-Accurately representing demographic realities is a critical component in creating useful, agent-based epidemiological models of infectious disease. Synthetic ecosystems are generated from Census data microsamples in a statistically-sound manner to maintain population-level demographic characteristics. These highly detailed representations of populations are the basis of many advanced simulations of infectious disease epidemics. Creating a standard, machine-readable representation of synthetic ecosystem data would enable easier use and integration with epidemic simulator software. Here we describe an ontologybased representation in Resource Description Framework (RDF) and Web Ontology Language (OWL) of version 1.0 of the 2010 U.S. Synthetic Population database by RTI International. Our representation draws upon applicable classes from several reference ontologies, including the Ontology of Medically Related Social Entities (OMRSE). After failing to find suitable ontological representations of several key data elements in the Synthetic Population dataset, we created new classes in OMRSE for representing employment status, employee roles, workplaces, residences, households, and age measurements. We loaded a test RDF dataset (structured according to ontologies in OWL) of synthetic individuals into a commercial triple store (Stardog) and validated the representation with SPARQL queries.
I. BACKGROUND
Disease transmission models (DTMs) are epidemiological
models that predict the future course of infectious disease
outbreaks under various assumptions. They are used to study
which strategies for controlling outbreaks are potentially the
most effective and thus for decision making during the course of
an outbreak. For example, researchers have used them to study
the effects of various vaccination control strategies on pandemic
influenza [
Agent-based disease transmission models (AB-DTMs) are a
class of DTMs that represent every host individual in the
population of interest, and sometimes individual vector
organisms as well, to increase the realism of the simulations and
thereby increase the accuracy of the predictions generated [
Census data are a key resource for matching simulated
populations to actual ones [
In addition to expanding microsample Census data to full
population size, researchers incorporate significant additional
information into synthetic ecosystems relevant to disease
transmission [
Researchers typically make these synthetic ecosystems
available as delimited text files in a format suitable for loading
into tables in a relational database management system, with
limited semantics and simple integer values for representing
categories such as race and gender. For example, see the
extensive collection of synthetic ecosystems available at [
In this work, we make available full-population synthetic
ecosystem data as Resource Description Framework (RDF) [
In previous work, we created the Ontology of Medically
Related Social Entities (OMRSE) to handle demographics such
as those represented in Census and electronic health record
(EHR) data [
Given the importance of school and workplace assignment and the data about them in synthetic ecosystems, it was critical to represent additionally the roles of students and employees. Furthermore, it was necessary to capture the relationships of these roles to the organizations that create them and to the individual facilities where they are realized. We also report here on the extent to which pre-existing ontologies fulfilled this need vs. the additional ontology development required.
II. METHODS
We reviewed the files generated by the Research Triangle
Institute’s Synthia synthetic population generator [
We reviewed each of the data fields in the following subset of Synthia files: synth_people.txt, synth_households.txt, schools.txt, workplaces.txt. Through an iterative process, we analyzed and described each data field and determined whether to include the field in this work. The most common reason we excluded a field from the final ontological representation was redundancy. For example, we excluded data fields from synth_households.txt, schools.txt, and workplaces.txt that represented the total number of individuals assigned to a household, school, or workplace since these values could be derived by counting in the underlying data. Other fields were excluded because they were determined to be of lesser immediate importance to epidemiologic simulation, such as the prek, kinder, gr01-12, and ungraded fields in schools.txt, which represent the total number of students in different grade categories in a given school.
We then determined whether each included data field could be accurately modeled using existing ontological classes from OMRSE or other established ontologies, or whether new classes were necessary. We created graphical models of how the data would be structured ontologically as an initial specification for transforming the data into RDF, as well as to identify any new classes that we would need to create. These graphical models depict the individuals, relationships between pairs of individuals, and classes to which individuals belong. These diagrams included specifications for associating people with their workplaces and schools as represented in the dataset.
We then manually created these individuals and
relationships for a single set of individuals in one household,
including their associated school and workplace, in a Web
Ontology Language (OWL) [
In accordance with OBO Foundry best practices, we reused as many classes and object properties from other ontologies as we could to generate the OWL file. After importing existing classes from OBO ontologies, it was still necessary to create new classes to represent several key elements of the Synthetic Population dataset. Specifically, we created new classes in OMRSE to represent employment status, employee roles, workplaces, residences, households, and age measurements.
We developed queries of the RDF datasets to validate our representations as well as to identify population characteristics that are likely to influence disease transmission. If these differences are signficant among regions, they could influence the choice of DTM used to study an infectious disease control strategy. For example, if two regions differ substantially in household and workplace composition, size of school-aged population, etc., an AB-DTM is likely to be the better choice. Furthemore, these queries could also be done as part of a simulation experiment to help explain differing results among geographical regions in incidence rates, peak dates, and choice of infectious disease control strategies output by the simulator.
Because the sizes of households, schools, workplaces, and the amount of overlap among them (e.g., households with an employee in the workplace and student in a school) influence disease transmission and thus potentially DTM results, we developed queries to find (1) the average numbers of individuals per household, workplace, and school; (2) the number and percentage of households with both an employee and a student; and (3) the number and percentage of workplaces with at least one employee who lives with a student. We executed these queries against both the Alachua County and Miami-Dade County datasets to contrast these locations based on characteristics relevant to disease transmission.
We loaded the RDF data into an instance of version 3 of the Stardog triple store from Complexible, Inc. This triple store runs on an Amazon Web Services r4.large instance (2 CPUs and 15.25GB of RAM). Queries were submitted from the Stardog command line on the same server on which the triple store was running. The timings we report here are from the Stardog command line output.
III. RESULTS
To accurately model the U.S. Synthetic Population Database, we created RDF representations of the data fields relating to individual persons, households, housing units, workplaces, and schools. We created graphical models of these representations (Figs. 1-4). Fig. 2 illustrates our representation of humans in a household. Fig. 3 illustrates our representations of workplaces and employment.
Many ontologies classify age as a physical quality, rather
than as a measurement of some temporal interval with respect to
the time the measurement was made. The Ontology for
Biomedical Investigations (OBI) [
Fig 1. Key for Graphical Models.
In analyzing Synthia data fields, we found that Synthia conflates households and housing units, despite being based on U.S. Census data that make the distinction clear. For example, Synthia assigns to households both the physical properties of a housing unit, such as latitude and longitude, as well as properties about the household as a social unit, such as total household income, race and age of the head of the household, and household size. Our approach distinguishes household from housing unit and asserts that housing units are individuated by their residence functions and that a household realizes the housing unit’s residence function by living there. In OMRSE, we define a household as a human or collection of humans that occupies a housing unit by storing their possessions there and habitually sleeping there thereby participating in the realization of its residence Fig 2. Graphical Model of RDF Specification of Household.
Fig 3. Graphical Model of RDF Specification of a Person’s Relation to a Workplace. Fig 4. Graphical Model of RDF Specification of Age. function, and add the following description logic equivalence statement: household =def ('Homo sapiens' or 'collection of humans') and ('participates in' some (process and (realizes some
'residence function'))) where residence function is defined as a function that inheres in a material entity and is realized by protecting persons and their possessions from weather and by some person or group of persons habitually sleeping in at least one site that is contained by that material entity.
We created a total of 11 new classes in OMRSE to support the representation of synthetic ecosystems. Each class has a textual definition adapted from U.S. Census. One major adaptation of the definitions was to put them in Aristotelian form with the name of the direct superclass as part of the definition. Other adaptations were necessary to eliminate ambiguity and to reuse other defined ontology terms. OMRSE is a publiclyavailable resource at the following permanent URL: http://purl.obolibrary.org/obo/omrse.owl.
The Alachua county dataset comprised ~13M triples, and the Miami-Dade County dataset comprised 133M triples (Table 1). The population totals for both counties are slightly lower than the 2010 Census numbers on which the Synthia datasets were based. The reason is that we did not incorporate group quarters such as nursing homes and military barracks, which is future work.
The execution time for the SPARQL queries ranged from a few milliseconds to 41 seconds. The longest of these was the query that counted all workplaces with at least one employee who lives at home with at least one student.
The housing unit totals for both counties match the 2010 Census numbers. The data show distinct differences, as expected, between Miami-Dade—a large urban county—and Alachua—a small county (in terms of population) where a large university is located. Miami-Dade has a larger household size and school size, a greater percentage of workplaces with at least one employee that lives with at least one school student, and a greater percentage of households with at least one workplace employee and school student. By contrast, Alachua has a higher average workplace size, even when the University of Florida is excluded from consideration. These differences are likely to impact simulator results—Miami-Dade will often have a larger incidence and prevalence of infectious disease that is spread from person to person such as influenza in the absence of control measures. Infectious disease control measures designed to reduce school and workplace transmission—such as school closure, voluntary or imposed absenteeism from work, and vaccination of the school and / or workplace population—are likely to have a greater predicted effectiveness (and thus perhaps actual effectiveness) in Miami-Dade than Alachua.
All materials created for this paper—the graphical models (including additional ones not shown here), the SPARQL queries, and the OWL files with the entire datasets for Alachua and Miami-Dade counties—are freely available under a Creative Commons Attribution (CC BY 4.0) license at: http://tinyurl.com/syneco-queries.
IV. DISCUSSION
We developed a Semantic Web and realism-based representation of the entire populations of two counties in Florida. We built SPARQL queries to assess differences between the two populations that are likely to influence disease transmission, as well as the results of experiments conducted using DTMs. The approach is generic and could be applied to any other synthetic ecosystem data, including for additional geographical regions. The queries are generic and could be applied to any additional county-based datasets (or datasets at other levels of geographical granularity such as Census tract) similarly transformed via our processes and representations.
We have demonstrated the feasibility of using Semantic Web technologies for representing entire populations, and in particular for representing synthetic ecosystems for use in ABDTMs. Additionally, through additions to OMRSE and the creation of RDF synthetic datasets, we have developed some of the resources necessary to transform other U.S. Census data into Semantic Web representations. In so doing, we have made explicit much of the semantics that are implicit in those data and the synthetic ecosystems that are based on them. It is our conjecture for future work that the explicit semantics improve the ease with which synthetic ecosystems can be expanded to incorporate additional biological, social, and abiotic ecosystem elements.
Although we developed this work in the context of
agentbased DTMs, this resource and approach could also be leveraged
for social network analysis due to the graph-based nature of
RDF. For example, one could construct queries for finding hubs
in the network and people or places that a set of people have in
common. Furthermore, DTMs are increasingly taking into
account social networks as part of the synthetic ecosystem itself
(for example, see Frias-Martinez et al. [
Future work includes expanding the specification to include data related to group quarters, which will require additional ontological analysis and ontology development.
ACKNOWLEDGMENTS This work was supported by award UL1TR001427 from the National Center for Advancing Translational Sciences (NCATS) and award U24GM110707 from the National Institute for General Medical Science (NIGMS). The content is solely the responsibility of the authors and does not necessarily represent the official views of NCATS, NIGMS, or the NIH.