An ongoing need exists for an enhanced toolkit for predicting spatial patterns of pathogen transmission [1–4]. While current models incorporate spatial aspects of the host [5, 6], pathogen [7–9], or more rarely, both [10, 11], many current models of infectious disease ignore the more complex landscape features, including interactions between hosts, which can be mitigated or facilitated by landscape complexity [12, 13]. Pathogen transmission potential is an integrated measure of both infectivity and an individual’s opportunity for encountering the pathogen in the environment or through contact with other infectious individuals . Therefore, models of pathogen infection must examine this transmission potential and focus on how landscape features directly influence this potential and the resulting patterns of pathogen spread. The continued shift in research emphasis towards efforts concentrating on the underlying ecological determinants and spatial dynamics of pathogen transmission will result in more effective global public health policy [15–17].
Employing geographic information systems (GIS) data as a tool in epidemiologic analyses is not new, given the ability of GIS to incorporate spatial and non-spatial data in one system . Colwell and colleagues (1996) successfully implemented research programs using GIS data of Bangladesh to more completely understand the transmission of Vibrio cholerae by modeling it as a component of the environment. Outbreaks were shown to be both seasonal and geographically localized, influenced strongly by the presence of estuaries and major rivers . Modeling of pathogen transmission and spread of infectious diseases with a focus on GIS analysis has been undertaken in several outbreaks and epidemics, including plague (Yersinia pestis) in the Southwestern United States, rabies in Trinidad, and Chagas disease vectors in Colombian villages [20–22]. These studies demonstrate that analysis of pathogen transmission patterns is enhanced through the flexibility in analyzing spatial data inherent to the GIS system.
Recently, agent-based models (ABMs), or individual-based models, have been effectively employed as an enhanced tool to address the spatial dynamics of pathogen transmission [7, 23, 24]. These models explicitly represent individual entities in the system under study and can realistically accommodate extreme heterogeneity among the agents by allowing individuals to incorporate spatial interactions into the simulations directly . This flexibility permits ABMs to account for population outliers and long-tailed distributions and to model rare, albeit important, events in the system under study . Agent-based modeling is therefore ideal for addressing complex questions regarding how hosts and pathogens navigate a complex landscape. Recently developed ABMs have been used to elucidate infectious disease dynamics in systems as disparate as demonstrating the process of granuloma formation following a tuberculosis infection , evaluating influenza vaccination strategies in Italy, with a focus on implementation campaigns mitigating a global pandemic to H5N1 , and understanding the relationship between vector ecology, human behavior, and spread of African sleeping sickness .
Host behavior and ecology
Macaque species are found throughout Asia and in parts of Africa, with the fascicularis subgroup having an extensive range throughout much of Southeast Asia. Long-tailed macaques (Macaca fascicularis) thrive in a variety of habitat types, including forests, grasslands, semi-deserts, and most especially, urban landscapes , often living commensally with humans. While macaques are generally considered to be frugivorous, long-tailed macaques are known to have a highly flexible diet and can be considered, in parts of their range, to be omnivorous. Male dispersal is common, while females remain in their natal group. Little is known about dispersal duration or distance ; however, long-distance dispersals have been documented . Gene flow between population groups is maintained by male dispersal as well as by group fission events, especially common as population size increases. Thus, long-tailed macaques thrive in complex, anthropogenic landscapes and can disperse across wildly variable habitats.
On the island of Bali, Indonesia, a system of temple complexes act as core use areas for long-tailed macaques (Macaca fascicularis) [29–32]. While the macaques’ home ranges extend well beyond the confines of the temple complexes, a substantial segment of a given population can be found in and around these temples on a regular basis. Dispersing male macaques may act as both units of gene flow between seemingly isolated macaque populations and as mechanisms of pathogen transmission across the island . Human land use patterns have resulted in a mosaic of riparian forest, small forest patches, agricultural lands, and urban areas across much of the island. The broad distribution of macaque populations on Bali suggests that macaques use this human-modified landscape by exploiting agriculturally-dominated, riverine links between populations for dispersal and the sanctuary nature of temples as stabilized food resources . This protection and resource availability has allowed macaques to exist in moderately high densities alongside high human densities .
Pathogen ecology and epidemiology
Gastrointestinal parasites are among the most prevalent suite of parasites and pathogens globally, with representatives found in nearly all mammal species and causing morbidity in nearly all individuals at some point in their lifetime . The success of this suite of parasites is due, in large part, to their mode of transmission. Relying on the fecal-oral route and often occurring with environmentally stable infective stages, infectious agents pass through the gut of an infected individual, are deposited in water or on plant matter, and are ultimately consumed, completing the transmission cycle . The environmentally stable infective stage makes the spatial transmission of gastrointestinal parasites of special relevance. Landscape type and quality have been shown to be important in the prevalence and intensity of intestinal parasites [35–37]. For example, intestinal parasite burden was significantly greater in low quality, fragmented habitat in populations of two species of howler monkeys (Alouatta palliata and A. pigra) .
It is estimated that more than 500 million people are infected with at least one species of Entamoeba at any given moment . Infection rates increase with impoverished economies and lack of access to clean drinking water. Both Entamoeba histolytica and E. dispar, along with at least two other amoebas (Iodamoeba and Endolimax) infect humans, domestic animals, and wildlife species, including non-human primates [34, 35, 38, 39]. While E. histolytica is linked to numerous cases of diarrhea and more than 100,000 human deaths/year, E. dispar is largely un-symptomatic, causing neither disease nor tissue degeneration . Both species of Entamoeba have been found in macaques throughout their range, including on the island of Bali, Indonesia [35, 41, 42]. The similarity in transmission strategy and phylogeny coupled with highly disparate disease severities makes E. histolytica and E. dispar an ideal model system for examining the effect of landscape variability on host dispersal and pathogen transmission.
Modeling host movement and pathogen transmission
LiNK, the ABM presented here, incorporates landscape features critical to understanding pathogen transmission patterns by using GIS layers of the actual system’s landscape [29, 43]. The powerful spatial analysis permitted through the use of GIS data combined with the strength and utility of ABM provides a mechanism to understand the spatial context critical for understanding patterns of pathogen transmission. LiNK has the ability to generate predictions regarding host dispersal and pathogen distributions based on the anthropogenic landscape, human-wildlife interactions, host behaviors and interactions, and pathogen life histories at island-, population-, and individual levels.
Here, we present an agent-based model of host (macaques) and pathogen (gastrointestinal parasites) movement through the Bali landscape. First, we aim to determine the impact of the inclusion of landscape information on patterns of macaque dispersal. We hypothesize that the inclusion of landscape information into our model will alter the dispersal pattern of macaques from isolation by distance, as predicted in the absence of landscape information, to one of dispersal linked by habitat type. We then compare the difference between modeled dispersal patterns generated with the inclusion of landscape information to that of actual macaque gene flow patterns, as measured by genetic distance. We hypothesize that the inclusion of landscape information into our model of macaque dispersal will correlate better to measured genetic distance than when landscape information is excluded. Finally, we explore the likely path and rate of pathogen transmission of two gastrointestinal parasites – Entamoeba histolytica and E. dispar – modeled using varying pathogen virulence, infectivity, and infectiousness parameters. We hypothesize that the inclusion of landscape information will result in environmental context dependence in rate and route of infection, with landscape heterogeneity mitigating overall infection. We also hypothesize differences between the two parasites independent of landscape features with the less virulent parasite – E. dispar – reaching overall greater distances from the site of initial infection due the host’s ability to maintain dispersal patterns as though healthy.