Seeds clustering and sentinel farm identification for disease spreading on dynamical cattle trade networks
The propagation of directly transmitted diseases in a population crucially depends on the pattern of interactions among hosts and on its time evolution. Being able to identify the most vulnerable elements of the population is critical to disease control and important findings have been provided in this direction by network approaches analyzing the structure of the population interactions. The temporal dimension characterizing the pattern of interactions, however, introduces an additional difficulty in assessing the consequences of an outbreak and the potential spreading risks of the nodes of the system, thus hindering the development of efficient containment and eradication strategies. Here we address these aspects by focusing on the applied problem of the spread of livestock diseases, supported by a detailed dataset describing cattle displacements among premises in Italy at the individual level, with a daily temporal resolution. Using a networked model approach for disease simulations, where nodes are premises and a directed edge is drawn between two nodes whenever a displacement of bovines occurs between the corresponding premises, we assess the role of initial conditions in generating an outbreak and classify the seeds into clusters leading to similar disease invasion pathways. We investigate the temporal stability of the clusters and put forward a novel procedure to identify specific nodes (premises) that should be monitored as disease sentinels. Such sentinels are more likely to be infected if an outbreak occurs, and provide critical information on the origin of the infection. The proposed characterization of the nodes can be exploited for epidemic risk assessment and inform the design of optimal surveillance systems.
Scott Duke-Sylvester -University of Louisiana, Lafayette
Recovering epidemic processes from the molecular evolution of the pathogen
Microorganisms that cause infectious diseases evolve on a time scale that is compatible with the spread of infection. As a result, the rapidly emerging phylogeographic structure of a pathogen is shaped by the rise and fall in the number of infections. I present results from a model combining ecological and evolutionary processes. I use my model to explore the effects of different rates of host movement and different rates of infection. I find that there are several linearizable relationships between the rate of host movement and measures of phylogenetic structure. I also find that it is possible to recover the period of an epidemic from the viral phylogeny as well as detect different periods associated with different rates of infection. My approach extends the reach of current approaches to phylogenetic analysis by linking phylogeographic patterns back to specific ecological processes. These results suggest the possibility of resolving additional details of an unobserved epidemic from sequence data derived from contemporary viral samples.
Ken Linthicum - USDA, ARS
Advances in Forecasting Emerging Vector-Borne Diseases
Population growth, frontier agricultural expansion, and urbanization transform the landscape and the surrounding ecosystem, affecting climate, diseases, and interactions between animals and humans. Global climate greatly influences local conditions that can affect emerging vector-borne disease patterns because the pathogens, their vectors, and hosts are sensitive to temperature moisture, and other ambient environmental conditions. In this presentation we examine in detail linkages between climate, ecosystems and elevated transmission of dengue, chikungunya, Rift Valley fever, and malaria. Several examples will be used to demonstrate how global climate and the environment can be used to forecast emerging vector-borne diseases. First, we will describe how temperature plays a major role in the ability to forecast how Aedes aegypti mosquitoes transmit dengue virus in Southeast Asia and possibly chikungunya virus in Africa and Asia, and how Anopheles mosquitoes transmit malaria in the Republic of Korea. Second, we will describe how rainfall permits us to forecast when and where Aedes and Culex mosquito species transmit Rift Valley fever in sub-Saharan Africa and Middle East. During periods of elevated transmission there is a significantly increased risk of globalization of these and other arboviruses. The ability to predict periods of elevated risk permits us to design better prevention, containment, or exclusion strategies to limit globalization of emerging pathogens.
Huijuan Wang - Boston University
Dynamics on networks characterized by spectral radius
The largest eigenvalue of the adjacency matrix, also called spectral radius, is a powerful characterizer of dynamic processes on networks such as virus spreading and synchronization processes. We investigate how to decrease the spectral radius, by removing links. Furthermore, most networks in reality are not isolated, but instead, interact with other networks by nature or to achieve synergy in their functioning. Such mutual interactions as well influence the dynamic processes on the coupled networks. A disease, for example, may spread among two species via the contact network of each species and the interactions in between. We will discuss further the effect of adding interacting links on the spectral radius.
Duygu Balcan - ISI Turin Italy
Phase transitions in contagion processes mediated by recurrent mobility patterns
Human mobility and activity patterns mediate contagion on many levels, including: spatial spread of infectious diseases, diffusion of rumors, and emergence of consensus. These patterns however are often dominated by specific locations and recurrent flows and poorly modeled by the random diffusive dynamics generally used to study them. Here we develop a theoretical framework to analyze contagion within a network of locations where individuals recall their geographic origins. We find a phase transition between a regime in which the contagion affects a large fraction of the system and one in which only a small fraction is affected. This transition cannot be uncovered by continuous deterministic models due to the stochastic features of the contagion process and defines an invasion threshold that depends on mobility parameters, providing guidance for controlling contagion spread by constraining mobility processes. We recover the threshold behavior by analyzing diffusion processes mediated by real human commuting data.
Steve Warren -Kansas State University
Wearable/Ingestible Sensors to Acquire Animal-Level State of Health in Cattle Herds
Continuous animal-level, state-of-health measurements in large, roaming herds offer the potential to validate and challenge multiscale models for disease epidemiology. Each animal species offers its own data acquisition challenges. When considering such measurements in cattle, vital sign parameters such as heart rate and core body temperature come to mind, as they provide well-understood clinical precursors for illness and disease. However, obtaining these data in the field using traditional methods is time and labor intensive, which speaks to the need for solutions that provide continuous and automatic parameter acquisition. This presentation addresses wearable/ingestible sensors and supporting wireless telemetry technologies that in aggregate offer the potential to provide substantive, animal-level data useful for herd health management and computational model validation.
Mike Sanderson - Kansas State University
Multiscale Modeling issues for E. coli O157 in Cattle
Escherichia coli O157:H7 is an important enteric disease of humans with a significant reservoir in cattle and their environment including feed, water and the pen surface. It causes no disease in cattle and thus the ultimate concern is contaminated meat products and the subsequent human disease. It is assumed that decreasing shedding in cattle will decrease disease burden in humans. Modeling of E. coli O157 in cattle production systems has generally involved the use of an SIS type model tracking the transition between states of the cattle. Heterogeneity of transmission rates and infectious periods has been included between animals. Interaction of animals with the environment and transmission of O157 through the environment have also been modeled. Recently fitting of models to cross-sectional prevalence data has produced a hypothesis that some small proportion of cattle (8-10%) shed at higher levels and termed these individuals “super-shedders”. The term has been loosely defined but generally includes individuals that shed at greater than 103-104 cfu/g usually based on one time sampling. These “super-shedders” have been reported to account for the majority of O157 transmission, contamination of the herd environment (>90%) and the risk for human disease.
Alternately, experimental inoculation data with varying doses suggest that many or most cattle may shed at high concentrations for a period of time during their shedding cycle. Cross-sectional data may be unable to distinguish between the possibility of a few cattle shedding at high concentrations while most shed at low concentrations, and the temporal dynamics of shedding concentrations within cattle where all or most cattle may shed at high concentrations for some period of time during their shedding cycle. This issue is of more than passing academic interest. If a few “super-shedders” within a herd are responsible for the majority of transmission and contamination of the environment then identification and removal of these cattle is central to control of the pathogen within the herd and subsequently to reducing human health burden. However, if most cattle shed at high levels for some period then searching for “super-shedders” will not be effective and control measures will need to be applied at the herd and or environmental level to result in effective control.
To study this possibility we have implemented a multiscale model of fecal shedding concentration dynamics for individual cattle integrated with a pen-level E. coli O157 transmission model. This allows us to study the temporal shedding concentrations of individuals over time and the aggregated shedding concentrations and prevalence at the pen level. A discrete-time, stochastic individual-based model tracks daily shedding concentration of individual cattle which are linked to environmental levels of O157 resulting in exposure and shedding in additional cattle, assuming the pathogen is transmitted indirectly (i.e., from environment to cattle). Tracking of daily prevalence at different concentrations allows us to summarize the shedding pattern of the pen and the estimate the daily cross-sectional prevalence of each shedding concentration.
Initial results indicate that even with no variation in individual shedding (all cattle have identical shedding concentration curves different only by the start time) there is marked variation in daily fecal O157 concentrations including counts in excess of 104. This indicates super-shedders may not be necessary to produce observed cross-sectional data and that modeling of individual variability may be necessary to capture the dynamics of E. coli O157 transmission and understand control efforts.
Tim Carpenter - University of California, Davis
Simulating the spread and control of foot-and-mouth disease, from one to a million plus
Animal agriculture in the US is comprised of livestock housed on more than one million premises in all 50 states. When considering the spread and subsequent control of diseases in this population, it is important to consider the individual animal located on and moving among these various livestock premises. An individual-based, spatial, stochastic simulation model, the Davis Animal Disease Simulation (DADS) model simulates the disease process in the individual animal in a herd. It then simulates the spread of disease within the herd, as well as movement of infected animals and potentially infectious contacts by humans, machinery and other fomites to neighboring as well as distant herds. With this information, alternative control and surveillance strategies may be simulated to evaluate their value under a variety of scenarios. The underlying mechanisms of this process and simulation examples of the DADS model will be presented.
Morgan Scott - Kansas State University
Wanted: robust mathematical models of enteric bacterial resistance to antimicrobials
The current paradigm of infectious diseases modeling tends to lean heavily on the SEIR framework (or some derivative thereof). These models generally assume that there exists a temporally fluid mixture of individuals in a defined host population that are susceptible to, exposed to, infected with or recovered from pathogens such as bacteria. When extended to the issue of resistance among enteric bacteria to antimicrobials (whether in human medicine or animal agriculture) these models rapidly lose their appeal since the ‘infection’ or ‘colonization’ of all hosts with a resistant strain of bacteria is typically the rule, rather than the exception. Instead, explicitly quantitative ecological models that estimate the bacterial load (e.g., colony-forming units per gram of feces) and that allow for competition among antimicrobial-resistant and susceptible strains for finite host-gut resources (e.g., nutrients, space) are much more useful in predicting short- and medium-term responses of gut bacteria to perturbations such as antimicrobial use or even dietary and environmental changes. Unfortunately, these models also have shortcomings since most predict a return to ‘baseline’ levels of resistance following the removal of any such perturbations. That the populations return to baseline isn’t inherently problematic; that the baseline itself imperceptibly changes (rises) over much longer periods of time is a problem. In fact, the current risk assessment / risk management paradigm adopted by regulatory agencies largely fails due to its ignorance of the long-term adaptive nature of bacterial populations to evolve in response to cumulative selection pressures. Modeling current antimicrobial use / current antimicrobial resistance is naïve at best, and harmful at its worst. My presentation will not by itself offer a quick-fix solution to the obvious problems that are present in contemporary resistance modeling; rather, I will provide a bucket list of desired capabilities for future modeling efforts and point out the strengths and applications that each of SEIR, ecological patch (metapopulation), and evolutionary microbiological modeling still can bring to this important interface of science and policy.
Bryan Lewis - Virginia Tech
Applications of Highly Detailed Simulations
The Network Dynamics and Simulation Science Laboratory at the Virginia Bioinformatics Institute has worked on highly detailed simulations of real-world systems for over a decade. The detailed approach to simulation is useful for studying the processes involved in complex phenomena. Representing all individual elements in the real-world system allows the flexible integration of a wide variety of data sources. Using similar levels of resolution and structure as the real-world system also facilitates the analysis of the simulation. These advantages come with the cost of computational complexity as well as time-consuming customization of each simulated system. Recent developments have allowed these costs to be diminished.
Recent work has focused on the use of simulations for public health decision making, providing frameworks for evaluation of public health evaluation methodologies, and developing environments to facilitate collaborative decision making. Simulation studies have focused on the importance of representing individual behaviors to the overall behavior of epidemics. Infrastructure development has focused on generic simulation configuration and management, enabling web-service directed simulation on high performance computing platforms, and digital library development for managing large and complex collections of simulation results. Future plans include the incorporation of the human-animal interface into these frameworks.
Alex Vespignani - Indiana University
The global epidemic and mobility model: a computational approach to infectious disease spreading
Computational models play an increasingly important role in the assessment and control of public health crises, as demonstrated during the 2009 H1N1 influenza pandemic. We present the global epidemic and mobility computational model, a publicly available software system that simulates the spread of emerging human-to-human infectious diseases across the world. The tool integrates worldwide high-resolution demographic and mobility data and its design aims at maximizing flexibility in defining the disease compartmental model and configuring the simulation scenario and allows the user to set a variety of parameters including: compartment-specific features, transition values, and environmental effects. We discuss the avenues as well as the major roadblocks to the extension of this approach in the field of zoonotic epidemics and animal infectious diseases.
Lee Cohnstaedt - USDA, ARS
Enhancing landscape modeling with population genetics data
The impact of environment on the abundance, distribution and movement of organisms may be assessed by combining population genetics analysis with continental or landscape level modeling. Population genetic analysis requires collecting the organism of interest throughout the study area and finding genetic differences between and within populations. This picture of the extant genetic structuring can then be combined with geographic information systems (GIS) data or remotely sensed environmental data to correlate past, present, and future organism distributions with biotic and abiotic factors. Examples of this combined analysis are demonstrated using the sand fly, Lutzomyia verrucarum which is a leishmaniasis disease vector in Peru. A mitochondrial population genetics analysis is combined with a GIS analysis of abiotic factors to explain identify population structuring, barriers to migration, migration rates, historical movement trends and the correlation of leishmaniasis prevalence with sand fly genotypes. A more powerful analysis using next generation sequencing on a North American continental scale will also be discussed.
Karen Garrett - Kansas State University
Multi-scale modeling of the effects of climate variability on disease risk
In addition to changes in the mean of climate variables such as temperature and precipitation, increased climate variability is predicted for many areas of the world. Changes in the variation of weather drivers can have important effects even if the mean remains unchanged. We evaluated the effects of climate variability in simulated epidemics. We developed a model of agricultural yield loss as a function of weather that demonstrates two qualitatively different results for changes in system variance. If median conditions are conducive, increasing system variance decreases mean yield loss. On the other hand, if median conditions are intermediate or poor, such that conditions are conducive no more than half the time, increasing system variance increases mean yield loss. If hosts are meshed in a landscape with spatial autocorrelation in disease risk, the results are more pronounced. A linked model of decision making based on past or current information about yield loss also shows changes in the performance of decision rules as a function of system variance.
Ling Xiu - Kansas State University
Network-based metapopulation model for Rift Valley fever
Rift Valley fever is a vector-borne disease which affects ruminants and humans. Rift Valley fever virus (RVFV) has been expanding its geographical distribution with important implications for both human and animal health. The emergence of Rift Valley fever (RVF) in the Middle East, and its continuing presence in many areas of Africa, has negatively impacted both medical and veterinary infrastructures and human morbidity, mortality, and economic endpoints. Furthermore, worldwide attention should be directed towards the broader infection dynamics of RVFV, since suitable host, vector and environmental conditions for additional epidemics likely exist on other continents; including Asia, Europe and the Americas. We propose a new compartmentalized model of RVF and the related ordinary differential equations to assess disease spread in both time and in space; with the latter driven as a function of contact networks. Humans and livestock hosts and two species of vector mosquitoes are included in the model. The model is based on weighted contact networks, where nodes of the networks represent geographical regions and the weights represent the level of contact between regional pairings for each set of species. The inclusion of human, animal, and vector movements among regions is new to modeling RVF. The movement of the infected individuals is not only treated as a possibility, but also an actuality that can be incorporated into the model. We have tested, calibrated, and evaluated the model using data from the recent 2010 RVF outbreak in South Africa as case study; mapping the epidemic spread within and among three South African provinces. An extensive set of simulation results shows the potential of the proposed approach for accurately modeling the RVF spreading process in additional regions of the world. The benefits of our proposed model are twofold: not only can our model differentiate the maximum number of infected individuals among different provinces, but also it can reproduce the different starting times of the outbreak in multiple locations.
Hong Liu - Kansas State University
Epirur_Cattle: an agent-based model of beef cattle movements in U.S.A.
Epirur_Cattle is an agent-based simulator for synthesizing beef cattle movements and forecasting zoonotic disease spread in American beef cattle industry. The main contribution of our work is to give realistic beef cattle movements and epidemic predictions in lack of publicly available fine-granularity data on cattle trades and transports. The Epirur simulates the entire life cycle of individual beef cattle. There are two categories of agent models: one is the model of intelligent operators in beef cattle industry, including cow/calf farmers, backgrounders/stockers, feedlot owners, and meat-packing factories, the other is the model for cows and beef cattle. From the information we gathered from interviewing local farmers and beef cattle experts, we find that the key rules that operators adopt to make decision on whether to buy or sell are the climate and the quality of the cattle. The transactions between two types of operators through sale barns, together with private pacts between operators or Internet auctions, form the long-distance links between cattle. Except the few days moving between premises, cattle agents gain weights and walk within each of their confinements. Spatial proximity forms the contact networks. The agent models are tested both by the statistics of cattle contact rates and by the characteristics of the contact networks.
Berhanu Tameru - Tuskegee University
The Role of the Epidemiologic Problem Oriented Approach (EPOA) Methodology in the Development of Computational Models of Diseases
Epidemiologic research involves the study of a complex set of host, environmental and causative agent factors and determinants as they interact to impact health and diseases in any given population. A detailed understanding of the epidemiology of a given disease provides the essential framework for model development and enables the laying out of the comprehensive and fundamental structures for the models. The EPOA methodology was utilized to illustrate on how the knowledge bases for different diseases and pests of plants, animals and humans such as HIV/AIDS, are developed. Using information from various sources, the EPOA triads are decomposed into their respective pillar variables and parameters. In the problem identification and characterization triad: the agent pillar identifies the causative agent and its characteristics; the host pillar identifies and characterizes the host (plant/animal/human/); and the environment pillar characterizes the physical, biological and socioeconomic environments for both the host and agent. In the problem management/solution triad: the therapeutics/treatment pillar considers the treatment options for the pest or disease; the prevention/control pillar considers prevention and control measures; and the health maintenance/promotion pillar considers measures for the health maintenance and promotion of the given population. The dynamics of the relationships and interactions between the pillars are described by means of ordinary/partial differential equations.
Models for diseases such as HIV/AIDS can be conceptual, in vivo or in vitro, systems analysis, mathematical, or computational just to name a few. The knowledge bases developed using the EPOA methodology provide well-organized structured sources of information, which are used in the variable and parameter estimations as well as analysis (biological, mathematical, statistical and computer simulations), which are all crucial in the epidemiologic modeling of diseases and pests. The EPOA methodology has become an important tool in the development of models that are vital and informative in the decision making process in the public health arena.