What is TITAN?

TITAN (Treatment of Infectious Transmissions through Agent-based Network) is an agent-based simulation model used to explore contact transmission in complex social networks. Starting with the initializing agent population, TITAN iterates over a series of stochastic interactions where agents can engage in risk behavior with one another, transmit infections through various medium, and enter and exit the care continuum. The purpose of TITAN is to evaluate the impact of treatment models on incidence and prevalence rates of the targeted disease(s) through the use of data fitting simulated trajectories and rich statistics of primary/sub-population attributable proportions.

TITAN is a has a sophisticated social network structure which allows for various forms of care cascade, interventions and barriers, and dynamic agent parameters such as:

  • Stocahstic enrollment/discontinuation of treatment as prevention strategies such as ART (antiretroviral therapy), PrEP (Pre-exposure prophilaxis), and VCT (voluntary testing)
  • High risk group behavior such as sex workers, injection drug use, or post-incarceration behavior
  • Variable transmission probability based on mode of transmission (anal, vaginal, needle, etc.) and viral load (acute, AIDs, virally supressed)
  • K-nearest neighbor partnering algorithm finds partners based of fitness rather than randomness
  • Heterogeneous agent populations including dynamics such as: racial disparity, sexual orientations (MSM, WSW, bisexual), sexual activity and promiscuity (main/casual/one-ofs), and assortive age mixing.
  • Disruptive barriers in form of incarceration or migration

Agent populations are defined as graphs (nodes connected by edges). Nodes in the graph are used to represent the attributes (or collection of attributes) of an agent (person), and edges define the type of relationship between agents. In practice, a graph represents a social network of connected people through various relationship types, and provides the medium for which agents can interact.

TITAN is developed using the Python 2.7 (https://www.python.org/download/releases/2.7/) programming language. TITAN also implements standard numberical libraries Numpy (http://www.numpy.org/) and SciPY (https://www.scipy.org/), as well as a set of Python v2 bindings for The Qt Company’s Qt application framework PYQT (https://wiki.python.org/moin/PyQt) for graphical user interfacing.

To promote efficiency and minimal computational effort, TITAN uses the python library NetworkX (https://networkx.github.io/) to create network graphs and perform network analysis/vizualization such as connectivity measures or Fruchterman-Reingold force-directed diagrams.

TITAN’s Developement website: http://titan-simulation.org/

Version History

v1.0 February 2017: Initial Release

Agent-based Modelling

A sub-class of complex systems methods, agent-based models (ABMs) are individual-based models and are increasingly commonplace in studies investigating the social and structural determinants of population health. In brief, an ABM consists of agents (i.e., ‘’nodes’‘) who are connected to each other by links (i.e., ’‘edges’‘), through which information, behaviors, or some other social process can be transmitted. Every agent’s internal state (e.g., HIV infection status and disease stage) are updated at each discrete time step based on pre-programmed rules and interactions with other agents.8 In this model, links represent sexual and injection-related risk behavior between two agents through which contact infectious transmissions can occur. In this model, links represent sexual and injection-related risk behavior between two agents through which contact transmission can occur.

The driving forces of the model itself is a connected tree of components which have orders of influence on transmission events, as seen in the figure below The risk behav- iors, such as the number of interactions an agent engages in with its partners, are the main source of transmission events of the modeled infection. In this model, agent de- mographics (dfined by the input parameters of the sim- ulation) dominate the risk behavior of a particular agent, and also determines the care continuum of the agent in its environment. The agents’ risk to transmit is damped by external preventative measures that are engaged in each component of the model. Testing of the agents can lead to a direct reduction in risk behavior (an agent aware of their infection status may engage in less risk acts or partners), and also cascades downwards into direct pre- ventative measures such as condoms or safe needle ex- changes (for PWID). Additionally, agents may enroll in transmission preventing programs or treatments, classified as the care continuum, which also directly inuences the agents risk behavior and transmissibility. These components work in concert to provide the overall dynamic of the simulation, and we continue to employ complexities that interact with these components in either preventa- tive or disruptive impacts.


Who is TITAN for?

TITAN is primarily intended for academic research of …