ABSTRACT. One of the central tenants of biology is that our genetics—our genotype—influences the physical characteristics we manifest—our phenotype. But with more than 25,000 human genes and more than 6,000,000 common genetic variants mapped in our genome, finding associations between our genotype and phenotype is an ongoing challenge. Indeed, genome-wide association studies have found thousands of small effect size genetic variants that are associated with phenotypic traits and disease. The simplest explanation is that these genetic variants work synergistically to help define phenotype and to regulate processes that are responsible for phenotypic state transitions. We will use gene expression and genetic data to explore gene regulatory networks, to study phenotypic state transitions, and to analyze the connections between genotype, gene expression, and phenotyope. We have found that the networks, and their structure, provide unique insight into how genetic elements interact with each other and the structure of the network has predictive power for identifying SNPs likely to be associated with phenotype through genome wide association studies. I will show multiple examples, drawing on my work in cancer, in chronic obstructive pulmonary disease, and in the analysis of data from thirty-eight tissues provided by the Genotype-Tissue Expression (GTEx) project.

ABSTRACT. Crater-forming impacts represent a class of extreme events involving high energy release and short time scales. Impacts constitute major geologic processes shaping the surfaces and evolution of planetary bodies. Formation of large craters involves high pressures and temperatures resulting in intense deformation, fracturing and melting. Impacts produce deep transient cavities, with excavation to deep levels in the crust, fragmentation, and removal of large volumes of rock. Here, we analyze the Chicxulub impact and its effects on the Earth´s climate, environment and life-support systems, in relation to the Cretaceous/Paleogene boundary. The boundary represents one of the major extinction events in the Phanerozoic, which affected ~75 % of species in the continental and marine realms. Effects of the impact have been intensely investigated, where the affectation in the evolution patterns was profound and long-lasting. The disappearance of large numbers of species including complete groups severely affected the biodiversity and ecosystem composition in the marine and continental realms. There are several aspects involved in addressing the Chicxulub impact as an extreme event, which involves a complex interplay of processes in the Earth´s system. First we examine the impact event and cratering, time scales involved and energy released. Next, we assess the impact´s regional and global effects, which involve major perturbations in the ocean and atmosphere, including feedback mechanisms and nonlinear effects. From here, we discuss how and to what extent life-support systems are affected by large impacts, and what the fossil record tells about the extinction event and biotic turnover. In particular, how sudden or extended are the processes, the extinction event and the biodiversity recovery processes. 

ABSTRACT. It is widely held that individual entities are more likely to exhibit the same or similar behavior if they are equal to each other--imagine animals using the same gait, lasers pulsing together, birds singing the same notes, and agents reaching consensus. In this presentation, I will show that this assumption is in fact false in networks of interacting entities. This surprising observation is rooted in a new network phenomenon we term “asymmetry-induced symmetry” (AIS), in which the state of the system can be symmetric only when the system itself is not. Using spontaneous synchronization as a model process, I will discuss scenarios where the state in which all nodes exhibit identical dynamics (a state of maximum symmetry) can only be realized when the nodes themselves are not identical. AIS can be seen as the converse of the well-studied phenomenon of symmetry breaking, where the state has less symmetry than the system. AIS has far-reaching implications for processes that involve converging to uniform states; in particular, it offers a mechanism for yet-to-be-explained convergent forms of pattern formation, in which an asymmetric structure develops into a symmetric one. AIS also has implications for consensus dynamics, where it gives rise to scenarios in which interacting agents only reach consensus when they are sufficiently different from each other.

ABSTRACT. Multistate dynamical processes on networks – where nodes can be in one of a finite number of states – are widely used as modelling frameworks in the epidemiological, physical and socio-technical sciences. In epidemiological settings, for example, complex multistate models are employed to forecast the evolution of real diseases, offering more realistic and accurate predictive tools than the simpler Susceptible-Infected-Susceptible and Susceptible-Infected-Recovered models.

In this work, we present a general formulation of Markovian multistate dynamical processes on complex networks that is representative of a broad range of models in the complex systems domain. In this formalism, the dynamics are specified by probabilistic rate functions Fm(a → b), indicating the rate at which a node in state a changes to state b depending on the states of its neighbours, encoded in the vector m. With this formalism, we construct theoretical frameworks of varying complexity and accuracy. Such frameworks allow for the analysis of multistate dynamics processes on networks, and the effect that the degree distribution of the network has on the evolution and equilibrium behaviour of the dynamics. We illustrate the power of our approach with novel analysis of a recently proposed model of co-operative diseases, gaining a deep understanding of the stability of the disease-free and endemic states of the model and the effect of network structure on these equilibrium states. Finally, we present detailed analysis of the special case of linear rate functions; such rate functions correspond to independent pairwise interactions, covering a broad range of epidemiological and spreading models. Employing our developed theoretical framework, we derive an expression for the position of the epidemic threshold of the dynamics, a manifold in parameter space that separates a disease-free equilibrium state from one which is endemic. This remarkable result is highly powerful because of its generality for arbitrary multistate linear rate functions, and is a major step in understanding the effects of highly complex spreading mechanisms on viral outbreaks.

ABSTRACT. Dealing with missing data in networks is a long-standing problem with important applications. For temporal networks, where missing data result in missing links, a major challenge is to extract information relevant to recover a faithful outcome of spreading processes occurring on them. To this aim, we leverage the fact that temporal networks (here, human proximity networks) can be naturally represented as 3-dimensional tensors whose slices correspond to the adjacency matrices of the network at a given time. This enables us to devise an approach based on tensor decomposition (NTF) – a dimensionality reduction technique from multi-linear algebra – to recover some properties, related to the nodes, whose some of the links are missing, and exploit these properties to build a surrogate of the network with the aim of reproducing a faithful spreading dynamics. The key idea is to learn a low-dimensional representation of the partial network and adapt it to take into account temporal and structural heterogeneity properties we know to be crucial for spreading processes occurring on networks. To test our approach we take into account three human proximity networks of different sizes and measured in different contexts. We simulate a loss of data for 10, 20 and 40% of the network nodes by removing their links for half the time span of the dataset taken continuously. Using our method on the resulting partial networks, we build a surrogate version for each of the networks. We compare the outcome of a Susceptible-Infected-Recovered (SIR) process on the original networks (non altered by a loss of data) and on the surrogate networks. We observe that the epidemic sizes obtained with the surrogate networks are in good agreement with those measured on the original networks. Finally, we propose a natural extension of our framework, to take into account additional data sources, when available. Here, we have access to the approximate location of individuals of the networks in time. The extension relies on joint tensor factorization (JNTF), that decomposes several tensors at a time. The rationale behind is that the information of the tensor representing node locations in time can help to recover their missing information in the node proximity tensor. This approach enables to face drastic cases where either all nodes miss some links or some nodes completely lose their links in the proximity network. We consider the two following scenarios to evaluate the method: a) 100% of nodes loses 50% of their links at random b) 50% of nodes loses all their links. We apply our method based on the joint factorization the partial proximity network and the node location network, to build a surrogate version of the proximity network. We simulate SIR processes on the partial, original and surrogate networks. The epidemics sizes measured on the surrogate network build through the JNTF method provide a good estimation with respect to the original case. To summarize, we propose a framework, based on tensor decomposition, to face missing data in temporal networks to estimate the outcome of spreading processes in various context.

ABSTRACT. When will an idea or behavior successfully spread through a population? Many different theoretical models of ``complex contagion'' processes have been proposed and studied to answer this question. Similarly, an ever increasing amount of network data is available to empirically study and to predict the outcome of such processes. However, although we understand how some of the global and local characteristics of network structure affect the global outcome of these spreading processes, little is understood about the general stability of even the simplest complex contagion models under small perturbations to network structure. Here we study perturbations to a network that leave the global and local structure statistically unchanged but have large effects on global spread. Specifically, we consider contagion according to a fractional threshold, which is one of the best-understood complex contagion mechanisms.

Community structure and modularity have been shown to play an important role in shaping the global outcome of contagion. Under fractional threshold contagion, communities foster spread originating within the community through reinforcement effects, but also obstruct the spread that originates outside. Prior work finds that there is a window of intermediate inter-modular connectivity that sustains optimal global spread [1]. These findings consider randomly structured inter-modular connectivity. However, real-world networks commonly exhibit a hierarchical structure where high-degree (central) nodes are more likely to provide inter-modular connectivity than low degree (peripheral) nodes. Interestingly, we find that adding a very small number of inter-modular connections between peripheral nodes in different modules quickly enables global spreading. On the other hand, a much larger number of connections between central nodes in different communities are needed to generate global spread. We derive general conditions for the structure of modules and inter-modular connectivity that exhibit this strong sensitivity to the degree-degree structure of inter-modular links.

Social networks typically have marked community structure, degree heterogeneity and stratification and connections between peripheral individuals in different communities are rare. In other words, they frequently exhibit the structural characteristics that sustain the structural instability we have identified. Thus, our findings have many important implications for spread in social systems. For example, consider the case where we want to stimulate the spread of a behavior. If links can be added, the most efficient way to ensure global spread is to add connectivity between peripheral nodes in different modules. Interestingly, this implies peripheral nodes can have a huge global impact where central nodes cannot. Another interesting application concerns evaluating the predictive power of contagion models simulated on measured networks. When the conditions for structural instability are satisfied, predictions made are highly sensitive to small measurement errors perturbing the network structure. This has a clear consequence on the limits of predictability of complex contagion in many real world systems, especially given that in many systems peripheral nodes are the hardest to observe, and links between peripheral nodes even harder.

[1] Nematzadeh, Azadeh, et al. "Optimal network modularity for information diffusion." Physical review letters 113.8 (2014): 088701.

ABSTRACT. The problem of modeling the spread of a disease among individuals has been studied in deep over many years. The development of compartmental models, that divide the individuals among a set of possible states, has given rise to a new collection of technics that enables, for instance, the analysis of the epidemic threshold or the study of the impact of a prophylactic campaign. After the initial epidemiological studies on well-mixed populations, it has been recognized that complex networks constitute a better description for the substrate on top of which the epidemic spreading takes place. Among the many available epidemic models, the Susceptible-Infected-Susceptible (SIS) has become a cornerstone in the study of epidemic spreading in complex networks. From the initial analysis of SIS using heterogeneous mean field approximations to determine the epidemic threshold [1], to the recent ones in which the probability of being infected is determined at the level of node [2], there have been uncountable advances on this topic [3].

In this work we analyze the SIS model in complex networks at the level of edges. In particular, we propose the definition of the epidemic conductance as the probability that a link is in condition of spreading the epidemics. We show how to obtain equations for the conductance of all the links, which can be solved by iteration in a similar way to the Microscopic Markov Chain Approach (MMCA) in [2]. These equations provide a more accurate description of the global epidemic incidence and the epidemic threshold than previous methodologies.

The relevance of the epidemic conductance is proved in a set of experiments, in which we show that removing the edges of maximum conductance leads to a much faster way of leaving the endemic state than by taking out the most infected nodes. This shows the importance of analyzing epidemics at the level of links, and the need to consider them in any protocol to try to limit the incidence of epidemics.

[1] Pastor-Satorras, R. and Vespignani, A.: Epidemic spreading in scale-free networks. Physical Review Letters, 86 (2001) 3200.

[2] Gómez, S., Arenas, A., Borge-Holthoefer, J., Meloni, S. and Moreno, Y.: Discrete-time Markov chain approach to contact-based disease spreading in complex networks. EPL (Europhysics Letters), 89 (2010) 38009.

[3] Pastor-Satorras, R., Castellano, C., Van Mieghem, P. and Vespignani, A.: Epidemic processes in complex networks. Reviews of Modern Physics, 87 (2015) 925.

[4] Matamalas, J. T., Arenas, A. and Gómez, S.: Epidemic conductance in complex networks. Preprint (2017) in preparation.

ABSTRACT. What is the most effective way to spread a behavior on a network? The recent literature on network diffusion has focused mostly on models of simple contagion–where contagion can result from contact with a single “infected” individual–and complex contagion–where contagion requires contact with multiple “infected” sources. While in the case of simple contagion, strategies focused on central nodes are known to be effective, the strategies that are most effective in the case of complex contagion are relatively unknown. Here, we study the strategies that optimize the diffusion of a behavior on a network in the case of complex contagion by comparing algorithms that choose which nodes to target at each step. We find that, contrary to the case of simple contagion where targeting central nodes is an effective strategy, in the case of complex contagion minimizing the total diffusion time requires the use of dynamic strategies that target less connected nodes in the beginning and hubs at a critical intermediate time. That is, the strategic question in the case of complex contagion is when to target hubs. We solve the model analytically for simple network structures and also use numerical simulations to show that these dynamic strategies outperform simpler strategies that could be hypothesized to be effective, like always choosing the node with the highest probability of infection. These findings shed light on the dynamic strategies that optimize network diffusion in the case of complex contagion.

ABSTRACT. Despite great advances in neuroscience and genetic studies, our understanding of the biological sources of psychiatric disorders is still quite limited. An important reason for this is not having objective psychiatric clinical tests, in particular in the case of neurodevelopment, involving the structural growth and functional maturation of the central nervous system. Humans develop a variety of brain functions through this process such as learning ability, memory, and psychomotor skills. Abnormalities during this process, either due to genetic or environmental factors, can lead to a series of neurodevelopment disorders (NDD), including autism spectrum disorder (ASD), intellectual disability (ID), etc. A crucial challenge facing the field now is the lack of scientific or biological explanations in the current psychiatric diagnostic/classification systems. We have found a quantitative biomarker applicable to neurodevelopment disorders by carefully studying the statistical properties of how people move. No two people move in exactly the same way. It turns out that people that suffer from cognitive deficiencies have noticeable heterogeneous movement impairments when compared with typical development (TD). Recent advances of high-resolution wearable electromagnetic sensing devices enables continuous motion recordings at milliseconds time scales, away from detection of the naked eye. Using this technology, we have extracted information leading to unraveling quantitative neurodevelopment biomarkers. In this talk I will first set up the general problem in particular for the non-specialists. Then I’ll discuss, briefly our neurodevelopment results, that falls into the emerging field of Precision Quantitative Psychiatry. By studying the statistical properties of hand movement’s we discovered a new data-type which allowed us to catalog the continuity property of the body dynamics. Using correlation functions, nearest neighbor speed-spike statistics plus other statistical metrics, typical of studies of complex systems, we were able to quantitatively characterize each person cognitive abilities. We applied our metrics to individuals with ASD. Our quantitative statistical analysis led to a parameter phase space that provides an automatic ASD severity classification comparing it, a posteriori, with over 90% precision, to their diagnosed verbal speaking abilities. We also found, unexpectedly, many similarities in the parent’s biomarkers with those of their ASD diagnosed progeny.

ABSTRACT. Social interaction is impaired in High-Functioning Autism (HFA). This impairment is less pronounced in simplified technologically mediated circumstances such as when interacting with humanoid robots. This study aims to investigate quantitatively the real-time dyadic interaction between HFA patients and healthy participants in a minimal virtual reality environment known as the perceptual crossing experiment (PCE). An independent control group of healthy participants was also included. In the PCE pairs of blindfolded participants were embodied as avatars in a one-dimensional virtual space and moved their avatars with a mouse computer. A tactile vibration stimulus was delivered when the avatar crossed another object in the space. Each participant could encounter three objects: a static decoy, the avatar of her partner, and a shadow of her partner’s avatar. Hence, the only event when both partners received feedback simultaneously was when they crossed each other. The task was to mark these encounters by clicking but not those with the decoy or partner’s shadow. Here we treated each participant’s movement in the virtual space as a point process by defining as salient events the zero-crossings in acceleration. An index of multi-scale similarity between partners was computed by applying complexity matching (CM), which compares two power-law distributions and represents the maximization of information exchange between complex networks. CM can be obtained for point-processes by comparing Allan Factor (AF) distributions, where AF quantifies the variation of event occurrences across different time scales. Using surrogate analysis where the observed CM was compared to the CM of shuffled partners we found evidence for multi-scale coordination in both HFA-controls and controls-controls pairs. Moreover, CM was not significantly different between these two groups. There are two ways of explaining our findings. On the one hand, machine-mediated interactions might help in reducing the cognitive, sensory, and affective load coming from other dimensions in typical social interaction. Thus, autistic people might engage in the task very similar to healthy participants, consistent with previous research. On the other hand, the constraints of the virtual environment could enhance repetitive movement patterns, a well-documented autistic trait, making it easier for the healthy participants to adapt their trajectories to the patients, and ultimately achieve complex coordination without relying on social co-regulation of movement. Importantly, it could be that in both groups the healthy participants are doing all the adaptation and coordination. However, this is to be determined on the basis of extra analysis. This research highlights the multi-scale and distributed character of dyadic interaction, which marks it as a complex system. Additionally, it is consistent with recent investigations on CM and dyadic interactions. Remarkably, the PCE, although minimal, has proven to be an experimental setup capable of eliciting meaningful interactions and forms of alignment found in real-life social interactions. Furthermore, this paradigm allows researchers to systematically and quantitatively assess dyadic interaction in real-time, a feature which is crucial for understanding and studying online social interaction. Therefore, this work supports the implementation of human-computer interfaces and real-time paradigms in the context of psychopathologies when regarded as social interaction disorders.

ABSTRACT. Human and nonhuman animals often show flexible grouping patterns, in which temporary aggregations or subgroups come together and split, changing their size and composition in short time scales. While this flexibility confers clear advantages in exploring and exploiting heterogenous environments, it also increases the uncertainty faced by group members, who have to deal with social relationships using partial information in a frequently changing social environment. Here we develop and validate a method to measure the flexibility in composition of these subgroups using Shannon’s entropy, which captures the degree of predictability of the composition of a given subgroup over time. We formulate null expectations of entropy that consider subgroup size variation and sample size, against which the observed entropy can be compared. Using the theory of set partitioning, we also develop a method to estimate the number of subgroups that the group is likely to be divided into, based on the composition and size of the observed subgroups. We use these methods to analyze the composition of spider monkey, chimpanzee and gelada baboon subgroups. The three species are known for their high variation in subgroup size, but only the first two show highly flexible subgroup compositions, whereas geladas split their groups along more predictable lines. Our results serve to quantify the differences in social flexibility between these three species, which are consistent with the differences based on qualitative characterizations of their grouping dynamics. When measured at an individual level and for different interactions, our results can also be interpreted as the degree of social uncertainty faced by individual members of a given group. We discuss the implications of these results for the evolution of social complexity and cognition.

ABSTRACT. Many of the situations we deal with on a daily basis do not allow for an accurate prediction of their outcome and pose a potential loss of control of action, which in turn may lead to undesirable negative emotional states for the agents. And yet, we have learned to live with a constant conceptual indeterminacy, where previous experiential as well as rational knowledge are not enough to simulate or run symbolically an event and its possible implications. Situations with high degrees of uncertainty, such as an upcoming exam, a meeting with the boss, or a work presentation, to name a few, are perceived as an antagonistic force that impinges decisively on the emotional state of an agent and potentially inhibits its range of action. In social interaction, this antagonistic force may be minimized or controlled with the help of a second agent boosting a more assertive behavior and actions. In this talk we will weave together three different cognitive processes; forces, metaphoric thinking, and conceptual integration that evidence the complex interactional force dynamics that is present in linguistic instantiations such as the Mexican Spanish ¡Échale ganas!. The expression and its underlying enactive metaphoric structure help to promote the needed shared cognition between agents enabling them to strengthen agonist power, while stabilizing negative emotions when facing a conceptually indeterminate event. Our analysis supports the idea that an important part of our interaction depends on joining conceptual forces via languaging to minimize antagonistic actional forces, especially when the environment fails to provide us with all the information needed for a thoughtful analysis of the outcome of an uncertain event.

ABSTRACT. Breast cancer is a complex heterogeneous disease. Common hallmark features of cancer can be found. Their origin may be traced back to their intricate relationships governing regulatory programs during the development of this disease. To unveil distinctive features of the transcriptional regulation program in breast cancer, a pipeline for RNA-seq analysis in 780 breast cancer and 101 healthy breast samples, at gene expression and network level, was implemented. Inter-chromosomal relationships between genes resulted strikingly scarce in a cancer network, in comparison to its healthy counterpart. We suggest that inter-chromosomal regulation loss may be a novel feature in breast cancer. Additional evidence was obtained by independent validation in microarray and Hi-C data as well as supplementary computational analyses. Functional analysis showed upregulation in processes related to cell cycle and division; while migration, adhesion and cell-to-cell communication, were downregulated. Both the BRCA1 DNA repairing signalling and the Estrogen-mediated G1/S phase entry pathways were found upregulated. In addition, a synergistic underexpression of the gamma-protocadherin complex, located at Chr5q31 is also shown. This region has previously been reported to be hypermethylated in breast cancer. These findings altogether provide further evidence for the central role of transcriptional regulatory programs in shaping malignant phenotypes.

ABSTRACT. Reconstruction of a network structure from observations is an important problem relevant for many different areas such as neuroscience, physiology, climatology, genetics, ecology, etc. A group of established reconstruction techniques relies on analysis of the system response to a specially designed perturbation, i.e., on invasive measurements. However, often invasive measurement is not an option, e.g., in problems related to climatology, physiological studies, and medical diagnosis. In such cases one is restricted to analysis of observations of the free-running system.

In this work we develop a method of reconstruction relying only on observation of the free-running system. We address the case when the signals are spiky, namely, that the measurements between the spiking events are dominated by noise and only determination of the times of spikes is reliable. Hence, the data we analyze are spike trains and estimation of time-continuous phase is not feasible.

The reconstruction routine is of iterative nature. First, since we do not have any knowledge of the system yet, we evaluate the phase response curve (PRC) in the mean-field approximation, i.e., all-to-all equal coupling. Next, using the PRC estimate, we obtain an approximation of the network, which is then in turn used to obtain a better approximation of the PRC, and so on, continuing this procedure until the error of the reconstruction falls bellow a chosen value. Assuming that the outputs of all nodes are known and that the coupling between the elements is sufficiently weak to justify the phase dynamics description, we recover the connectivity of the network and properties of all its nodes. We perform thorough statistical analysis to quantify the robustness of our method.

Finally, we test our method on a network of 20 Morris-Lecar oscillators, to see how it behaves for a realistic neuronal model.

ABSTRACT. Background: The mechanisms underpinning the pathogenesis of asthma remain poorly understood. Given the complexity and heterogeneity observed at the organ level, characterising asthma requires computational approaches that offer tightly controlled variables and boundary conditions. Such models can offer mechanistic insights into drug pharmacodynamics and guide drug optimisation.

Objective: We aimed to develop a clinically-relevant computational model of airway remodelling capable of predicting the impact of therapeutic interventions targeting asthma at patient level. We hypothesised that capturing interactions between the epithelial, mesenchymal, and inflammatory elements within airways will capture airway remodelling, with the crucial implication that abnormal variations in model parameters will result in the hallmarks asthma.

Methods: The Flexible Large-scale Agent-based Modelling Environment (FLAME) was employed to model the airway. Cylindrical model of a Strahler Order 3 airway was developed, and featured epithelial (columnar and basal), mesenchymal (fibroblasts and muscle), and inflammatory (mast and eosinophil) agents. Rules captured interactions between these cells. Parametric analyses were conducted by altering model variables independently and collectively. Model validation was achieved by comparing model output – epithelial integrity, eosinophils/mm2 sub-mucosa, and muscle/mm2 wall area – against clinical data. The following markers reflected the hallmarks of asthma: <50% intact epithelium, >10 eosinophils, and >10% muscle mass. Significance was assessed using two-way ANOVA and Bonferroni multiple comparison. Each simulation was run equivalent to 6 months of physical time.

Results: Of the 25 boundary conditions studied only 4 captured all three hallmarks of asthma. Of these 4, the most parsimonious set of boundary conditions was selected as the virtual patient (42.8% epithelium; 22 eosinophils/mm2 sub-mucosa; and 26.5% muscle mass). The virtual patient was intervened with two therapeutic strategies: (i) eosinophil apoptosis, and (ii) reduced recruitment of inflammatory cells. The pro-apoptosis intervention reduced eosinophil activity by 54.1%, which agreed with the impact of mepolizumab (55% reduction in eosinophil activity). The anti-recruitment intervention reduced eosinophil activity by 81.4%, which agreed with the reduction in eosinophil activity following intervention with fevipiprant (79.6%). When employed to predict loss in muscle mass, the pro-apoptosis simulation showed a maximum relative reduction of ~35% for the highest ‘dose’. For the same ‘dose’ the anti-recruitment model reported a muscle loss of ~70%.

Conclusions: The model accurately predicted the impact of mepolizumab and fevipiprant on airway eosinophilia, and suggested that the intervention with a therapeutic that reduces eosinophil recruitment (over inducing eosinophil apoptosis) will be more successful in reducing muscle mass. This demonstrates the clinical and design significance of the model, and suggests the model as a useful tool to guiding drug development in respiratory medicine.

This study was funded by AirPROM.

ABSTRACT. Most biological functions are the result of coordinated interactions in groups of molecules, known as pathways. Typically, pathways are not isolated and can interact with other pathways through shared molecules, a phenomenon known as pathway crosstalk. Altered physiological states such as disease can lead to perturbations in these pathways. These perturbations not only affect the pathways themselves, but also the communication existing between them, which may be associated with the pathological features observed in disease. Therapeutic agents may correct this pathway deregulation, but at the same time they can also affect untargeted pathways, and cause further changes in the connectivity of pathways. We examined how an anti-diabetic agent pioglitazone alters the connectivity among functional pathways observed in a murine model of type 2 diabetes mellitus (T2DM). Pioglitazone, a peroxisome proliferator-activated receptor gamma (PPARG) agonist, is an insulin-sensitizing agent used for treating T2DM with various effects such as lipid/glucose lowering and anti-inflammation. We used RNA-Seq data in sciatic nerve tissues from our previous study, which included diabetic (db/db) either with or without pioglitazone treatment, and non-diabetic (db/+) mice as control. Normalized expression counts were used to quantify transcriptional deregulation at the pathway level in diabetic mice, either in the presence or absence of pioglitazone treatment, using a cutoff-free enrichment algorithm. Alterations in communication between pathways were evaluated through the enrichment analysis of the crosstalking sections of deregulated pathways, which was used to construct pathway perturbation networks. We identified significant perturbations in 91 pathways from the Reactome database in diabetic mice without pharmacological treatment. These pathways formed a network with 229 connections, which included a major subnetwork consisting of 61 pathways that are centered on neuronal system and transmission across chemical synapses, GABA and rhodopsin-like receptors, GPCR signaling, trans-membrane transport and ion channels. The pioglitazone treatment induced perturbations in 107 pathways, forming a network with 258 connections. Notably, a pathway associated to lipid and lipoprotein metabolism gained a higher degree-centrality in this pioglitazone-treated network, and pathways related to hemostasis and platelet activation emerged as “bridges” between neurological, signaling, developmental and metabolic pathways. In this work, we explored how pioglitazone modulates the communications between significant functional pathways altered in the context of diabetic neuropathy and induces rewirings that may indicate the importance of lipid metabolism and hemostasis pathways in the therapeutic action of pioglitazone in diabetic neuropathy.

ABSTRACT. The autonomic nervous system modulates the cardiac cycle through central (e.g. vasomotor and respiratory centers) and peripheral (e.g. arterial pressure and respiratory movements) oscillators. Parasympathetic modulation decreases the heart rate and cardiac contractility, whereas activity of the sympathetic branch opposes these effects and regulates peripheral vasoconstriction [1-4]. Autonomic nervous system has been found to break down under pathologic conditions like Diabetes Mellitus (DM) type II [5-9]. Thus, the correlation between blood pressure and heart rate can be used evaluate vagal and sympathetic activity, providing a non-invasive method for understanding autonomic control. An effective way to modulate blood pressure and heart rate is breathing rhythmically at 0.1 Hz. It is known that for control subjects, the respiratory resonant frequency of 0.1 Hz can induce periodic modulations in the cardiac rhythm with a frequency that depends on each person around 0.1 Hz [9]. Pathological conditions like DM breaks such modulation [8]. Our objective is to evaluate the vagal and sympathetic activity damage on diabetic patients using non-invasive measurements. For that, in this work, we compare for control subjects and DM patients, the correlations between systolic blood pressure (SBP) and interbeat intervals (IBI) records simultaneously taken by a portapress® [10] under rhythmic breathing at 1 Hz during 5 minutes. We found that the moments of the SBP and IBI fluctuations are statistically different between control and DM, thus it can be used as early-warnings auxiliary for diagnoses [7]. Moreover, there is a loss of the respiratory modulation of the heart rate as diabetes evolves [8].

Financial funding for this work was partially supplied by UNAM under grant DGAPA-PAPIIT IV100116. Thanks to Laboratorio Nacional de Ciencias de la Complejidad, México for the facilities to meetings of the working group.

References 1. Levy MN. Circulation Research 29:437–445, 1971. 2. Task Force of the European Society of Cardiology. European Heart Journal 17:354–381, 1996. 3. Robles-Cabrera A, et al. Revista de Neurología 59.11:508-516, 2014. 4. Estañol B, Rivera AL, et al. Physiological Reports 4.24:e13053, 2016. 5. Michel-Chávez A, et al. Arquivos Brasileiros de Cardiologia 105.3:276-284, 2015. 6. American Diabetes Association & American Academy of Neurology. Diabetes Care 18:59, 1995. 7. Rivera AL, et al. PLoS ONE 11.2:e0148378, 2016. 8. Rivera AL, Estañol B, et al. PLoS ONE 11.11:e0165904, 2016. 9. Ziegler D, et al. Diabetes Care 31:556-561, 2008. 10. Eckert S, Horstkotte D. Blood Pressure Monitoring 7.3:179-183, 2002.

ABSTRACT. Epithelium to mesenchymal transition (EMT), is complex phenomenon of cellular transdifferentiation through which an epithelial cell looses its characteristic phenotypic epithelial markers and becomes a mesenchymal cell which has the an ability to invade other tissues and which is resistant to chemotherapy. Under homeostatic conditions, EMT participates in wound healing, development and organogenesis. However EMT is also involved in the characteristic tissue remodelling of pathological chronic degenerative processes such as fibrosis and carcinomas, accounting for 51% of worldwide deaths. It remains to be investigated how EMT-driven tissue remodelling, needed for the maintainance of tissue homeostasis, becomes aberrant and a driving force of pathologies such as cancer and fibrosis. The EMT is controlled by the complex interactions between Transcription Factors, operating within individual cells in the tissue, and changes in the surrounding micro-environment, given by the composition of the extracellular matrix (ECM, which determines the stiffness of the tissue), and the levels of pro-inflammatory cytokines such as TGFb. In turn, the mesenchymal cells produce ECM components and cytokines, forming a positive feedback loop between the phenotypes of the cells and the properties of the surrounding tissue. While in equilibrium this complex feedback control structure preserves homoeostasis, we hypothesized that when the system is perturbed by genetic or environmental risk factors known to predispose to fibrotic or oncogenic conditions, the equilibrium between positive and negative feedback loops is impaired, which can lead to the onset and progression of pathology. Here, we use a systems biology approach in which we represent this complex multi-scale feedback control structure with a mathematical model. Using control theoretical approaches, we analyse this model to identify the different perturbations that can drive aberrant tissue remodelling processes. With this analysis, we were able to identify the different risk factor combinations that drive the transition from a homoeostatic to pathological tissue repair process in hepatocytes. Model simulations and analysis of the proposed core gene regulatory network attains only three steady–states, corresponding to the epithelial, mesenchymal and an intermediate senescent phenotypes observed during the EMT. We could also quantitatively characterise the different microenvironmental signals, in terms of the minimal input amplitudes and frequencies required to break the stability of the system and, forcing the transition from an epithelial to a mesenchymal state. Simulations of our multi-scale model show that an increased strength of in the positive feedback loop between the phenotypic decision-taking and the microenvironment cues can lead to an abrupt transition from a homeostatic to a pathological tissue with an over-accumulation of mesenchymal cells. In conclusion our analysis demonstrates how a system biology approach for the identification of underlying mechanisms in the onset and progression of complex diseases.

ABSTRACT. In recent years, the availability of detailed data on human behavior has lead to great advances in the understanding and modeling of phenomena such as natural disasters and epidemic outbreaks, which constantly affect the lives of the most vulnerable. At UNICEF Office of Innovation we are collecting, combining and analyzing data from private sector and open sources to generate insights so that UNICEF country offices and other stakeholders can take better informed decisions to respond to the needs of children worldwide.

ABSTRACT. The control of Tuberculosis (TB) is one of the largest endeavors undertaken by public health authorities. Recently, the development of global strategies for diagnosis and treatment optimization have led to TB burden decay worldwide to the point that the scientific community has dreamed with its eradication by 2035. Nonetheless, such goal is yet far away, and TB still constitutes a major Public Health problem, provoking 1.8 million deaths in 2015. Among all possible novel epidemiological interventions against TB, improved preventive vaccines hold the promise of offering substantial reductions of TB burden worldwide. Accordingly, several vaccine candidates are currently under development, each of which, depending on their immunogenic properties, might show differences in protection when applied to different age groups. In this context it would be required to develop a model for TB spreading capable of successfully describing the different age-dependent processes that affect the transmission and evolution of the disease. We present a TB epidemiological model which, capitalizing on publicly available data from different sources (World Health Organization (WHO), United Nations (UN) Population Division and published studies on contact-patterns surveys at an international scale), formalizes a data-driven description of most relevant coupling mechanisms between populations' age structure and TB dynamics. Our results shows that the global demographic shift projected by the UN for the next decades in certain regions is to be accompanied by a shift in the age-distribution of TB burden in some of the regions most affected by the disease. We show that for Africa and the East Mediterranean this could lead to revert the projected global decline of TB unless novel epidemiological measures are deployed. This phenomenon appears mainly as a consequence of a large increase of the infection prevalence in the eldest strata of the population. Also, a proper calibration of disease burden distribution across age groups and the incorporation of heterogeneous mixing patterns, yield to significantly different forecasts with respect to the state-of-the-art modelling framework. Regarding the comparison of different vaccination strategies, an adolescent-focused global immunization campaign appears to be more impactful than newborn vaccination in the short term. We demonstrated that TB indicators and vaccination strategies remarkably depend on how the disease dynamics is coupled to the demographic structure of the population. Capitalizing on a data-driven approach, we identified substantial biases in epidemiological forecasts that are rooted on an inadequate description of age-dependent mechanisms, among others. Our findings provide fundamental insights if novel age-focused epidemiological interventions, such as preventive vaccines, are to be considered and established.

ABSTRACT. Background: Pre-exposure prophylaxis (PrEP) can confer nearly 90% protection against HIV acquisition amongst men who have sex with men (MSM). If scaled up to sufficient coverage, PrEP could significantly reduce HIV transmission within MSM. However, increases in risk amongst MSM on PrEP resulting from lower perceptions of HIV-risk (risk compensation) could negate some of the benefits of PrEP, and may alter the epidemics of other sexually transmitted infections (STIs) among MSM. We explore the impact of PrEP, and possible consequences in terms of reduced condom use amongst PrEP users and changes in sexual mixing patterns on the HIV and hepatitis C virus (HCV) epidemics among MSM.

Methods: We developed a joint MSM HIV/HCV transmission model parameterised with UK behavioural data, which captures biological (heightened HCV infectivity and reduced spontaneous clearance among HIV-positive MSM) and behavioural heterogeneities (preferential sexual mixing by HIV-status, differences in condom use and risk heterogeneity). Consistent with UK guidelines, we classified MSM not using a condom within the past 6 months or having a regular HIV positive partner as eligible for PrEP, which applies to 25.9% of MSM from our data set; the European MSM Internet Survey (EMIS). Of these we assumed 50% would take-up PrEP. We considered two risk compensations: (i) Lower condom use by PrEP users with their partners, from 68% to 13% (reflective of the chance MSM without a HIV diagnosis use a condom with partners during anal intercourse to the chance two HIV diagnosed MSM); (ii) PrEP users no longer preferring to mix with the same HIV status; (iii) both scenarios combined. We ran our model to an endemic HIV prevalence of 5% and HCV prevalence amongst HIV infected MSM of 10%, and then projected the impact of PrEP (efficacy 86%) over 20 years under different scenario assumptions.

Results: Our model projects that PrEP use reduces HIV prevalence from 5.0% to 3.8% over 20 years when no compensatory risk behaviours are modelled, see Figure 1. However, HCV became more concentrated within HIV-positive MSM due to a smaller self-mixing group (rising to 11.9%) and HCV prevalence increased from 1.0% to 1.1% overall in the MSM population. If condom use reduced among PrEP users, the beneficial effect of PrEP was negated by 19.2%, resulting in a HIV prevalence of 4.0% after 20 years and 7% greater prevalence of HCV infections in the MSM population overall. When PrEP users no longer preferred to mix with HIV-negative MSM, the impact of PrEP on HIV prevalence was the same but the overall burden of HCV was reduced by 9% and became less concentrated in HIV positive MSM (prevalence reduced from 10% to 9.7%).

Conclusion: The advent of PrEP may not only shape the future of the HIV epidemic but change the pattern of HCV infection due to resulting behaviour change. Increased screening of HCV among PrEP users is likely warranted.

ABSTRACT. Obesity is linked to insulin resistance, high insulin levels, chronic inflammation, and alterations in the behavior of CD4+ T cells. Despite the biomedical importance of this condition, the system-level mechanisms that alter CD4+ T cell differentiation and plasticity are not well understood. We model how hyperinsulinemia alters the dynamics of the CD4+ T regulatory network, and this, in turn, modulates cell differentiation and plasticity. Different polarizing micro-environments are simulated under basal and high levels of insulin to assess impacts on cell-fate attainment and robustness in response to transient perturbations. In the presence of high levels of insulin Th1 and Th17 become more stable to transient perturbations and their basin sizes are augmented, Tr1 cells become less stable or disappear, while TGFβ producing cells remain unaltered. Hence, the model provides a dynamic system-level explanation for the documented apparently paradoxical role of TGFβ in both inflammation and regulation of immune responses and the emergence of the adipose Treg phenotype. Furthermore, our simulations provide novel predictions on the impact of the micro-environment in the coexistence of the different cell types, proposing that in pro-Th1, pro-Th2 and pro-Th17 environments effector and regulatory cells can coexist, but that high levels of insulin severely affect regulatory cells, specially in a pro-Th17 environment. This work provides a system-level formal and dynamic framework to integrate further experimental data in the study of complex inflammatory diseases.

ABSTRACT. We study a multilevel regulatory network of key inflammatory and metabolic cellular factors such as cytokines, transcription factors, cell receptors, hormones, etc., involved in inflammatory-related diseases. In the case of diabetes 2, we consider two connected networks for hepatocytes and beta pancreatic cells. The model is implemented in terms of continuos Boolean propositions satisfying the axioms of fuzzy logics. The stationary states of the dynamical system are consistent with patterns of health, metabolic syndrome, or diabetes disease. We determine the network nodes that mainly determine the transit between these states and their critical expression levels that induce these transitions.

ABSTRACT. Healthcare system is characterized by multitude of aspects (health, economics, organization of the process, technological and informational support, social aspects, behavior, etc.), stakeholders (society, government, insurance companies, various kind of institutions and organizations), actor (patients, physicians, nurses, pharmacy, etc.). Moreover, the system is continuously changing, has the explicit and implicit relationship between its elements, as well as significant uncertainty due to limited observations, high variation in scenarios, subjective and local decision making. To investigate the systems on various scales (country, city, hospital, department) a whole system should be considered to manage the uncertainty and reactively modify the model of the system in respect to changing conditions. We propose an approach to holistic modeling of the healthcare process to identify the multi-scale processes in the healthcare systems, where limited or weak structured observations are available. Within our cases, the observations are mainly represented by electronic health records of in- and out-patients of selected hospital(s). In such conditions, we apply a combination of process mining, data mining, and text mining techniques to identify the structure of the process. The model of the complex process can be considered as an extended clinical pathway, describing multi-aspect, multi-scale dynamic complex process. The identified model of the system and complex processes cover the aspects mentioned earlier and serve as a basic framework to incorporate models of various aspects, parts, and sub-processes of the system. This includes models for simulation, data-driven prediction models, surrogate models for unknown aspects of the system, etc. For example, it incorporates agent-based modeling of activity within a hospital, models for prediction of length of stay or disease complications, etc. The model of the complex process can be used in various ways: comparison of different solutions, decision support (for different decision makers), optimization of hospital structure, etc. The key objective of the proposed approach is forming a framework for the model-based assessment of healthcare systems within a context of value-based approach, where the important role is played by resource usage and life quality improvement. We consider a modeling and simulation of complex health care process as a way to assess resources of the system which have high uncertainty, lack of observations, and undefined (or partly defined) structure. Moreover, incorporation of social, psychological, and behavioral aspects of the system enable quantitative assessment of personal life quality. Finally, identified and controllable (within a simulation) diversity of patient flow extends the basic approach with a new level of personalization on different levels of abstraction where even rare cases, events, and conditions are considered and analyzed. To validate the proposed approach, we consider a flow of patients with selected nosologies observed in 2010-2016 in Federal Almazov North-West Medical Research Center (Saint Petersburg, Russia), one of the leading cardiological research centers in Russia, and described with electronic health records. The selected nosology includes in-patients with acute coronary syndrome and planned stent implantation, out-patients with arterial hypertension. The solution considers a group of key hospital departments involved in the treatment as a system which has the issues mentioned earlier.

ABSTRACT. Dense monolayers of living cells display intriguing relaxation dynamics, reminiscent of soft and glassy materials close to the jamming transition, and migrate collectively when space is available, as in wound healing or in cancer invasion. Here we show that collective cell migration occurs in bursts that are similar to those recorded in the propagation of cracks, fluid fronts in porous media and ferromagnetic domain walls. In analogy with these systems, the distribution of activity bursts displays scaling laws that are universal in different cell types and for cells moving on different substrates. The main features of the invasion dynamics are quantitatively captured by a model of interacting active particles moving in a disordered landscape. Our results illustrate that collective motion of living cells is analogous to the corresponding dynamics in driven, but inanimate, systems.

ABSTRACT. Many real complex systems cannot be represented by a single network, but due to multiple sub-systems and types of interactions, must be represented as a multiplex network. This is a set of nodes which exist in several layers, with each layer having its own kind of edges, represented by different colours. An important fundamental structural feature of networks is their resilience to damage, the percolation transition. Generalisation of these concepts to multiplex networks requires careful definition of what we mean by connected clusters. We consider two different definitions. One, a rigorous generalisation of the single-layer definition leads to a strong non-local rule, and results in a dramatic change in the response of the system to damage. The giant component collapses discontinuously in a hybrid transition characterised by avalanches of diverging mean size. We also consider another definition, which imposes weaker conditions on percolation and allows local calculation, and also leads to different sized giant components depending on whether we consider an activation or pruning process. This 'weak' process exhibits both continuous and discontinuous transitions.

ABSTRACT. The organization of constituent network layers to multiplex networks has recently attracted a lot of attention. Here, we show empirical evidence for the existence of relations between the layers of real multiplex networks that go beyond degree correlations. These relations consist of correlations in hidden metric spaces that underlie the observed topology. We discuss the impact and applications of these relations for trans-layer link prediction, community detection, navigation, game theory, and especially for the robustness of multiplex networks against random failures and targeted attacks. We show that these relations lead to fundamentally new behaviors, which emphasizes the importance to consider organizational principles of multiplex networks beyond degree correlations in future research.

ABSTRACT. We show that real multiplex networks are unexpectedly robust against targeted attacks on high degree nodes, and that hidden interlayer geometric correlations predict this robustness. Without geometric correlations, multiplexes exhibit an abrupt breakdown of mutual connectivity, even with interlayer degree correlations. With geometric correlations, we instead observe a multistep cascading process leading into a continuous transition, which apparently becomes fully continuous in the thermodynamic limit. Our results are important for the design of efficient protection strategies and of robust interacting networks in many domains.

ABSTRACT. We present a continuous formulation of epidemic spreading on multilayer networks using a tensorial representation, extending the models of monoplex networks to this context. We derive analytical expressions for the epidemic threshold of the susceptible-infected-susceptible (SIS) and susceptible-infected-recovered dynamics, as well as upper and lower bounds for the disease prevalence in the steady state for the SIS scenario. Using the quasistationary state method, we numerically show the existence of disease localization and the emergence of two or more susceptibility peaks, which are characterized analytically and numerically through the inverse participation ratio. At variance with what is observed in single-layer networks, we show that disease localization takes place on the layers and not on the nodes of a given layer. Furthermore, when mapping the critical dynamics to an eigenvalue problem, we observe a characteristic transition in the eigenvalue spectra of the supra-contact tensor as a function of the ratio of two spreading rates: If the rate at which the disease spreads within a layer is comparable to the spreading rate across layers, the individual spectra of each layer merge with the coupling between layers. Finally, we report on an interesting phenomenon, the barrier effect; i.e., for a three-layer configuration, when the layer with the lowest eigenvalue is located at the center of the line, it can effectively act as a barrier to the disease. The formalism introduced here provides a unifying mathematical approach to disease contagion in multiplex systems, opening new possibilities for the study of spreading processes.

ABSTRACT. In a world more connected and digitalized than ever, the steadily growing availability of data, along with the advances in fields such as Data Science, Complex Systems, Artificial Intelligence or Computational Sociology have profoundly change whole businesses such as marketing and advertising, revolutionizing related areas such as political campaigns or governance. How can we leverage these advances for the most vulnerable? How can we make sure that the existing data divide does not increase the gaps of inequality? How can we make the invisible visible and heard? How can we push for algorithmic and scientific advances to have equity and the most vulnerable at their core? How can we integrate these advances into the humanitarian and development systems?