Homeodynamic feedback inhibition control in whole-brain simulations Authors: Jan Stasinski, Halgurd Taher, Jil Mona Meier, Michael Schirner, Dionysios Perdikis, Petra Ritter, Lyle J. Graham, Marcus Kaiser, Lyle J. Graham, Marcus Kaiser, Lyle J. Graham, Marcus Kaiser Published: 2024 ABSTRACT Simulations of large-scale brain dynamics are often impacted by overexcitation resulting from heavy-tailed structural network distributions, leading to biologically implausible simulation results. We implement a homeodynamic plasticity mechanism, known from other modeling work, in the widely used Jansen-Rit neural mass model for The Virtual Brain (TVB) simulation framework. We aim at heterogeneously adjusting the inhibitory coupling weights to reach desired dynamic regimes in each brain region. We show that, by using this dynamic approach, we can control the target activity level to obtain biologically plausible brain simulations, including post-synaptic potentials and blood-oxygen-level-dependent functional magnetic resonance imaging (fMRI) activity. We demonstrate that the derived dynamic Feedback Inhibitory Control (dFIC) can be used to enable increased variability of model dynamics. We derive the conditions under which the simulated brain activity converges to a predefined target level analytically and via simulations. We highlight the benefits of dFIC in the context of fitting the TVB model to static and dynamic measures of fMRI empirical data, accounting for global synchronization across the whole brain. The proposed novel method helps computational neuroscientists, especially TVB users, to easily “tune” brain models to desired dynamical regimes depending on the specific requirements of each study. The presented method is a steppingstone towards increased biological realism in brain network models and a valuable tool to better understand their underlying behavior. Introduction Brain network modeling (BNM) allows us to investigate how phenomena observed in electrophysiological and neuroimaging experiments relate to underlying biology, without the necessity of invasive and resource-consuming experiments. Mean-field simulations of whole-brain dynamics enable formulating and testing hypotheses about the underlying mechanisms mainly at mesoscopic and macroscopic scales, while at the same time considering individual differences in brain anatomy, including the roles of subunits such as specific cell types, connections, or synaptic mechanisms. The Virtual Brain (TVB; thevirtualbrain.org) offers a multifunctional platform for running individualized brain simulations using BNMs [1–3]. In the TVB framework, simulations are designed by coupling neural mass models, each corresponding to a brain region, i.e., a node, in the structural brain network. The edges of the structural brain network consist of white-matter tracts and are approximated from diffusion-weighted Magnetic Resonance Imaging (MRI) data using tractography [4], resulting in weights (structural connectivity (SC)) and tract lengths matrices. Typically, the output of BNM simulations is fitted to empirical data to explain high-level properties of empirical signals, including functional magnetic resonance imaging (fMRI) networks, electroencephalography (EEG), or firing rates of neuronal populations [5–8]. In recent years, this approach has found its use in many theoretical and clinical applications. Whole-brain simulations were used to investigate brain dynamics in Alzheimer’s disease [9–11], epilepsy [12,13] and schizophrenia [14]. Theoretical applications include studies of excitatory-inhibitory balance and emergent neural mechanisms over a range of temporal and spatial scales [15–17]. One challenge of large-scale brain simulations is to obtain ‘biologically plausible’ brain signals that represent aspects of real brain phenomena, such as network interactions and dynamic changes due to interventions or pathology. Here we describe brain signals as ‘biologically plausible’ if they resemble characteristics of empirical data, e.g., firing rates, synchronicity, or functional connectivity (FC). Experimental studies have shown that neuronal populations in resting state exhibit on average low to moderate levels of asynchronous activity, with low variability between regions. Unlike the saturation of activity or the large spread of firing rates that can be obtained by a connectome that spans several orders of magnitude [18–21]. For example, in vivo studies report a ∼3 Hz firing rate for excitatory neurons and irregular (Poisson-like) cortical activity [22,23]. Thus, simulations should produce similar characteristics. For this purpose, an established method is to fit the global coupling parameter to adjust the input variability across regions. However, global coupling has two distinct effects: (i) it controls interactions and therefore correlations among regions and (ii) at the same time impacts their excitability. On the one hand, lowering global coupling to avoid overexcitation can result in decreased network interactions, potentially rendering them incomparable to empirical data. On the other hand, overexcitation and highly correlated activity driven by strong coupling can be considered implausible and unlikely to exhibit switching between different regimes as some of them require small inputs and low levels of activity to be present. Additionally, we argue that for the simulated signals the considered ‘biologically plausible’ they should exhibit complex dynamics like criticality and multistability, which have been discussed as important for brain dynamics [24–27], both in resting state and in the context of the brain function. Previous research suggests their role in working memory, decision making [28,29], and processing of information [30–32] as they are often associated with a balance between global integration and local segregation in brain networks [33–36]. Complex behavior in dynamical systems corresponds to e.g. (i) multistable dynamics, which in the presence of varying input can lead to switching between different stable attractors, (ii) metastable dynamics (slowing down around attractor "ghosts" [30], before switching to other attractors’ neighborhoods in state space), or (iii) chaos, i.e., unstable but bounded dynamics. The presence of these complex dynamics hinges on a calibration of the model towards critical points, i.e., bifurcations or multistable regimes, where local input perturbations can cause a qualitative change in the model’s dynamics [37,38]. However, in the context of BNMs, the coupling weights are often characterized by heavy-tailed distributions [39–42] which leads to a high variability across regions in terms of their external inputs in a BNM. Therefore, controlling the dynamics of a BNM towards criticality or multistability is non-trivial, while being crucial for fitting procedures and model calibration. Experiments conducted in vivo, in vitro, and in silico have demonstrated that steering the system toward or away from criticality can be achieved by dynamic adjustments of the excitation-inhibition (E-I) balance [43–45]. From a modeling perspective, two closely related plasticity mechanisms have been proposed: non-local E-I balancing [46] and local feedback inhibition control (FIC) [16]. At its core, FIC is a mechanism to locally regulate inhibition in mean-field models [47] and enforce a target firing rate. Schirner and colleagues successfully used such synaptic plasticity-like learning rule, which dynamically adjusts E-I balance between proximal nodes in the context of decision making [46]. In the presence of FIC, this dynamic balancing can control synchrony between pairs of regions, to explicitly reproduce empirical FC patterns in the context of fluid-intelligence performance. Similar work was done with the Wilson-Cowan model to steer the system toward criticality [48–51]. Our study implements a homeodynamic plasticity mechanism (dFIC) in the Jansen-Rit (JR) neural mass model [52]. This versatile model generates slow- and fast-periodic oscillations in the δ- and α-bands, making it popular for studying network dynamics and fitting simulated time-series to empirical data including EEG, MEG, and resting-state fMRI signals [7,9,14,53–58]. Multistability is an inherent property of this model, as two fixed points (FPs), two limit cycles (LCs), or a FP and a LC, can coexist for specific parameter configurations [53]. We consider this property to be relevant for simulating blood-oxygen-level-dependent (BOLD) fMRI signals. In particular, BOLD FC dynamics (FCD) are characterized by the reoccurrence of FC networks (e.g., default-mode network) over time windows of tens of seconds, as opposed to the fast dynamics of many neural mass models which occur on a millisecond timescale [29]. Occasional, noise-induced switching between different attractors of these fast systems can give rise to the additional slower timescales observed in FCD. Our aim is to promote the above-mentioned complex dynamics through FIC. So far, the only known implementation of FIC in the JR model was proposed in the work of Coronel-Oliveros et al. [57], where a similar mechanism was used to constrain the activity of the JR model in the context of Alzheimer’s disease. In this work, we study the function and limitations of dFIC in the JR model. Our motivation is two-fold. First, to counteract heavy-tailed SC matrices and to compensate the over-excitation due to long-range coupling. This can be seen as decorrelation of large-scale network effects and local activity of BNMs. Second, to tune the activity and the range of traversed dynamical regimes of JR nodes in a network, regardless of the chosen method of structural connectivity normalization. In other words, long-range coupling and its strength can be chosen without impairing criticality and multistability, as well as without violating empirical constraints. The structure of the article is as follows: we first provide an analysis of the dFIC mechanism in the uncoupled JR model. Further, we demonstrate the dFIC mechanism in a small network of 4 nodes before moving to a whole-brain network using the average connectome from the Human Connectome Project Young Adult data set [58,59]. We analyze how the global dynamics of a large-scale brain network model are affected by locally regulating feedback inhibition. Through stochastic simulations and analytical methods, we highlight the impact of using dFIC focusing on specific activity targets and their consequences in parameter estimation procedures in a whole-brain simulation paradigm. We demonstrate the benefits of applying dFIC with a specific focus on increased switching between models’ attractors, both on the single-node and whole-network level. Finally, we discuss the process of fitting the simulated to empirical data using a custom fitness function, accounting for static and dynamical FC as well as for the overall level of synchronization in BOLD signals. Methods The following section outlines the dFIC mechanism and provides detailed descriptions of steps taken to introduce and set up dFIC in the JR model. First, we introduce the equations, state variables and parameters that were added to the original JR model and outline the workflow used to obtain the simulated data. Next, we describe the structural connectivity matrices used as a basis for our simulations. Finally, we list the used measures and statistical tools to describe and quantify the results. Equations of the JR model The JR model represents a local cortical circuit of three interconnected populations captured by six differential equations, two for pyramidal projection neurons Eqs (1A–1B), two for excitatory Eqs (1C–1D) and two for inhibitory interneurons Eqs (1E–1F), which form feedback loops. Neural populations are characterized by a second-order differential operator that converts the average incoming spike rate into the average membrane potential, whereas a non-linear, sigmoidal, function is applied to transform the average membrane potential into the average output spike rate. The state variables y0, y1 and y2 describe the outputs of pyramidal, excitatory, and inhibitory postsynaptic potential blocks (EPSP and IPSP) respectively. The postsynaptic potential (PSP) of the i-th pyramidal population is given by the difference PSPi = y1,i−y2,i. We implemented one additional state variable to the tuning model, namely wFICi, which describes the adaptive strength of the inhibitory input onto the pyramidal cells of region i in Eq (1B). The graphical representation of the model can be found in the Fig 1. The graphical representation of the JR model with the introduced heterogenous inhibitory scaling variable wFICi. The three populations Excitatory Interneurons, Pyramidal Cells and Inhibitory Interneurons are marked in orange, red and blue respectively. The sigmoidal is denoted with S and he(t) and hi(t) represent the impulse response resulting from the model equations Eqs (1A–1F). In the following equations, C denotes the SC matrix and Cij are the entries of that matrix, i.e., the connection weights between regions i and j. The symbol NROIs = 84 denotes the number of nodes in the connectome. Analogously, θij is the time delay and is computed from the tract lengths between regions i and j, as well as the conduction speed value, which was set at 5 m/s. Additionally, we introduce Iext,i—the total external current received by region i, consisting of a constant component μ and a term representing long-range excitatory coupling, which sums over the outgoing firing rates of all regions j, weighted by the corresponding SC element Cij Eq (3). The parameter values used for simulations are listed in Table 1 and based on physiological considerations in the original Jansen and Rit publication [52]. The amplitudes of the EPSP and IPSP are determined by A = 3.25 mV and B = 22 mV for the excitatory and inhibitory cells, respectively. The corresponding durations of EPSP and IPSP are inversely proportional to a = 0.1 ms–1 and b = 0.05 ms–1. As in the original work [52], the firing threshold is given by v0 = 6 mV. The remaining parameters of the sigmoidal transfer function in Eq (2) were set at vmax = 0.0025 s−1, e0 = 2.5 s–1 and r = 0.56 mV–1. The constant J = 135 corresponds to the number of the synaptic contacts, which is scaled for individual connections with c1, c2, c3 and c4. 10.1371/journal.pcbi.1012595.t001Table 1Parameters and their values.These values are fixed throughout this work, if not stated differently. EPSP: excitatory post-synaptic potential, IPSP: inhibitory post-synaptic potential, PC: pyramidal cells, EI: excitatory interneurons, II: inhibitory interneurons.SymbolDescriptionValue A Maximum amplitude EPSP3.25 mV B Maximum amplitude of IPSP22 mV a Reciprocal of excitatory time constant0.1 ms−1 b Reciprocal of inhibitory time constant0.05 ms−1 v 0 Firing threshold6 mV r Steepness of sigmoidal0.56 mV−1 c 1 Local coupling from PC to EI1 · 135 c 2 Local coupling from EI to PC0.8 · 135c3Local coupling from PC to II0.25 · 135 c 4 Local coupling from II to PC0.25 · 135 ν max Maximum firing rate of neural population0.0025 s−1 η Adaptation rate0.005 · 1/ms·(mV)2τdAdaptation window1000 ms Parameters and their values. These values are fixed throughout this work, if not stated differently. EPSP: excitatory post-synaptic potential, IPSP: inhibitory post-synaptic potential, PC: pyramidal cells, EI: excitatory interneurons, II: inhibitory interneurons. Regimes of the JR model The JR model exhibits a complex repertoire of regimes and bifurcation points that is dependent on its parameters depicted in Fig 2. The presence of bistability, where two stable states coexist within a specific parameter range, is one of the fundamental features of this model. In the case of the uncoupled JR model one of the regimes generates the ⍺-band (8−12 Hz) oscillatory activity–here referred to as fast LCs with frequencies in the range 10−12 Hz. Another regime gives rise to slower, δ-band (1−4 Hz) and θ-band (4−5 Hz) rhythm oscillations—referred to as slow LCs. In the following we will omit the region index i for the case G = 0, corresponding to an isolated JR model. Given a deterministic setting, as the input μ is increased from Iext = 0 to Iext<0.12, the postsynaptic potential increases steadily, and the model’s behavior is determined fixed points (non-oscillatory regime). This state configuration of the system can be described as sub-bistable and corresponds to 0.0070.12, the system starts exhibiting bistability between different regimes. The first bistability exists between the low- and high-activity fixed points (FPs) for Iext∈[−0.012, 0.09], the second one between the low activity FPs and the fast LC for Iext∈[0.09, 0.113], and the third one between the fast and slow LCs for Iext∈[0.113, 0.137]. Further increases to the input, Iext>0.137, place the system in the fast LC and Iext>0.31 leads to a high-activity FP (non-oscillatory regime), which we jointly refer to as super-bistable regime, where y0>0.1. Importantly, when coupling multiple JR nodes, the network effects can give rise to dynamics not captured by the bifurcation diagram of a single, uncoupled node. Nevertheless, the bifurcation diagram for the uncoupled node can approximate network effects under the assumption that each node receives a similar slowly varying “mean-field” input with small noise fluctuations, Therefore, the bifurcation diagram of a single node only approximates the behavior of a network node and may not be precise. Bifurcation diagrams of a single uncoupled JR model. A: State variable y0 of fixed points (FPs) and limit cycles (LCs) vs. constant external input Iext = μ = const. At low inputs Iext≈−0.1, one can find a branch of stable FPs (black line) which undergoes a series of bifurcations. The branch folds at Iext≈−0.1, where a saddle-node bifurcation (SN, black dot) destabilizes it (dashed black line). A second SN follows, without changing the stability. The two SN result in an S-shaped bifurcation structure. The upper branch stabilizes via a subcritical Hopf bifurcation near Iext≈0, which gives rise to unstable LCs (dashed orange line). The LC branch becomes stable (solid orange line) through a SN of LCs (purple dot) and terminates through a saddle-node on invariant circle bifurcation (SNIC, green cross). The upper stable FP branch destabilizes and stabilizes again through two supercritical Hopf bifurcations, between which one can find another family of LCs. B: Same bifurcation structure showing the postsynaptic potential PSP vs. Iext. The bifurcation diagram was computed numerically using AUTO-07p [60]. The parameter values are given in Table 1. Dynamic feedback inhibitory control Importantly, due to recurrent activation in the network case, the overall sum of incoming activity to a network node can increase significantly compared to the single-node scenario. The increase in the input to multiple single nodes results in an increase in the overall excitation in the system and an effectively reduced repertoire of available regimes that the individual nodes can achieve. Such a system exhibits behaviors associated with high-activity states, making it impossible to reach the regimes requiring weaker inputs. In the case of JR, such configuration places most nodes in the fast LC or, in extreme cases, near the high FP and leads to increased synchronization of the activity across nodes. Another consequence is the decreased variability of derived slow timescale signals such as BOLD, since they are computed from PSPs of low variability. dFIC was conceived as a solution that allows counteracting over-excitatory network effects by locally adjusting the inhibitory strength of nodes to reflect the overall strength of incoming connections from other nodes such that the long-term average output PSP activity of pyramidal cells of each node is fixed at a chosen level. Our solution splits the simulation process into two parts: the deterministic tuning process and the stochastic post-FIC simulations. In the first step, we implement the tuning process, which dynamically adjusts the weights of the wFICi state variables to bring the model’s activity to the desired level. To achieve this, we add two additional, activity-detection equations Eqs (4A–4B) and one equation describing the evolution of the wFICi variables Eq (4C), based on a gradient-descent learning rule proposed by Vogels [61], first implemented in large-scale whole-brain networks by Schirner et al. [17]. The activity-detection equations are used to extract a slowly varying average of the two state variables necessary for computing the learning process during the tuning step: y0,id and y2,id, for averaged PSP output of the pyramidal and inhibitory populations, respectively. Thus, fast oscillations and short-term variability of y0,i and y2,i are averaged out. We chose to include such equations to facilitate our analytics by making a fast- and slow-time-scale separation more explicit, as well as in order to disassociate the slow averaging dynamics from the particular form of the learning dynamics, and given that some slow averaging is in all cases necessary to filter out fast noisy fluctuations (in the case of FP targets) or oscillations (in the case of LC). The Eq (4A) provides a time-averaged slow detection of the ongoing firing rate as fitting target, which ensures that wFICi variable does not fluctuate or oscillate. The crucial parameter defining the target of the dFIC procedure is y0target, which is the desired value to which the long-term (i.e., at a slow time scale) average activity of the system is being brought to by the tuning process. This value must correspond to one of the attractors (FP or LC) of the uncoupled, original JR model (Fig 2). For point attractors, the value corresponds to their exact location, whereas for LCs the value corresponds to their centers (in practice approximated as the average activity value over several cycles). The η parameter describes the learning rate at which the wFICi variable is being adjusted during tuning to reach y0target. The time constant of the detection equations is denoted by τd. The parameters η and τd can be adjusted depending on the desired target dynamical regime of the model and simulation length. For example, η determines how quickly the gradient in the dFIC algorithm descends. If η is chosen too large, the values of the state variables wFICi may oscillate. Once the values of wFICi have converged, the tuning is completed and post-FIC simulations can be performed, for which wFICi is replaced by a constant parameter pFICi Eq (5A). pFICᵢ=1L∑tk 0.25, slightly above the average of the empirical FC matrix (FCmean = 0.23). This process helped us to reject biologically unplausible and over-synchronized FC data (S4 Fig). FCDs were computed using multiple FCs per simulation obtained by sliding window of size of 60 seconds and increment with 1 TR = 720 ms. The fit was quantified by the Kolmogorov-Smirnov distance (1−KSDFCD) between the distributions of empirical FCD values and each of the simulated FCDs. Additionally, we combined the fits of RFC, 1−KSDFCD, and a synchronization threshold into a single multimodal factor (MMF) which is described by Eq (6). The coefficients d1, d2 were set at 1, and 0.75, respectively. The selection of coefficients was motivated by the general idea that both RFC, and 1−KSDFCD should contribute to the final measure equally despite having different maximum values, with 0.6 and 0.8 respectively. The coefficient d2 was set at 0.75 to fulfil this requirement. We found this combined method useful for assessing the biological plausibility of our simulated data, but it is worth noting that the threshold value and coefficients are likely dataset specific and should not be used without numerical justification and testing. To confirm that no additional simulations are needed for a robust estimation of the MMF measure, we performed a bootstrapping statistical analysis, for each parameter combination (G, y0target) and for each condition (post-FIC and no-FIC simulations). To that end, we generated a bootstrapping sample distribution by resampling with replacement from an original set of–per parameter combination–MMF values, computed for overlapping windows of a shorter time length (19.2 min that we extracted from the original 30 min simulations. The results show that the MMF values we computed were robust already across such shorter time windows, for more details see S1 Appendix. Results Control of a single Jansen-Rit population In this section we present results on the control of a single JR node in presence of the homeodynamic inhibitory plasticity mechanism dFIC. We outline how dFIC achieves a desired activity level when the external input Iext = const., since this is a prerequisite for the functioning of dFIC in a network setup, where Iext is time dependent. The following semi-analytical methodology is based on numerical bifurcation analysis and slow-fast dissection [71,72]. To start with, we note that the full system for an isolated node, described in Eqs (1),(2) and Eq (4), can possess different timescales, depending on the parameter values. The JR dynamics are governed by the time constants 1/a = 10 ms and 1/b = 20 ms. Typical values for the detection time constant τd and the learning rate η used in this work are τd = 1000 ms and 1/η = 400 (mV)2 ms. With this configuration the full system has three timescales. For simplicity however, we assume a two-timescale system, by posing that τd and 1/η are large enough, such that (i) y0d and y2d converge to long-term temporal averages and (ii) changes in wFIC are slow, giving time to the JR state variables to converge to the proximity of attractors present at constant wFIC. This assumption allows us to apply slow-fast dissection, to understand the dynamics of the full system (i.e., single node with dFIC), which emerges from the dynamics of the no-FIC system through a slow variation of the wFIC. In the limit case 1/τd→0, η→0, corresponding to no-FIC, wFIC becomes a parameter in the remaining equations Eqs (1) and (2) y0d and y2d become the long-term temporal averages of y0 and y2, respectively. To understand the dynamics in this limit case, we average the LCs of the original JR model (Fig 2) over one period, as shown in Fig 3A for wFIC = 1. Bifurcation structure of the uncoupled JR model with averaged limit cycles (LCs). A: Average y0d versus external current Iext for wFIC = 1. The solid (dashed) black lines correspond to families of stable (unstable) fixed points (FPs). For families of stable (unstable) LCs, the average value of y0 over one cycle is computed and depicted as solid (dashed) orange lines. B: Bifurcation structure in wFIC‐Iext‐y0d space. The bifurcation structure of panel A is depicted at wFIC = 1. When computed on an interval of wFIC, stable branches of this structure form an attracting surface with a lower and an upper sheet corresponding to families of stable FPs (gray surface) and LCs (yellow surface). The upper boundaries of the stable LC surface are given by a family of Hopf bifurcations (blue line) and saddle-node bifurcations of LCs (purple line). The bifurcation diagrams were computed numerically using AUTO-07p [60] and the parameter values given in Table 1. Notably, wFIC = 1 yields the original bifurcation landscape of the JR model, whereas values wFIC≠1 reshape this structure. To capture this alteration, one can compute y0d vs. Iext bifurcation diagrams for an interval of wFIC values and obtain surfaces of FPs and LCs, as shown in Fig 3B. Initial conditions and a given point (Iext, wFIC) in the parameter plane will determine which of the attractors of the JR model the dynamics will approach. At the same time, y0d and y2d will approach the corresponding temporal average of that attractor. This attractor, be it a FP or a LC, will not satisfy y0d=y0target, unless (Iext, wFIC) is chosen appropriately. In other words, the average activity is usually lower or higher than the desired target. In the case of a small η>0, wFIC is a subject to a slow drift, proportional to y0d−y0target. An increase of wFIC will cause a decrease in y0d and vice versa. Therefore, if y0d>y0target, wFIC slowly drift towards larger values, causing a decrease of y0d, whereas for y0d0 holds in the parameter intervals considered in this work. A linear stability analysis at the above equilibrium yields dgdwFIC=ηy2ddy0ddwFIC. Therefore, if dy0ddwFIC<0 is fulfilled at the dFIC equilibrium, it will be a stable equilibrium. Given the JR model with the implementation of dFIC as an inhibitory plasticity mechanism, this condition—by construction—is always fulfilled. To sum up, a dFIC equilibrium exists at a given Iext if a stable FP or LC with y0d=y0target at this exact Iext exists and it follows automatically that this equilibrium is stable. In the following, we provide simulation results to support the above analysis, using three representative examples shown in Fig 4. They also give insight into the functioning of dFIC and allow to relate wFIC dynamics to the JR bifurcation structure. In the first case, shown in Fig 4A1, 4B1 and 4C1, we set Iext = 0.2, for which the original JR model only possesses stable LCs with activity levels beyond the desired y0target=0.01. Consequently, in the absence of dFIC, the system converges to this LC. The trajectory with active dFIC however, exhibits a drift in wFIC, forcing the activity level to the desired target, which is a fixed point of the JR model. The second case in Fig 4A2, 4B2 and 4C2, on the other hand aims to reach a LC at a target of y0target=0.1. In the absence of dFIC, the system goes beyond this target and converges to a fixed point on the upper FP sheet. With dFIC, the activity is reduced and forced onto a fast LC of the JR model. In the last case, Fig 4A3, 4B3 and 4C3, the activity is increased from the fast LC to a larger level, corresponding to a fixed point on the upper sheet. Comparison of no-FIC (η = 0) and dynamic FIC (η>0) of a single Jansen-Rit node. A, B: y0d(t) and wFIC(t) vs. time t, respectively, obtained by direct simulation of Eqs (1), (2) and Eq (4). Each column corresponds to different values of y0target and Iext. C: Bifurcation structure of the uncoupled Jansen-Rit model with averaging in wFIC‐Iext‐y0d space. In C, y0target is marked by a red plane and the intersection of this plane with stable equilibria or limit cycles is shown as a red line and marks dFIC equilibria. Results obtained from direct simulation of the no-FIC (η = 0) and dFIC (η>0) are superimposed. Circles mark the start and asterisks the end of these trajectories. The no-FIC and dFIC initial conditions are shifted slightly for visual clarity. The parameter values for the direct simulations are τd = 400 ms and 1η=400ms×(mV)2, other parameter values are given in Table 1. The bifurcation diagrams were computed numerically using AUTO-07p [60]. Having dFIC equilibria over a wide span of Iext values is desirable: first, the external current is often a free parameter in neural mass models and second, Iext will be dynamic and subject to fluctuations in a network setup, as we discuss in the next section. Hence, we look for intersections of attractors, depicted in Figs 3B and 4 with the plane y0d=y0target. These intersections are shown in Fig 4 as red lines and represent the set of (Iext, wFIC) pairs, for which y0d=y0target and dFIC is in equilibrium. For the examples in Fig 4, this set is a continuous line. However, the surface of attractors exhibits gaps in the interval y0d∈[0.02,0.095]. Hence, targets in this interval might lead to a corresponding gap in the set of (Iext, wFIC) values for which dFIC equilibria exist. In other words, dFIC can fail for targets in this range if μ is not chosen appropriately. Outside y0d∈[0.02,0.095] on the other hand, dFIC will function as intended and control the system towards the desired target, independent of Iext. In summary, the correct application of the proposed homeodynamic inhibitory plasticity mechanism for a single JR node requires the existence of an attractor located at the desired y0target value (for LCs, this location is defined through their center), for a given constant Iext. The attractor does not need to be present at this location for the original JR model (wFIC = 1) but can be found through variation of wFIC. dFIC in a brain network model Moving on to the network case, Iext,i now contains the network input, is time-dependent and spans over a range of values, which is determined by the network activity and indegree di=∑j=1NROIsCij of each node i. At the same time, wFIC is now a vector with NROIs entries wFICi, one for each brain region. An analysis as above is considerably more intricate and beyond the scope of this work. Nevertheless, we want to point out how the above requirement for successful control through dFIC can be translated to a network setup, by making some assumptions. If the network connections are not strong enough to induce fast transients or overly synchronized oscillations, Iext,i can be separated into three components, Iext,i(t)=μ+diR(t)+ξi(t), (9) namely the constant offset μ, a slowly varying indegree-dependent mean-field input diR and small fluctuations ξi, which arise from the network and are distinct to each node. If we neglect the last contribution, the constant μ remains and shifts the entire network towards larger external currents. The mean-field input diR on the other hand introduces a spread of the network along Iext, depending on the indegree distribution. Given that R(t) is subject to only slow changes, the above slow-fast analysis for the single JR model holds valid, since R can be treated in the limit case, just as wFIC before. Hence, the behavior of each node in the network can be approximated by the single-node bifurcation structure present in the plane (Iext, wFIC). The major difference to the single-node case is the distribution of nodes along Iext. As shown for the single node, outside the interval [0.02, 0.095] of the y0target parameter, i.e., the gap in the attractor surfaces (Figs 3B and 4), we find a continuum of Iext values, for which dFIC convergence is guaranteed. In the network case and with the assumption of weak coupling, this continuum of Iext will allow the individual node to approach the dFIC equilibrium, no matter its precise location on the Iext-axis. Within the gap on the other hand, there can be nodes with Iext values for which no dFIC equilibrium exists. In other words, in the case of the JR model, dFIC equilibrium on the network level can fail for y0target∈[0.02,0.095]. In the upcoming sections, we will provide numerical evidence for the functioning of dFIC in different network setups. We assume the success criterion of tuning parameter convergence to be 1% of the distance of the mean of the last L = 5 seconds of the y0 time-series from the y0target. Before convergence of wFICi one can encounter damped oscillations towards the dFIC equilibrium. Hence, we choose initial conditions based on the bifurcation structure to reduce these transients and obtain faster convergence of wFICi. When the tuning was successful, we averaged wFICi over last L seconds to obtain a single pFICi parameter per node in the brain network. We present the simulation results in two separate parts, one for a small network, used to explore the dFIC performance for different values of the time constant and learning rate parameters, τd and η, respectively, and then, for a biologically realistic whole-brain network. Small network simulations The results of η and τd parameter explorations demonstrate the convergence of the tuning algorithm (Fig 5) for y0target=0.01 for all and τd values in the cases of η = 0.01 and 0.005. This means that all single-node wFICi state variables converged asymptotically to a stable value necessary for the system to exhibit the desired activity level with the accuracy of 1%. The tuning to a y0target=0.103 was successful across all conditions except for short τd = 10ms and a few η values for G = 10, G = 25, and G = 100 as depicted on Fig 5. Based on these results, τd = 1000, η = 0.005 were chosen as parameters for further for simulations as they proved to be robust across both y0target and G values. We have also tested a faster solution with τd = 100, η = 0.0025, which, as expected, yielded almost identical results across both y0target and all G values. Fig 6 demonstrates the resulting pFICi values as functions of G and indegree, showing a monotonously increasing—albeit nonlinear—relation between the respective values. Analysis of η and time window (τd) parameter explorations for y0target=0.01 (A) and y0target=0.103 (B) under different G values (columns). Successful tuning (y0 mean within 1% of y0target) is marked by yellow crosses. The colors of the squares indicate the percentage of the distance of the y0 state variable activity from the y0target value after tuning. In the two cases of y0target = 0.01 and y0target = 0.103, the evolution of the wFIC variable was on the correct trajectory but the tuning required a longer simulation to converge. Small-network differences in inhibitory control pFIC parameters per node obtained from the tuning procedure for different y0target values (A: y0target = 0.01, B: y0target = 0.103) and G values with the color intensity marking the indegree of a given node. Plot shows the relationship between G values, pFIC values with color-coded indegrees of the nodes of the small connectome. Two different target values were chosen to correspond to sub-bistable and super-bistable JR configurations. Due to the additive impact of noise in post-FIC simulations we chose the y0target = 0.01 further to the left from the saddle node bifurcation and y0target = 0.103 directly at the edge of saddle node bifurcation between the LCs (Fig 2A). Afterwards, we run the post-FIC simulations to assess the impact of tuning in the small network. The time-series coming from simulations with G = 1, 10, 25 demonstrate the change in activity levels caused by the introduction of the pFIC parameter (Fig 7). It becomes evident that noise and network effects cause an offset between the long-term average activity level and the y0target for two reasons: (a) predominantly—due to stochastic switches between attractors, which can lead to a net-positive (negative) effect for the sub-bistable (super-bistable) case (respectively), and (b) due to the (sigmoidal) nonlinearity of the model’s activation function, which leads always to a net-positive offset, even in the absence of any switches. As a reminder, here we consider sub-bistable targets to be 0.007 1. For G = 10, the deviation from the target is visible, which can be explained by the noise induced switching between two dynamical regimes of the system. Note that for y0target = 0.01 the y0 average is higher than the target due to switches between FP and slow LC and for the y0target=0.103 the effect is opposite, due to switches between dominant fast LC and slow LC. Whole-brain post-FIC results In the second part of the results, we focus on the comparison between post-FIC and no-FIC simulations on the whole-brain network. The resulting PSP time-series show the desired effect of tuning, which, as expected, restricts the baseline activity of each of the nodes to the y0target value. The difference between the y0target and the mean post-FIC y0 values is caused by the introduction of noise as well as the ability of noise to cause nodes to periodically jump between the regimes of the JR model. The time-series of the state variable y0 show a predictable deviation from the target activity level, proportional to both the sum of weights and the chosen G parameter value, further highlighting the fact that the dFIC for y0target = 0.007 and y0target<0.11 places the system in the proximity of multi-stability. The switches between different regimes are less visible for the respective no-FIC simulations (S3 Fig), a phenomenon further investigated in the next section. We have computed and compared power spectra for the best fitting post-FIC simulations (using the Welch method), and best fitting no-FIC simulations. The power spectra analysis shows a significant post-FIC increase in power in the respective ⍺ and θ frequency bands when compared with the corresponding no-FIC simulations (Fig 8). Increases for y0target=0.007 occur predominantly in the θ band as the tuning target places the system near the low FP (Fig 8A), where network fluctuations and noise move the system towards the bistable regime between the LCs (to the right on the bifurcation diagram). The presence of multiple peaks in the power spectra, corresponding mainly to θ and slow ⍺ rhythms (<10 Hz), suggests that it is bistable switching and not the slow LC that explains the increased theta power. Similarly, the y0target=0.01 is a critical point on the fast LC exhibiting alpha oscillations (Fig 8B), where changes in input and noise cause the periodical switches between both alpha and theta LCs. We conclude that the dFIC mechanism is responsible for the increase in power in the respective rhythms. Further support of this claim comes from the Poincaré map analysis. The power spectra of all nodes for the two best post-FIC simulations for two different fit measures y0target values (panel A: FC-fit and panel B: MMF fit, blue) and two best fitting no-FIC simulations (red). dFIC Improves the fit to empirical data The results of the fitting based on RFC are shown in Fig 9 and example matrices are available in the S4 Fig. The results indicate that dFIC improves fits between the simulated and empirical FC data as long as y0target≤0.11 allows the system to reach multiple dynamical regimes. When y0target>0.11, all nodes are forced by the pFICi parameter to remain on the fast LC and the repertoire of possible behaviors is significantly diminished. We assumed that the 30 min simulation for BOLD signals approximates well the stationary value of the FC statistic for a given parameter set (including the amplitude of noise). Results of functional connectivity fitting. A-B. The heatmap with the resulting fit measure for all post-FIC and the best fitting simulations without FIC (no-FIC). The red dots mark the best-fitting G parameter per y0target, white square marks the best fitting simulation overall. C-D. Plots C-D illustrate the direct comparison of Pearson-R values for best fitting simulation per y0target C: for sub-bistable and D for super-bistable target values. Max R.: synchronization-adjusted Person correlation coefficient between simulated and empirical data. FCD fitting results show an overall improvement of fit for most sub-bistable targets and significantly higher best similarity score (1−KSDFCD) for post-FIC simulations compared with no-FIC settings (Fig 10). Heatmap of Kolmogorov-Smirnov-Distance-based similarity measure (1-KSD) for both post-FIC (A) and no-FIC (B) simulations. The optima are in a very similar range, but they differ slightly suggesting that this measure captures different (dynamical) properties of the BOLD dynamics. We have combined the results from synchrony-adjusted FC and FCD fitting into a one Multimodal Factor fitness function (MMF). The resulting y0target optima have largely remained similar, but the global optimum shifted to y0target=0.01, G = 11 (Fig 11). Interestingly, the FCD distribution of this simulation visually fits the empirical FCD better than the best qualitative 1−KSDFCD fit (Fig 12B compared with Fig 12C) as the best 1−KSDFCD only fit contained a second peak. Heatmap of Multimodal Factor fit based for both post-FIC (top) and no-FIC (bottom) simulations. The result that MMF maximizes not directly at the critical point of saddle-node bifurcation (y0target≈0.0189) but at y0target=0.01 can be explained by two factors: (1) we sampled the y0target space more densely for the sub-bistable regimes than for the super-bistable and (2) ones, as we suspected this range to be optimal, second: the network effects and noise have primarily additive effect, which causes a shift in effective network average of Iext,i to the right of the y0target value. Example of FCD matrices and their distributions. A: shows the FCD data for the best fitting post-FIC simulation chosen solely based on the Pearson correlation with empirical FC. B: for the generic empirical FCD data, C: for the combined MMF fit, D. shows the FCD data for the best fitting no-FIC simulation based on the combined 1−KSDFCD fit. On all the plots average empirical FCD distribution is plotted in black. The FCD matrices were threshold at 96th percentile for clarity. The plot of all the MMF fits in Fig 13 shows the wide range of good fits for a wide range of sub-bistable y0target∈[0.007,0.0189] and G values and for super-bistable y0target=0.1. The difference between the wider sub-bistable y0target range and a single super-bistable value can be attributed to two factors: first, the denser sampling of the former parameter space and the noise, since the offset it induces via network effects is mainly net-positive for the sub-bistable regime due to switches from the FP to the LC. Comparison of best fitting results for different y0targets. A: Sub-bistable targets. B: Super-bistable targets. These plots show the viability of the Multimodal Factor (MMF) fitting approach and the positive effect of the dFIC on the range of sub-bistable cases and directly super-bistable case of y0target=0.1. To test if differential MMF fitting results between the post-FIC and no-FIC simulations are statistically significant, we conducted a permutation statistical analysis by sampling without replacement from the same set of MMF values computed for the above shorter time windows, but this time randomly mixing the two conditions of post-FIC and no-FIC simulations. Statistical analysis showed that the difference between post-FIC and no-FIC simulations was significant at a level of p < 10−6 for almost all parameter combinations. Details of both procedures are further described in the S1 Appendix. dFIC intensifies “switching” between dynamical regimes To investigate the switching behavior of the model, we computed Poincaré maps of synaptic potential time-series. This step allowed us to highlight and quantify the effect of the dFIC on the model’s underlying activity. First, we have identified a threshold of activity (c = 6 mV) which can be applied to the resulting maps to identify the regime of the JR model being exhibited by any node at any given time. More in detail, the lower activity FP of JR in presence of noise, yields points (PSPi,k, PSPi,k+1) in [0 mV, 6 mV]×[0 mV, 6 mV], the fast LC in [6 mV, 18 mV]×[6 mV, 18 mV], whereas one period of the slow LC will have one point in [0 mV, 6 mV]×[6 mV, 18 mV] and one in [6 mV, 0 mV]×[0 mV, 6 mV], due to the formation of two maxima within one period. If a node traverses multiple regimes however, the point cloud of a single node is going to fulfill multiple of those conditions, allowing to identify which regimes were visited and therefore which dynamics exhibited. By analysis of the pattern formed by individual nodes over time and the pattern of all nodes together, it is possible to extend the understanding of the overall behavior of the model. Further we have calculated how many nodes exhibit behaviors in one, two or three different regimes for more than 7,5% of the simulation time Tp, as a measure of complexity of the resulting signals. For most of targets in the proximity to critical points of the JR model (0.007≤y0target≤0.1), Poincaré analysis shows higher variability of exhibited regimes. That means, more regimes can be reached by any single node and a higher number of nodes are ‘visiting’ a higher number of regimes in post-FIC simulations compared to their no-FIC counterparts. The dFIC algorithm for those targets increases in overall switching behavior for a high number of tuning targets. In contrast for some low G values and y0target ∈ {0.004, 0.007, 0.01} dFIC is binding the activity of the nodes to a single regime. Expectedly, for y0target≤0.004 and y0target>0.11, dFIC does not significantly impact the switching behavior. Fig 14 shows the y0target and G values for which FIC caused an increase in the number of showcased behaviors. We demonstrate this effect based on two examples: Fig 15 shows best fitting no-FIC (y0target=0.007,G=9) compared with best fitting post-FIC (y0target=0.01,G=11), where tuning increases this number significantly (Fig 15A1 and 15A2). This effect can also be observed when the y0target is super-bistable, in which case the no-FIC method is compared with the post-FIC one, both with y0target=0.1andG=11 (Fig 15B1 and 15B2). Summary heatmap of increases in number of regimes traversed by the nodes. If the number of regimes of post-FIC simulations is greater than for no-FIC simulations the map is marked in yellow, in the opposite case in black, and in yellow if there is no difference. The tuning targets that correspond to a single attractor, with large distance to any bistability show lack of improvement with dFIC or decrease in number of nodes in multiple exhibited regimes. In contrast the targets where the system is tuned to a sub-bistable regime and directly super-bistable regime show increases in that measure. Inc: increased, Dec: decreased, NC: no change. Poincaré maps for the best no-FIC (A1) and post-FIC (B1) MMF fit. Each point (PSPi,k, PSPi,k+1) maps one PSP maximum PSPi,k of an individual node i to the next one PSPi,k+1. The node indegrees are color-coded (low indegree: dark blue, high indegree: yellow). The set threshold at c = 6 mV consistently (visually) separates the FPs and LCs of the JR model. Bar plots at the bottom right corner of each subplot show the percentage of nodes exhibiting 1,2 or 3 regimes during at least 7,5% of the time of the simulation, showing the effect of dFIC. The A2 and B2 panels compare no-FIC with post-FIC simulations for super-bistable settings (y0target=0.1,G=11). The node indegrees are color-coded low indegree: dark blue, high indegree: yellow. The set threshold at c = 6 mV consistently (visually) separates the FPs and LCs of the JR model. Both plots demonstrate a difference in number of regimes traversed by each (map) of the nodes during the simulation. A decrease in number of nodes in a single regime or an increase in the number of nodes in two or three regimes is considered a richer repertoire of behaviors as a result of implementing dFIC. Overall, Poincaré mapping analysis for the best fitting simulations according to the MMF showed the positive impact of dFIC on the number of nodes traversing 2 and 3 regimes (Fig 14). The overlap between high values of fits of RFC, 1−KSDFCD, and MMF in the (y0target,G) parameter plane and the results of our Poincaré mapping analysis suggest a strong link between the overall quality of fits between the simulated and empirical BOLD data and the switching behavior in the PSP time-series. This effect is visible while comparing the heatmaps (Fig 16), suggesting a similar range of potentially optimal JR and dFIC parameters. Our results provide further evidence that in simulations with JR model the switches between different regimes leading to different amplitudes and frequencies of oscillations in the underlying PSP data result in more variable and hence more biologically plausible BOLD signals and BOLD-derived FC matrices. The heatmaps showing FC, FCD, MMF fits and the heatmap of increase switching behavior in underlying PSP activity as shown by Poincare analysis. Heatmaps for all 3 fitting measures (A-C) show a similar range of good fit parameter values spanning from y0target=0.007 to 0.1. and for wide range of G values. Interestingly a very similar range shows an increased number of switches during post-FIC simulations in PSPs (D). Discussion The work presented in this paper addresses issues of over-excitation and large divergence in individual node activity arising from coupling multiple nodes of neural mass models in a whole-brain network with heavy-tailed weight distributions (39,41,57). The goal was to provide a solution that (i) does not involve changing the underlying structural connectivity (ii) allows for decoupling the effects of global coupling on correlations among regions and excitability for each single region, (iii) permits direct analysis of adaptation dynamics. Established methods of counteracting this issue include: transformation of the connectome weights (i.e. log-scale transformations) [73,74], fitting a G parameter to reduce the relative strength of all network effects on a single node, and various implementations of inhibitory plasticity mechanisms [16,46,50,51]. Since those methods did not fulfill all three of those conditions, we proposed dynamic Feedback Inhibition Control (dFIC) as a solution. We introduced a previously established FIC mechanism [16,50,57,59] into the differential equations of the model and provided an analytical treatment of the adaptation procedure. We determined the conditions under which dFIC control is successful in the JR model, as well as the practical implications of using dFIC in terms of fitting data and the dynamical repertoire of the regimes of the model. We analyzed the case of an isolated node during tuning and established limits for tuning targets. To the best of our knowledge, this was the first attempt at providing analytical results on the functioning of a homeodynamic synaptic plasticity mechanism in a BNM. We demonstrated that—for a wide range of G values—we can tune the values of the wFICi such that the deterministic system is kept at a predetermined y0target value. The presence of a stable attractor associated with the target constitutes the main constraint of the y0target selection. Using small network simulations, we have studied the interplay between τd, η, and wFICi to converge and performed small network post-FIC simulations to assess the effects of applying dFIC in a network. Further, we have analyzed the effects of dFIC in a whole-brain BNM setup and compared the simulation results with the original JR model, including PSP data analysis and both static and dynamic BOLD time-series analysis. There are several important outtakes from our results. First, our work shows that dFIC can be used to regulate the dynamical regimes of individual nodes in the whole-brain network. Based on the single-node bifurcation diagram of the model (Fig 1), one can choose a y0target to determine which regimes of the model will be traversed when the system is coupled, and noise added. The noise levels need to be within certain limits to be investigated either computationally or via an analysis of the resulting stochastic system. Too low noise will likely not lead to any switches between attractors, whereas too high noise will lead to dynamics hardly determined by the specific structure of the system. In network simulation results, we have observed three outcomes of dFIC depending on the target: (i) dynamics of a node reflect a regime at set target, (ii) dynamics correspond to a regime at the set target with a small positive offset, and (iii) dynamics correspond to a regime at the set target with a small negative offset. The offsets can be attributed to network effects, noise, and nonlinearities. In the JR model, this feature allows for the selection of the frequencies of interest, and control over the overall variability in the system, which was not the case in known-to-us implementations of JR model coupled in a BNM. As an example, if y0target corresponds to an oscillatory regime, dFIC results in the increase in power in the associated dominant ⍺ and θ frequencies in obtained signals (compared to the established JR implementations [52,54–56]. Second, BOLD signals generated in presence of dFIC capture characteristics of empirical data better over a broad parameter region than without dFIC. The complexity of the PSP time-series, assessed through Poincaré maps, increases in presence of dFIC: the individual nodes of the BNM traverse more dynamical regimes. In particular, our findings show that when the tuning target is set near criticality (i.e. either in proximity to a saddle node bifurcation 0.0070.11 values, yielded worse fits than the respective no-FIC simulations, due to all regions being constrained to the fast LC, providing additional argument for the role of dynamical variability in BOLD fitting. Lastly, by exploring several tuning targets and G values, we have showcased our solution’s ability to decouple the G parameter’s influence on inter-regional correlations from the overall excitability level, in line with the FIC tuning results by Deco et al. [16]. This has an interesting consequence, as one can a priori decide on the activity target of interest, and, thus, determine the range of available dynamical regimes of the system. This constitutes a more informed starting point for the exploration of the effect of other model parameters or of stimulation. dFIC, E-I balance and criticality The balance of excitation and inhibition, postulated as a central property of a healthy neural system, has had a few explicit implementations in research dedicated to brain simulations [8,46,51,75]. Studies show the significance of homeodynamic plasticity in the context of maintaining E-I balance and simulating healthy and more realistic neuronal dynamics, but the connection to the notion of criticality remains debated [25,76,77]. Both concepts have been argued to require some sort of plasticity mechanism [33,36,49,78,79]. It is worth noting that several mechanisms collectively contribute to the brain’s ability to self-regulate E-I balance through synaptic plasticity [80,81]. Recent studies suggest the roles of short-term [82] and long-term synaptic plasticity [83], and structural plasticity [84] in maintaining this balance. Additionally, there are cellular mechanisms distinct from plasticity, but with capability of self-regulation. As an example, spike-frequency adaption, i.e., the property of certain types neurons to reduce their firing frequency upon sustained input, could effectively act as a negative feedback similar to feedback inhibition [85]. Overall, mechanisms like dFIC are implementations of generalised inhibitory synaptic plasticity mechanisms in the BNMs. However, their effectiveness in maintaining E-I balance in simulated networks does not allow for determining the specific biological underpinnings of such mechanisms. Even though, the E-I balance does not necessarily entail criticality in the context of BNM [86], there has been evidence suggesting that E-I balance, and optimal information processing based on criticality are two tightly related phenomena [24,31,87,88]. For example, researchers have postulated that E-I balance is directly responsible for bringing the cortical activity to criticality in a rat’s visual cortex in vivo [89]. Their research strongly suggests that homeostatic plasticity, linked to E-I balance in general, is responsible for bringing the cortex to the critical state. A recent review on inhibitory plasticity mechanisms points out that E-I balance in neural networks enables complex, non-linear collective behavior [75]. Furthermore, E-I balance can be a necessary condition for more complex dynamics on a whole network scale. Proponents of E-I balance have argued for the role of inhibitory plasticity mechanism in keeping the BNM at criticality in the Deco-Wang model and Wilson-Cowan model, respectively [18,49]. They have provided evidence for the role of such a process suggesting that the brain uses such mechanisms to stay in a state that prevents runaway activity such as epileptic seizures and enriches the dynamics (thanks to proximity to criticality), while staying relatively immune to the strength of noise or global coupling. Interestingly, their results, similar to ours, showed improved correspondence between simulated and empirical functional resting-state fMRI connectivity compared to models without inhibitory plasticity mechanism. Our findings support the idea that E-I balance plays an important role in maintaining rich and biologically plausible dynamics in BNMs. Depending on the y0target, dFIC positively affects the desired characteristics of simulated neural signals such as variability and richness of behavior. Here, we have provided experimental arguments for choosing the sub-bistable or directly super-bistable targets in the context of the JR model. However, our solution delineated limits and conditions for a wider range of tuning targets including the low-activity or high-activity states, which in future research can be valid tuning targets depending on the specific research question. The mechanisms related to inhibitory control, similar to dFIC, have potential relevance to modeling E-I balance also in the clinical context. Parameters responsible for the balance between excitation and inhibition have been linked to abnormalities in Alzheimer’s Disease, such as synaptic degeneration, hyperexcitation, white matter atrophy, as well as amyloid-beta and tau levels. In previous studies, inhibitory parameters of BNM were fit to empirical functional data of patients [9,90]. Recently, the JR model has been used to derive biomarkers related to inhibitory control, emphasizing the potential of model-based approaches in the early diagnosis of dementia and Alzheimer’s disease [57,91]. Similarly, the role of inhibitory control in the progression of dementia has been studied using the Wilson-Cowan model [92] and the Deco’s model from 2018 [93,94]. As we have shown, dFIC improves the overall model fitting, facilitating the study of how different dynamical regimes and transitions between them may be linked to various pathological conditions. Additionally, the resulting inhibitory plasticity parameters, can be interpreted as measures of dynamical differences in clinical setting. To sum up, dFIC is designed to maintain the activity of all nodes at a certain level, by adjusting local inhibition and, in that manner, maintain E-I balance at both local and—indirectly—at the global scale regardless of the chosen target. Our exploration of different activity targets suggests that tuning the system to the proximity of bistability results in increased switching between regimes and better fits to the empirical data. dFIC provides improvement in characteristics observed in the empirical signals. Previous works have treated FIC as an iterative process of simulating, computing activity levels, adjusting weights and then simulating again [8,16,17]. In this work, we embed dynamic FIC as an extension into the JR model, similar to Santos et al. in Wilson-Cowan model [51]. The dFIC equations lead to a slow, asymptotic convergence towards a target average activity of the JR nodes in the whole-brain network model, under conditions relatively easy to be met in most cases (of neural mass models and target dynamical regimes thereof). The analytical treatment of the single-node tuning dynamics reveals the conditions under which convergence towards a target average activity can be guaranteed and provides guidelines for employing dFIC in practice. Limitations We have implemented the dFIC mechanism in the JR model, because it is widely used in BNM studies as well as because it is complex enough to allow for multiple dynamical regimes and possible bifurcations among them. However, our results are applicable to other models which have a control quantity (here y0d) which monotonously decreases with an increasing control parameter (here wFIC). One can assume that a dFIC mechanism, potentially implemented in other neural mass models, would possess this property, since increased inhibition should naturally lead to reduced activity. Beyond this, the stability of the dFIC dynamics only depends on the presence of stable fixed points and LCs (in the uncontrolled model), which are located at the desired target activity. Further analysis is needed to assess how dFIC would change a system’s behavior which possesses chaotic attractors. Therefore, applying dFIC to other neural mass models should be straightforward, a bifurcation analysis being a prerequisite. Future work can take up this task and provide examples of dFIC application to other neural mass models implemented in TVB and frequently used in BNM studies. In general, although our results assume deterministic dynamics during the dFIC tuning process, we expect them to remain valid in the presence of relatively low additive, white noise, which is generally used in stochastic simulations of BNMs, as well as for our post-FIC simulations. Such a stochastic version of dFIC could also resolve the problem of a possible offset between the y0target and the average y0 value resulting from stochastic post-FIC simulations, which was observed in our study. Alternatively, tuning in presence of noise could be achieved by running a short stochastic dFIC simulation right after the deterministic dFIC has converged, by using the result of the latter as the initial condition of the former. Our work utilized fitting static and dynamic characteristics of the simulated BOLD signals to assess their ‘biological plausibility’, including FC, FCD and synchrony. However, it did not extend to the high-temporal resolution signals such as MEG and EEG. Even though, we presented the basic properties of the generated PSP time-series (with and without dFIC), in the frequency domain and via Poincare Maps (Figs 8,14 and 15), we did not extend our fitting procedures to include those other signal modalities. Such multimodal fitting can help in the future determine the relationship between different modalities and further improve both the fitting methods and the homeodynamic fitting procedures. Further research can also investigate dFIC as a homeodynamic mechanism, possibly relying on inhibitory plasticity, which maintains the state of the brain within limits and responds to external perturbations induced by e.g., cognitive stimuli. This kind of slow timescale modulation of brain dynamics could be of great interest in understanding data, e.g., on FCD dynamics of BOLD signals. In this context it is important to note that we based our simulation on average empirical structural and functional data: average SC, average FC, and average distribution of FCDs were derived from a large HCP dataset. Hence, when considering data of individual subjects, dFIC might yield slightly different results. The overall findings suggesting sub-bistable and directly super-bistable targets as optimal should in the future be tested using the individual subject data, both for normative and clinical cases. Conclusion We demonstrated how dFIC can be used to regulate network activity at the single-node level, stabilizing it at a predetermined target value. The analytical treatment of the single node tuning allowed us to determine the ranges of possible activity-level targets. The comparison of post-FIC simulation results with the JR model without such mechanism, both shows the practical value of using the proposed mechanism and provides supporting arguments for extending similar solutions to other models used for whole-brain simulations. We intended dFIC to be a useful addition to the widely used JR model, which can improve the control over networks of JR nodes by addressing the issue of over-excitation. We kept the proposed solution flexible–and easy to extend to other neural mass models—such that future users can test it further and decide how to set up a BNM depending on their own research goals, and independent of their decisions regarding the changes to the SC, or the intended effect of the global coupling. Supporting information Example Poincaré map of the PSP from a single simulation. Each point (PSPi,k, PSPi,k+1) maps one PSP maximum PSPi,k of an individual node i to the next one PSPi,k+1. The node indegrees are color-coded low indegree: dark blue, high indegree: yellow. The set threshold at c = 6 mV consistently (visually) separates the FPs and LCs of the JR model. (TIF) Schematic of the dynamic Feedback Inhibition Control procedure. Steps: (i) μ and y0target selection, (ii) deterministic tuning resulting in wFICi vector (which needs to be checked for convergence), (iii) the last 5 seconds of the vector need to be averaged per node to obtain pFICi parameter, (iv) stochastic (optional) post-FIC simulation, (viii) simulation output. (TIF) Comparison of the mean of whole-brain time-series with dFIC (post-FIC) vs. without dFIC (no-FIC). For A: y0target=0.01 and B: y0target=0.1. In both cases, the difference in the baseline of the displayed timeseries and illustrates the desired effect of tuning: In case A the dFIC limits the network-induced increase in activity despite the addition of the noise, preventing the nodes from exhibiting fast limit cycle oscillations. In case B, dFIC places the nodes at the upper boundary of bistability, allowing the system to exhibit not only fast limit cycle oscillations but also slow limit cycle oscillations–due to noise and network effects. (TIF) Graphical representation of the best fitting functional connectivity (FC) matrices. Based on pure FC correlation (top-left) and best fitting simulation after correcting for synchrony (top-right), best fitting FC matrix based on MMF measure (bottom-left) and average empirical FC (bottom-right). empFC: empirical functional connectivity; FC: functional connectivity; MMF: multi-modal factor. (TIF) Statistical analyses. (PDF) 30 Jul 2024 Dear Mr Stasinski, Thank you very much for submitting your manuscript "Homeodynamic feedback inhibition control in whole-brain simulations" for consideration at PLOS Computational Biology. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments. We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation. When you are ready to resubmit, please upload the following: [1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. [2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file). Important additional instructions are given below your reviewer comments. Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts. Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments. Sincerely, Marcus Kaiser, Ph.D. Academic Editor PLOS Computational Biology Lyle Graham Section Editor PLOS Computational Biology *********************** Reviewer's Responses to Questions Comments to the Authors: Please note here if the review is uploaded as an attachment. Reviewer #1: This paper presents an analysis of a recently developed method for the stabilisation of brain dynamics in large-scale neural mass model simulations. The method is based on the idea that activity homeostasis can be achieved by adjusting inhibitory weights in an activity-dependent manner to maintain a set-point firing rate. The key insight used here is that global stability is maintained when this set-point corresponds to an attractor of a single-node model (the Jansen-Rit model in this paper). The authors show that this method can control the target activity level to obtain biologically plausible neural activity, including central features of BOLD signals. Interestingly, the method is also shown to maintain to high variability of model dynamics corresponding to critical/bifurcation points hypothesised to be important for brain function. Although presented mainly as a technical solution to stabilise large-scale brain simulations, the work also adds support to the hypothesis that neural homeostasis through inhibitory plasticity may be functionally important - I wonder if this idea could be explored in a data-driven way with the methods presented here. I have no major comments on the paper, in my opinion it is well written and argued. I think though that equations 1-4 miss the dynamics (it looks like eqn. 4c should be the derivative of wFIC). Second, it would be prudent to document the published code to enable re-use. As it is, the repository only has some python files without any documentation or instructions. Reviewer #2: This article suggested the methodology to obtain biologically plausible simulation of a brain by improving widely used Jansen-Rit model, and performed thorough analyses about how the authors’ method could regenerate the characteristics of brain signal, especially the dynamical features of common complex system and neuroscientific metrics. Although the authors handled mean-field model and BOLD data proficiently and mostly convincing results, I would like to recommend this article for publishing only after significant efforts would be made for improving the quality of manuscript. 1. Overall readability should be significantly improved. 1-1) There is substantial number of honest mistakes across the whole paper. For instance, reference 18, 19, 59, 72, 84, 92 were obviously broken. Equation 6 was not indexed. At line number 485, referring figure number went wrong. I could find a dozen more even after initial superficial reading of the manuscript, and it is hard to list all that I found while repetitive readings. It could be simply minor points to improve but many mistakes arouse a doubt that there might be overlooked critical ones. I would like to recommend that the authors thoroughly check especially, figure, equation, reference number in the text. 1-2) Several acronyms were not introduced properly. Ex) JR at line number 152, FC at 164. 1-3) All the coefficients and variable names should be at least once briefly explained. Ex) T in eq 5a, R_FC in eq 6. 1-4) Table 1 was never referred in the text. In addition, though JR model and its coefficient set are well-known, the supporting prior works or reasoning to determine the values should be offered in the text. 1-5) ‘Introduction’, especially line 74-126 or line 127-188 can be substantially shrunk by removing redundant messages and rearrangement to ‘discussion’. The reason why I brought up this issue is that because of this long introduction compared to the very specific and neat motivation and summary of the work, the main novelty gets blurred, resulting in unnecessary complexity of the flow. 2. Since the authors emphasized that the introduction of dFIC into JR model improves the pipeline to construct a better BMN regenerating ‘biologically plausible’ brain signal, the novelty of this work was spotlighted mainly in the perspective on ‘observation-simulation matching.’ Although JR model and the other types of mean-field models don’t target the biological plausibility (or similarity) of local network architecture, dFIC seems capable of somewhat compensating such simplification. I would like to recommend for the authors to state this point briefly. Although somehow, in Discussion, the section stating E-I balance is weakly related to this point, it seems possible that a bolder interpretation, such as dFIC can be the biological mechanism of local E-I balance, can be suggested. 3. As the extension of 2, it would be much better to be discussed, the determination of wFIC and pFIC in the pathological conditions (one of the main purpose of TVB) and the implications of resultant wFIC and pFIC. 4. Poincare map 4-1) In Figure 2, I did not understand how Poincare map without varying lag time or transition map (between dots) could specify the dynamics of each regime such as slow limit cycle or fast limit cycle. 4-2) In the caption of fig2, ‘each color marks a single node’ and ‘colors on the map mark the indegree of the nodes’ are very confusing comments. 4-3) As an extension, it would be better to kindly offer the full details of the Poincare map analysis in Method section. 4-4) In Figure 15 in the subplots, traversing number 3 seems not changing in any case (~25). I would like to recommend that the authors offer the interpretation on it. Reviewer #3: In this manuscript, Stasinki et al. proposed a new version of the Jansen & Rit model which includes a Feedback Inhibitory Control (FIC) mechanism, similar to the one implemented in Abeysuriya et al. (Plos Comp Biol, 2018) and Deco et al. (Journal of Neuroscience, 2014). They first introduced the model, and then explored how the FIC can change the dynamics at the single region level through bifurcation analyses. The authors performed small network simulations to show the capability of the FIC mechanisms to preserve E/I balance in network simulations when increasing the global coupling. Finally, they characterized the model’s performance in fitting empirical fMRI BOLD functional connectivity and functional connectivity dynamics empirical data. I found the article very interesting to read, and methodologically robust. There’s no doubt that the authors performed not only exhaustive but also correct analyses to demonstrate the plausibility and usefulness of incorporating FIC into the Jansen & Rit model. The figures are of high quality, and the results completely support the main hypotheses. In the field of whole-brain modeling, an implementation of FIC mechanisms (or other plasticity-based mechanisms for preserving E/I balance) is needed in the field of neural mass modeling using the Jansen & Rit model. I would recommend this manuscript if the authors can address the following major and minor suggestions. Majors The article would benefit from a reduction of content, especially in the Introduction section, which is quite long. For example, the paragraph that starts in line 97 “from a dynamical systems perspective…” is redundant compared to the one before it. The two paragraphs of lines 114 and 127 are both about E/I balance, hyperexcitability, and FIC, and can be reduced and combined into one. The paragraph of line 143 can be reduced or removed. Other phrases, like “This version of FIC is formulated as a set of ordinary differential equations…” (line 180) can be moved to Methods, and other ones like “The analytical treatment of the single-node tuning dynamics reveals the conditions…” (line 185) may be more properly placed in the Results. What is the motivation of the authors for using three differential equations for including the FIC mechanisms, instead of just one (like in Abeysuriya et al., Plos Comp Biol, 2018)? For me, it wasn’t clear when reading the text. What is the advantage of including the additional two state variables in the model? Perhaps the authors can address this point in the Discussion, and/or compare the model’s outputs or performance with a version of the model that includes only the equation 4c. The authors discarded simulated FCs with a mean > 0.25. This can be avoided using metrics that “balance” connectivity pattern similarity and differences in global correlations, like the structural similarity index (like in Ipiña et al., 2020, Neuroimage). Why the authors didn’t use the SSIM or other composed metrics (similar to Pfeffer et al. 2021, Science Advances)? The authors showed that the model can fit fMRI BOLD FC and FCD. However, it would be good to show that the model is also able to fit some properties of empirical EEG (e.g., EEG FC, power spectrum). The authors can try to fit both fMRI and EEG dynamics using the model and the data from the HCP. If not, it would be nice at least to see what the EEG power spectrum looks like when using the optimal parameters of the fMRI fitting. I also recommend addressing this limitation of the work in the Discussion if not addressed. In the Introduction section, Line 152, the authors said that “… so far, an FIC mechanism is missing for the Jansen & Rit model”. Indeed, a mechanism for preserving the E/I balance in the Jansen & Rit model was proposed by Coronel-Oliveros et al. (2024, Alzheimer's & Dementia). I think that the authors should mention this paper as an antecedent of FIC mechanisms in the Jansen & Rit model. The mechanism proposed in Coronel-Oliveros et al. is very similar (but of course not equal) to the implementation proposed by Stasinki et al. here, without considering the additional activity detection equations (4a and 4b). Perhaps the authors can discuss the differences between these two versions of the Jansen & Rit model. It would be a nice addition to the Discussion to talk about possible clinical applications, in a very short paragraph, of the JR + FIC model, for example, proposing how the FIC can be used to study E/I disbalances in dementia (see for example Ranasinghe et al., 2022, eLife; Martinez-Cañada et al., 2023, Alzheimer’s & Dementia: Diagnosis, Assessment and Disease Monitoring; Coronel-Oliveros et al., 2024; Alzheimer’s and Dementia; Moguilner et al., 2024; Alzheimer’s Research & Therapy). In this other work, the authors used the JR model for proposing biomarkers from the model, and not from the data: Amato et al., 2024, Alzheimer’s & Dementia: Diagnosis, Assessment and Disease Monitoring). Minors Line 26 in Abstract: “Using a variety of metrics”. What metrics? Line 30-31 in Abstract: the two lines might be reduced to “… of fitting simulated to fMRI static and dynamic functional connectivity data”. The Figure 1 description (and the figure itself) should be moved to the Results section. It would be nice to include a scheme or drawing of the model. Perhaps the authors can move the Fig S1 in Supp Material to the Main Text. Line 346 in results: “Before convergence of wFIC_i one can encounter…” This may be moved to the Results section. Line 391: “We have identified…” - move to Results. Figure 2 should be in Results, not Methods. Line 398 Fitting methods: how many model realizations (random seeds) did you use for stochastic simulations? Line 413 in Methods: “Into a single multi-modal fitness …” add the acronym (MMF). Line 429 in Methods: “To test if differential MMF fitting…”. That entire part may be moved to Results. Line 538 in Results: “In summary, the correct application…”. That phrase is redundant and can be removed. Line 814 in Discussion: “In contrast, when the model was tuned to…”. Limit to discuss the results, not repeat them in the Discussion. Line 823 in Discussion: “Overall, dFIC gives more…”. I think this might be removed. Line 834 in Discussion: “The criticality hypothesis says that…”. This was explained in the Introduction, and the Discussion section ********** Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: Yes Reviewer #2: Yes Reviewer #3: Yes ********** PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: Yes: Matthias H Hennig Reviewer #2: No Reviewer #3: Yes: Vicente Medel Figure Files: While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Data Requirements: Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5. Reproducibility: To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols 30 Sep 2024 Submitted filename: Response2Reviewers_dFIC.pdf 25 Oct 2024 Dear Mr Stasinski, We are pleased to inform you that your manuscript 'Homeodynamic feedback inhibition control in whole-brain simulations' has been provisionally accepted for publication in PLOS Computational Biology. Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests. Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated. IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript. Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS. Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology. Best regards, Marcus Kaiser, Ph.D. Academic Editor PLOS Computational Biology Lyle Graham Section Editor PLOS Computational Biology Feilim Mac Gabhann Editor-in-Chief PLOS Computational Biology Jason Papin Editor-in-Chief PLOS Computational Biology *********************************************************** Reviewer's Responses to Questions Comments to the Authors: Please note here if the review is uploaded as an attachment. Reviewer #1: A good and comprehensive revision. Reviewer #2: The authors carefully addressed my questions and suggestions for correction and reinforcement of the manuscript, and the revised manuscript is adequately improved. Especially, the current version delivers their message in focused manner with full information about the simulation that they performed. Thus, I would like to recommend this version of manuscript to be published. Reviewer #3: The authors have succesfully answered all my inquiries. ********** Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: None Reviewer #2: Yes Reviewer #3: Yes ********** PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: Yes: Taegon Kim Reviewer #3: Yes: Vicente Medel 19 Nov 2024 PCOMPBIOL-D-24-00577R1 Homeodynamic feedback inhibition control in whole-brain simulations Dear Dr Stasinski, I am pleased to inform you that your manuscript has been formally accepted for publication in PLOS Computational Biology. Your manuscript is now with our production department and you will be notified of the publication date in due course. The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript. Soon after your final files are uploaded, unless you have opted out, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers. Thank you again for supporting PLOS Computational Biology and open-access publishing. We are looking forward to publishing your work! With kind regards, Lilla Horvath PLOS Computational Biology | Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom ploscompbiol@plos.org | Phone +44 (0) 1223-442824 | ploscompbiol.org | @PLOSCompBiol