2 +}]_{o}$ dependent scaling of release probability $U_{H i l l}$ : mM, the release probability $U_{S E}$ , the average number of vesicles in the release-ready pool $N_{R R P}$ , and the NMDA/AMPA ratio ${\hat{g}}_{r a t i o}$ are dimensionless.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp4-v1.pdf
Download elife-99693-supp4-v1.pdf
Supplementary file 5 Validation of PSP amplitudes. See Figure 3B1.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp5-v1.pdf
Download elife-99693-supp5-v1.pdf
Supplementary file 6 Validation of first PSP amplitudes’ CVs. See Figure 3B2.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp6-v1.pdf
Download elife-99693-supp6-v1.pdf
Supplementary file 7 Validation of mPSC frequency. See Figure 3E2.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp7-v1.pdf
Download elife-99693-supp7-v1.pdf
Supplementary file 8 Assumptions. Additionally, we assume that various input data generalizations between brain regions, organisms and animal ages do not affect validity. These are not listed again.: https://cdn.elifesciences.org/articles/99693/elife-99693-supp8-v1.pdf
Download elife-99693-supp8-v1.pdf
MDAR checklist: https://cdn.elifesciences.org/articles/99693/elife-99693-mdarchecklist1-v1.docx
Download elife-99693-mdarchecklist1-v1.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Mendeley

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

James B Isbister
Andras Ecker
Christoph Pokorny
Sirio Bolaños-Puchet
Daniela Egas Santander
Alexis Arnaudon
Omar Awile
Barros-Zulaica Natali
Jorge Blanco Alonso
Elvis Boci
Giuseppe Chindemi
Jean-Denis Courcol
Tanguy Damart
Thomas Delemontex
Alexander Dietz
Gianluca Ficarelli
Mike Gevaert
Joni Herttuainen
Genrich Ivaska
Weina Ji
Daniel Keller
James King
Pramod Kumbhar
Samuel Lapere
Polina Litvak
Darshan Mandge
Eilif B Muller
Fernando Pereira
Judit Planas
Rajnish Ranjan
Maria Reva
Armando Romani
Christian Rössert
Felix Schuermann
Vishal Sood
Aleksandra Teska
Anil Tuncel
Wener Van Geit
Matthias Wolf
Henry Markram
Srikanth Ramaswamy
Michael W Reimann

(2026)

Modeling and simulation of neocortical micro- and mesocircuitry (Part II, Physiology and experimentation)

eLife 13:RP99693.

https://doi.org/10.7554/eLife.99693.3

Share this article

Cite this article

Overview of the physiology and simulation workflow.

Modeling and validation of neuron physiology.

Modeling and validation of synaptic physiology.

Spontaneous activity: calibration, in vivo-like dynamics, linking structure to function.

Spontaneous activity calibration: estimating χ and ϕ.

Mean dendritic length and missing synapses by population.

Spontaneous activity firing rate distributions.

Firing rate statistics.

Over-/under-expression of neurons that maximize node participation.

Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for the central column of the seven-column subvolume for each of the 60 meta-parameter combinations.

Visualization of spontaneous activity for the seven-column subvolume for the parameter combination Ca2+=1.1 mM, PFR=0.9, ROU=0.2 after collapsing activity to flatspace (activity binned and smoothed).

Layer-wise population responses to single whisker deflection and active whisker touch stimuli closely match in vivo millisecond dynamics and response amplitudes.

Whisker deflection responses.

Active whisker touch responses.

Layer-wise E and I population rasters and max-normalized histograms of evoked activity during the 10 single whisker deflection protocol for each of the simulated meta-parameter combinations.

Trial-averaged max-normalized histograms of evoked activity during the 10 single whisker deflection protocol for each of the simulated meta-parameter combinations.

Visualization of trial-averaged response of the seven-column subvolume over the 10 single whisker deflection protocol for the parameter combination Ca2+=1.1 mM, PFR=0.3, ROU=0.4, FP=20% after collapsing activity to flatspace (activity binned and smoothed).

Reproducing and extending detailed experiments exploring the canonical model and encoding of contrast, rate-coded and synchronous information.

In silico experimental methods.

Additional validations: reproducing the experiments of Varani et al., 2022.

Calibration of the drifting grating stimulus based on peak firing rates.

Contrast tuning by layer and paradoxical effect of optogenetic PV+ stimulation.

Sigmoidal parameter fits to contrast tuning functions.

Modeling the effects of optogenetic stimulation of interneurons.

The effect of inhibitory targeting trends observed in electron microscopy on spontaneous and stimulus-evoked activity.

Re-calibration of SM-connectome.

Full nbS1 simulations: independent functional units, millisecond-precise stimulus-responses, selective propagation of stimulus-evoked activity through mid-range connectivity, linking structure to function.

Full nbS1—supplementary.

Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for hex0 for meta-parameter combinations: Ca2+=1.05 mM, ROU=0.4, and PFR=0.05−0.3.

Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for hex39 for meta-parameter combinations: Ca2+=1.05 mM, ROU=0.4, and PFR=0.05−0.3.

Visualization of spontaneous activity for the full nbS1 model for the parameter combination Ca2+=1.05 mM, PFR=0.15, ROU=0.4 after collapsing activity to flatspace (activity binned and smoothed).

Layer-wise E and I population rasters and max-normalized histograms of stimulus evoked activity for hex0 for each meta-parameter combination Ca2+=1.05 mM, ROU=0.4, PFR=0.15, and FP=0.1−0.25.

Layer-wise E and I population rasters and max-normalized histograms of stimulus evoked activity for hex39 for each meta-parameter combination Ca2+=1.05 mM, ROU=0.4, PFR=0.15, and FP=0.1−0.25.

Visualization of trial-averaged response of the full nbS1 model over the 10 single whisker deflection protocol for the parameter combination Ca2+=1.1 mM, PFR=0.15,ROU=0.4, and FP=20% after collapsing activity to flatspace (activity binned and smoothed).

Previously published modeling techniques and data to parameterize them that were combined in this work to model the physiology of nbS1 and conduct in silico experimentation.

Supplementary file 1

Supplementary file 2

Supplementary file 3

Supplementary file 4

Supplementary file 5

Supplementary file 6

Supplementary file 7

Supplementary file 8

MDAR checklist

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Spontaneous activity calibration: estimating $χ$ and $ϕ$ .

Visualization of spontaneous activity for the seven-column subvolume for the parameter combination $C a^{2 +} = 1.1$ mM, $P_{F R} = 0.9$ , $R_{O U} = 0.2$ after collapsing activity to flatspace (activity binned and smoothed).

Visualization of trial-averaged response of the seven-column subvolume over the 10 single whisker deflection protocol for the parameter combination $C a^{2 +} = 1.1$ mM, $P_{F R} = 0.3$ , $R_{O U} = 0.4$ , $F_{P} = 20 %$ after collapsing activity to flatspace (activity binned and smoothed).

Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for hex0 for meta-parameter combinations: $C a^{2 +} = 1.05$ mM, $R_{O U} = 0.4$ , and $P_{F R} = 0.05 - 0.3$ .

Layer-wise E and I population rasters and max-normalized histograms of spontaneous activity for hex39 for meta-parameter combinations: $C a^{2 +} = 1.05$ mM, $R_{O U} = 0.4$ , and $P_{F R} = 0.05 - 0.3$ .

Visualization of spontaneous activity for the full nbS1 model for the parameter combination $C a^{2 +} = 1.05$ mM, $P_{F R} = 0.15$ , $R_{O U} = 0.4$ after collapsing activity to flatspace (activity binned and smoothed).

Layer-wise E and I population rasters and max-normalized histograms of stimulus evoked activity for hex0 for each meta-parameter combination $C a^{2 +} = 1.05$ mM, $R_{O U} = 0.4$ , $P_{F R} = 0.15$ , and $F_{P} = 0.1 - 0.25$ .

Layer-wise E and I population rasters and max-normalized histograms of stimulus evoked activity for hex39 for each meta-parameter combination $C a^{2 +} = 1.05$ mM, $R_{O U} = 0.4$ , $P_{F R} = 0.15$ , and $F_{P} = 0.1 - 0.25$ .

Visualization of trial-averaged response of the full nbS1 model over the 10 single whisker deflection protocol for the parameter combination $C a^{2 +} = 1.1$ mM, $P_{F R} = 0.15$ , $R_{O U} = 0.4$ , and $F_{P} = 20 %$ after collapsing activity to flatspace (activity binned and smoothed).