Project
EF45-2
Effective Stochastic Simulation of Adaptive AB Models
Project Heads
Max von Kleist
Project Members
Nils Gubela
Project Duration
01.05.2023 – 31.12.2025
Located at
FU
Description
Agent-based models (ABMs) are increasingly used to study complex phenomena. We are interested in a class of ABMs that constitute stochastic (Markovian) contact-, as well as reaction dynamics. Using this setup, a state of the ABM can be represented by a temporal graph where the contact dynamics change the set of edges and the reaction dynamics alter the vertices.
While ABMs may differ in their setup, they typically involve adaptive processes. In epidemiological applications, adaptivity may entail measures of self-isolation (quarantine) upon infection, altering the dynamics of edges in temporal graph, whereas in biochemical systems, molecular complex formation alters diffusion and attraction of neighboring molecules. Thus, adaptivity creates feedback, which can lead to emergent complex behavior. Analytical solutions can only be obtained in very few special cases and thus numerical studies based on stochastic simulations are often indispensable to explore the dynamical behavior of these ABMs.
However, when simulating these ABMs, the main interest is in determining the evolution of the reaction dynamics. I.e. the main interest is in the state of the agents. Exact numerical methods suffer from the immense computational overhead of simulating the contact dynamics of agents in order to accurately capture the relevant reaction dynamics. Inexact stochastic simulation based on time discretization, and (a-)synchronous parallel updating of edges and vertices are most commonly implemented in epidemiological simulation software, yielding unreliable results.
We recently developed a rejection-based approach for simulating effective reaction dynamics on aforementioned ABMs, called SSATAN-X [1]. The central idea is to bulk update the contact dynamics between two epidemic reactions.
In this project, we aim to further advance SSATAN-X simulation method for adaptive ABMs. The goals of the project are threefold: error-analysis of the bulk-updating scheme, increasing computational efficacy and application to different (adaptive) network models.
High acceptance sampling of infection events
The epidemic ABM comprises two distinct processes: network processes and epidemic processes. While the update of the network process contributes significantly to the computational overhead, it offers minimal insight into the underlying epidemic events. To address this limitation, we propose an acceptance-rejection based algorithm that updates only the relevant network components at any given time point [2]. Our analysis demonstrates that this method is exact and, when compared to SSA, is multiple orders of magnitude faster, enabling computations that were previously infeasible. With this new model, we are able to analyze how changes in behavior shape the epidemic curve.
Application: Quantifying the impact of transient behavior changes on the 2022 mpox outbreak in Berlin
Mpox denotes a viral zoonosis caused by the Orthopoxvirus monkeypox (MPXV), which is endemic in West and Central Africa. In spring 2022, notable outbreaks of MPXV clade IIb were recorded in several high-income countries, predominantly affecting men who have sex with men (MSM). At the peak of the outbreak, over 200 new mpox cases per week were reported in Berlin, which constitutes one of the largest MSM population in Europe. Within the same year, the outbreak significantly declined, and it is unclear which factors contributed to this rapid decrease. To investigate the concomitant effects of sexual contact networks, transient contact reductions and the effect of infection- vs. vaccine-derived immunity on the 2022 mpox outbreak, we calibrated an agent-based model with epidemic, vaccination, contact- and behavioral data [3]. Our results indicate that vaccination has a marginal effect on the epidemic decline. Rather, a combination of infection-induced immunity of high-contact individuals, as well as transient behavior changes reduce the number of susceptible individuals below the epidemic threshold. However, the 2022 mpox vaccination campaign, together with infection-derived immunity may contribute to herd-immunity in the Berlin MSM population against ongoing clade I mpox outbreaks. Demographic changes and immune waning may deteriorate this herd immunity over time. These findings highlight that, in addition to vaccination, timely and clear communication of transmission routes may trigger spontaneous protective behavior within key populations; underscoring the importance of targeted sexual health education as a core component of outbreak response.
Project Webpages
Selected Publications
- Malysheva N, Wang J, von Kleist M. S̲tochastic S̲imulation A̲lgorithm For Effective Spreading Dynamics On T̲ime-Evolving A̲daptive N̲etworX̲ (SSATAN-X) Math. Model. Nat. Phenom. 17 (2022)
- Gubela, N., & von Kleist, M.Efficient and accurate simulation of infectious diseases on adaptive networks. PLOS Complex Systems 2(6) (2025)
- Gubela, N., Kim, H. Y., Lunchenkov, N., Stern, D., Michel, J., Nitsche, A., … & von Kleist, M. Behavior change and infection induced immunity led to the decline of the 2022 Mpox outbreak in Berlin. Communications Medicine (2026)
Selected Pictures
Please insert any kind of pictures (photos, diagramms, simulations, graphics) related to the project in the above right field (Image with Text), by choosing the green plus image on top of the text editor. (You will be directed to the media library where you can add new files.)
(We need pictures for a lot of purposes in different contexts, like posters, scientific reports, flyers, website,…
Please upload pictures that might be just nice to look at, illustrate, explain or summarize your work.)
As Title in the above form please add a copyright.
And please give a short description of the picture and the context in the above textbox.
Don’t forget to press the “Save changes” button at the bottom of the box.
If you want to add more pictures, please use the “clone”-button at the right top of the above grey box.