Learning interpretable dynamics of stochastic complex systems from experimental data.

Complex systems with many interacting nodes are inherently stochastic and best described by stochastic differential equations. Despite increasing observation data, inferring these equations from empirical data remains challenging. Here, we propose the Langevin graph network approach to learn the hidden stochastic differential equations of complex networked systems, outperforming five state-of-the-art methods. We apply our approach to two real systems: bird flock movement and tau pathology diffusion in brains. The inferred equation for bird flocks closely resembles the second-order Vicsek model, providing unprecedented evidence that the Vicsek model captures genuine flocking dynamics. Moreover, our approach uncovers the governing equation for the spread of abnormal tau proteins in mouse brains, enabling early prediction of tau occupation in each brain region and revealing distinct pathology dynamics in mutant mice. By learning interpretable stochastic dynamics of complex systems, our findings open new avenues for downstream applications such as control.

Correction: A spatial vaccination strategy to reduce the risk of vaccine-resistant variants.

[This corrects the article DOI: 10.1371/journal.pcbi.1010391.].

Epidemic spreading under mutually independent intra- and inter-host pathogen evolution.

The dynamics of epidemic spreading is often reduced to the single control parameter R0 (reproduction-rate), whose value, above or below unity, determines the state of the contagion. If, however, the pathogen evolves as it spreads, R0 may change over time, potentially leading to a mutation-driven spread, in which an initially sub-pandemic pathogen undergoes a breakthrough mutation. To predict the boundaries of this pandemic phase, we introduce here a modeling framework to couple the inter-host network spreading patterns with the intra-host evolutionary dynamics. We find that even in the extreme case when these two process are driven by mutually independent selection forces, mutations can still fundamentally alter the pandemic phase-diagram. The pandemic transitions, we show, are now shaped, not just by R0, but also by the balance between the epidemic and the evolutionary timescales. If mutations are too slow, the pathogen prevalence decays prior to the appearance of a critical mutation. On the other hand, if mutations are too rapid, the pathogen evolution becomes volatile and, once again, it fails to spread. Between these two extremes, however, we identify a broad range of conditions in which an initially sub-pandemic pathogen can breakthrough to gain widespread prevalence.

A spatial vaccination strategy to reduce the risk of vaccine-resistant variants.

The COVID-19 pandemic demonstrated that the process of global vaccination against a novel virus can be a prolonged one. Social distancing measures, that are initially adopted to control the pandemic, are gradually relaxed as vaccination progresses and population immunity increases. The result is a prolonged period of high disease prevalence combined with a fitness advantage for vaccine-resistant variants, which together lead to a considerably increased probability for vaccine escape. A spatial vaccination strategy is proposed that has the potential to dramatically reduce this risk. Rather than dispersing the vaccination effort evenly throughout a country, distinct geographic regions of the country are sequentially vaccinated, quickly bringing each to effective herd immunity. Regions with high vaccination rates will then have low infection rates and vice versa. Since people primarily interact within their own region, spatial vaccination reduces the number of encounters between infected individuals (the source of mutations) and vaccinated individuals (who facilitate the spread of vaccine-resistant strains). Thus, spatial vaccination may help mitigate the global risk of vaccine-resistant variants.

Distribution equality as an optimal epidemic mitigation strategy.

Upon the development of a therapeutic, a successful response to a global pandemic relies on efficient worldwide distribution, a process constrained by our global shipping network. Most existing strategies seek to maximize the outflow of the therapeutics, hence optimizing for rapid dissemination. Here we find that this intuitive approach is, in fact, counterproductive. The reason is that by focusing strictly on the quantity of disseminated therapeutics, these strategies disregard the way in which this quantity distributes across destinations. Most crucially-they overlook the interplay of the therapeutic spreading patterns with those of the pathogens. This results in a discrepancy between supply and demand, that prohibits efficient mitigation even under optimal conditions of superfluous flow. To solve this, we design a dissemination strategy that naturally follows the predicted spreading patterns of the pathogens, optimizing not just for supply volume, but also for its congruency with the anticipated demand. Specifically, we show that epidemics spread relatively uniformly across all destinations, prompting us to introduce an equality constraint into our dissemination that prioritizes supply homogeneity. This strategy may, at times, slow down the supply rate in certain locations, however, thanks to its egalitarian nature, which mimics the flow of the pathogens, it provides a dramatic leap in overall mitigation efficiency, potentially saving more lives with orders of magnitude less resources.

Topological synchronization of chaotic systems.

A chaotic dynamics is typically characterized by the emergence of strange attractors with their fractal or multifractal structure. On the other hand, chaotic synchronization is a unique emergent self-organization phenomenon in nature. Classically, synchronization was characterized in terms of macroscopic parameters, such as the spectrum of Lyapunov exponents. Recently, however, we attempted a microscopic description of synchronization, called topological synchronization, and showed that chaotic synchronization is, in fact, a continuous process that starts in low-density areas of the attractor. Here we analyze the relation between the two emergent phenomena by shifting the descriptive level of topological synchronization to account for the multifractal nature of the visited attractors. Namely, we measure the generalized dimension of the system and monitor how it changes while increasing the coupling strength. We show that during the gradual process of topological adjustment in phase space, the multifractal structures of each strange attractor of the two coupled oscillators continuously converge, taking a similar form, until complete topological synchronization ensues. According to our results, chaotic synchronization has a specific trait in various systems, from continuous systems and discrete maps to high dimensional systems: synchronization initiates from the sparse areas of the attractor, and it creates what we termed as the 'zipper effect', a distinctive pattern in the multifractal structure of the system that reveals the microscopic buildup of the synchronization process. Topological synchronization offers, therefore, a more detailed microscopic description of chaotic synchronization and reveals new information about the process even in cases of high mismatch parameters.

Alternating quarantine for sustainable epidemic mitigation.

Absent pharmaceutical interventions, social distancing, lock-downs and mobility restrictions remain our prime response in the face of epidemic outbreaks. To ease their potentially devastating socioeconomic consequences, we propose here an alternating quarantine strategy: at every instance, half of the population remains under lockdown while the other half continues to be active - maintaining a routine of weekly succession between activity and quarantine. This regime minimizes infectious interactions, as it allows only half of the population to interact for just half of the time. As a result it provides a dramatic reduction in transmission, comparable to that achieved by a population-wide lockdown, despite sustaining socioeconomic continuity at  ~50% capacity. The weekly alternations also help address the specific challenge of COVID-19, as their periodicity synchronizes with the natural SARS-CoV-2 disease time-scales, allowing to effectively isolate the majority of infected individuals precisely at the time of their peak infection.

IRS1 phosphorylation underlies the non-stochastic probability of cancer cells to persist during EGFR inhibition therapy.

Stochastic transition of cancer cells between drug-sensitive and drug-tolerant persister phenotypes has been proposed to play a key role in non-genetic resistance to therapy. Yet, we show here that cancer cells actually possess a highly stable inherited chance to persist (CTP) during therapy. This CTP is non-stochastic, determined pre-treatment and has a unimodal distribution ranging from 0 to almost 100%. Notably, CTP is drug specific. We found that differential serine/threonine phosphorylation of the insulin receptor substrate 1 (IRS1) protein determines the CTP of lung and of head and neck cancer cells under epidermal growth factor receptor inhibition, both in vitro and in vivo. Indeed, the first-in-class IRS1 inhibitor NT219 was highly synergistic with anti-epidermal growth factor receptor therapy across multiple in vitro and in vivo models. Elucidation of drug-specific mechanisms that determine the degree and stability of cellular CTP may establish a framework for the elimination of cancer persisters, using new rationally designed drug combinations.

Universal patterns in passenger flight departure delays.

Departure delays are a major cause of economic loss and inefficiency in the growing industry of passenger flights. A departure delay of a current flight is inevitably affected by the late arrival of the flight immediately preceding it with the same aircraft. We seek to understand the mechanisms of such propagated delays, and to obtain universal metrics by which to evaluate an airline's operational effectiveness in delay alleviation. Here we use big data collected by the American Bureau of Transportation Statistics to design models of flight delays. Offering two dynamic models of delay propagation, we divided all carriers into two groups exhibiting a shifted power law or an exponentially truncated shifted power law delay distribution, revealing two universal delay propagation classes. Three model parameters, extracted directly from dual data mining, help characterize each airline's operational efficiency in delay mitigation. Therefore, our modeling framework provides the crucially lacking evaluation indicators for airlines, potentially contributing to the mitigation of future departure delays.

Unusual changeover in the transition nature of local-interaction Potts models.

A combinatorial approach is used to study the critical behavior of a q-state Potts model with a round-the-face interaction. Using this approach, it is shown that the model exhibits a first order transition for q>3. A second order transition is numerically detected for q=2. Based on these findings, it is deduced that for some two-dimensional ferromagnetic Potts models with completely local interaction, there is a changeover in the transition order at a critical integer q_{c}≤3. This stands in contrast to the standard two-spin interaction Potts model where the maximal integer value for which the transition is continuous is q_{c}=4. A lower bound on the first order critical temperature is additionally derived.

Digitizable therapeutics for decentralized mitigation of global pandemics.

When confronted with a globally spreading epidemic, we seek efficient strategies for drug dissemination, creating a competition between supply and demand at a global scale. Propagating along similar networks, e.g., air-transportation, the spreading dynamics of the supply vs. the demand are, however, fundamentally different, with the pathogens driven by contagion dynamics, and the drugs by commodity flow. We show that these different dynamics lead to intrinsically distinct spreading patterns: while viruses spread homogeneously across all destinations, creating a concurrent global demand, commodity flow unavoidably leads to a highly uneven spread, in which selected nodes are rapidly supplied, while the majority remains deprived. Consequently, even under ideal conditions of extreme production and shipping capacities, due to the inherent heterogeneity of network-based commodity flow, efficient mitigation becomes practically unattainable, as homogeneous demand is met by highly heterogeneous supply. Therefore, we propose here a decentralized mitigation strategy, based on local production and dissemination of therapeutics, that, in effect, bypasses the existing distribution networks. Such decentralization is enabled thanks to the recent development of digitizable therapeutics, based on, e.g., short DNA sequences or printable chemical compounds, that can be distributed as digital sequence files and synthesized on location via DNA/3D printing technology. We test our decentralized mitigation under extremely challenging conditions, such as suppressed local production rates or low therapeutic efficacy, and find that thanks to its homogeneous nature, it consistently outperforms the centralized alternative, saving many more lives with significantly less resources.

Author Correction: Universal resilience patterns in complex networks.

In this Letter, in Fig. 3c and f the Saccharomyces cerevisiae and Escherichia coli networks were subject to both weight loss and node deletion, a combination of two types of perturbation, as opposed to weight loss only (as the labelling incorrectly indicated). The collapse in Fig. 3h was also obtained from this combined perturbation, and therefore the results displayed in Fig. 3h remain fully consistent with the theoretical framework presented in this Letter. Figure 1 to this Amendment shows the corrected Fig. 3c, f and h, in which Fig. 3c and f have been generated with weight-loss perturbations only, as originally reported, together with the originally published panels, for completeness and transparency. The codes used to generate the original and the corrected Fig. 3 are available at https://github.com/jianxigao/NuRsE . We thank Travis A. Gibson for alerting us to this error. The original Letter has not been corrected.

The Metastability of the Double-Tripod Gait in Locust Locomotion.

Insect locomotion represents a fundamental example of neuronal oscillating circuits generating different motor patterns or gaits by controlling their phase coordination. Walking gaits are assumed to represent stable states of the system, often modeled as coupled oscillators. This view is challenged, however, by recent experimental observations, in which in vitro locust preparations consistently converged to synchronous rhythms (all legs oscillating as one), a locomotive pattern never seen in vivo. To reconcile this inconsistency, we developed a modeling framework to capture the trade-off between the two competing mechanisms: the endogenous neuronal circuitry, expressed in vitro, and the feedback mechanisms from sensory and descending inputs, active only in vivo. We show that the ubiquitously observed double-tripod walking gait emerges precisely from this balance. The outcome is a short-lived meta-stable double-tripod gait, which transitions and alternates with stable idling, thus recovering the observed intermittent bouts of locomotion, typical of many insects' locomotion behavior.

Dynamic patterns of information flow in complex networks.

Although networks are extensively used to visualize information flow in biological, social and technological systems, translating topology into dynamic flow continues to challenge us, as similar networks exhibit fundamentally different flow patterns, driven by different interaction mechanisms. To uncover a network's actual flow patterns, here we use a perturbative formalism, analytically tracking the contribution of all nodes/paths to the flow of information, exposing the rules that link structure and dynamic information flow for a broad range of nonlinear systems. We find that the diversity of flow patterns can be mapped into a single universal function, characterizing the interplay between the system's topology and its dynamics, ultimately allowing us to identify the network's main arteries of information flow. Counter-intuitively, our formalism predicts a family of frequently encountered dynamics where the flow of information avoids the hubs, favoring the network's peripheral pathways, a striking disparity between structure and dynamics.

Universal resilience patterns in complex networks.

Resilience, a system's ability to adjust its activity to retain its basic functionality when errors, failures and environmental changes occur, is a defining property of many complex systems. Despite widespread consequences for human health, the economy and the environment, events leading to loss of resilience--from cascading failures in technological systems to mass extinctions in ecological networks--are rarely predictable and are often irreversible. These limitations are rooted in a theoretical gap: the current analytical framework of resilience is designed to treat low-dimensional models with a few interacting components, and is unsuitable for multi-dimensional systems consisting of a large number of components that interact through a complex network. Here we bridge this theoretical gap by developing a set of analytical tools with which to identify the natural control and state parameters of a multi-dimensional complex system, helping us derive effective one-dimensional dynamics that accurately predict the system's resilience. The proposed analytical framework allows us systematically to separate the roles of the system's dynamics and topology, collapsing the behaviour of different networks onto a single universal resilience function. The analytical results unveil the network characteristics that can enhance or diminish resilience, offering ways to prevent the collapse of ecological, biological or economic systems, and guiding the design of technological systems resilient to both internal failures and environmental changes.

Constructing minimal models for complex system dynamics.

One of the strengths of statistical physics is the ability to reduce macroscopic observations into microscopic models, offering a mechanistic description of a system's dynamics. This paradigm, rooted in Boltzmann's gas theory, has found applications from magnetic phenomena to subcellular processes and epidemic spreading. Yet, each of these advances were the result of decades of meticulous model building and validation, which are impossible to replicate in most complex biological, social or technological systems that lack accurate microscopic models. Here we develop a method to infer the microscopic dynamics of a complex system from observations of its response to external perturbations, allowing us to construct the most general class of nonlinear pairwise dynamics that are guaranteed to recover the observed behaviour. The result, which we test against both numerical and empirical data, is an effective dynamic model that can predict the system's behaviour and provide crucial insights into its inner workings.

Universality in network dynamics.

Despite significant advances in characterizing the structural properties of complex networks, a mathematical framework that uncovers the universal properties of the interplay between the topology and the dynamics of complex systems continues to elude us. Here we develop a self-consistent theory of dynamical perturbations in complex systems, allowing us to systematically separate the contribution of the network topology and dynamics. The formalism covers a broad range of steady-state dynamical processes and offers testable predictions regarding the system's response to perturbations and the development of correlations. It predicts several distinct universality classes whose characteristics can be derived directly from the continuum equation governing the system's dynamics and which are validated on several canonical network-based dynamical systems, from biochemical dynamics to epidemic spreading. Finally, we collect experimental data pertaining to social and biological systems, demonstrating that we can accurately uncover their universality class even in the absence of an appropriate continuum theory that governs the system's dynamics.

Network link prediction by global silencing of indirect correlations.

Predictions of physical and functional links between cellular components are often based on correlations between experimental measurements, such as gene expression. However, correlations are affected by both direct and indirect paths, confounding our ability to identify true pairwise interactions. Here we exploit the fundamental properties of dynamical correlations in networks to develop a method to silence indirect effects. The method receives as input the observed correlations between node pairs and uses a matrix transformation to turn the correlation matrix into a highly discriminative silenced matrix, which enhances only the terms associated with direct causal links. Against empirical data for Escherichia coli regulatory interactions, the method enhanced the discriminative power of the correlations by twofold, yielding >50% predictive improvement over traditional correlation measures and 6% over mutual information. Overall this silencing method will help translate the abundant correlation data into insights about a system's interactions, with applications ranging from link prediction to inferring the dynamical mechanisms governing biological networks.