The Rule-based Epidemic Modelling (RBEM) Project was funded by the UK Medical Research Council project (grant X/011658/1) to make rule-based modelling methodology more accessible to the infectious disease modelling community. A rule-based approach is in contrast to writing differential equations, and there can be human as well as technical advantages to this. Our pre-print (draft) paper illustrates how the approach can be applied to classical problems, in addition to our previous work (cited below) on novel problems during pandemic pressures.
In our team with Sándor Bartha, Akindele A. Onifade and William Waites, we modelled malaria as found in sub-Saharan Africa in areas of high prevalance of sickle-cell trait (HbAS). HbAS carriers typically do not develop malaria symptoms when tested but remain infectious to others. We considered the research question:
If testing for malaria is done according to observing symptoms, does the infectious resevoir of asymmptomatic HbAS carriers introduce a bias sufficient to affect the course of a malaria epidemic?
We were able to investigate this more readily with our rule-based techniques than with traditional approaches, concluding that observational testing bias of the HbAS resevoir can render testing ineffective for suppressing an epidemic, which we contrasted with the effect of randomised testing.
Overview
A significant part of the advantages to rule-based modelling relates to the natural history of a disease.
The natural history of a disease is the story of its progress in individual humans or animals. No matter how the disease starts, it has a process and an endpoint - the patient may be in many states including remission, recovered, chronicity or death. The natural history arises from a combination of the empathy of a physician with the observational skill of a scientist, and has been a foundation of medical care for thousands of years.
An epidemiological model takes another step, where the natural history can be calculated on computers so we can draw conclusions about what happens across entire populations as we manage a disease.
The RBEM project addresses a difficulty in epidemiological models where the computer language expressing the disease equations can look very different from the natural history it is modelling. This in turn makes it difficult to explain what the model is doing, often not immediately obvious even to experts which natural history a model is encoding. This also makes it difficult for people elsewhere to reproduce results. These factors become a problem in a emergency.
In this project our goal is to:
Provide a good language for writing down the natural history of a disease as a story, describing and explaining how a person moves from one state to another, and what comes next. This language needs to be easy to read so humans can understand and improve it, but also suitable for computers to perform calculations on it. The language needs to be easy to set up and run so that it can become part of the daily discourse in investigating disease.
We hope this may be a useful additional tool to the many which already exist.
Origins of rule-based epidemic modelling
The first application of rule-based modelling to epidemics was under the extreme challenges of COVID-19. As the Royal Society’s 2020 initiative Rapid Assistance in Modelling the Pandemic made clear, existing techniques were often insufficient. We agreed, and contributed a paper to the Society’s 2021 review in which we said:
Real-time modelling and the vast amount of models developed in the last 24 months have highlighted gaps in the existing technical frameworks that ought to be addressed in preparedness for possible future pandemics.
This was confirmed in 2025, when the UK COVID-19 enquiry published its first two reports containing about 250 references to mathematical modelling. Epidemiologists providing advice in times of crisis need the tools to demonstrate their reasoning in a reproducible manner, easily explainable to their scientific peers.
All modelling is communication, and we believe our technique has communication advantages for readers in terms of accessibility and explainability, and potentially also in ease of use for those creating models.
As we said in our 2021 paper in the Journal of Theoretical Biology:
Rule-based models allow to combine transparent modelling approach with scalability and compositionality and therefore can facilitate the study of aspects of infectious disease propagation in a richer context than would otherwise be feasible.
Having validated this approach with many examples in the paper, we now seek to make it more accessible. Here we bring together existing techniques and tools and demonstrate practically how to:
- easily get up and running to produce epidemiological results
- have a reproducible environment, which is a precondition for reproducible results
- computational techniques can be overlaid on the basic rule-based system
- integrate into existing disease modelling workflows
- which additional work is needed to make the rule-based modelling systems better