odin dust monty

ODM is an open-source modelling stack for infectious disease research. odin specifies models in a clear declarative language, dust runs them as fast stochastic simulations, and monty fits them to data through Bayesian inference — one open workflow from idea to evidence.

📦 Packages and workflow

odin
Model specification
Declarative modelling language compiling to C++.
|
dust
Simulation
Scalable stochastic simulation engine.
|
monty
Inference
Monte Carlo inference for simulation-based models.

Together, these components form a modular modelling–inference pipeline used across UK infectious disease research.

Composable

Use components independently or combine into a full modelling pipeline.

Efficient

Compiled models and scalable simulation enable fast, large-scale workflows.

Robust

Modern inference supports principled uncertainty quantification.

🌱 Our vision for ODM

What we're trying to do

Make the work of building, running and fitting epidemiological models transparent, reproducible, and within reach — for research groups, outbreak responders, and students.

How we want to get there

  • As a community. Widening the circle of contributors, institutions and users who shape the stack.
  • Distributed stewardship. Maintainership shared across institutions and career stages, with a public roadmap.
  • Accessible. Lower barriers to contribute and to use, with documentation, contributor guides and transparent governance.

Why it matters

Robust, accessible tools — ready for use now, and for the next emergency. Transparent and well-maintained, sustained by a community that knows how they work and trusted by the researchers and responders who depend on them.

⚕️ Example: public health workflow

Specify model
Define a transmission model (e.g. malaria or respiratory infection), including compartments, parameters, and interventions.
|
Simulate scenarios
Run fast stochastic simulations across settings or intervention strategies to generate incidence and burden projections.
|
Fit to data
Estimate parameters and uncertainty by fitting the model to surveillance or trial data using Monte Carlo methods.
Example compartmental model diagram
Example simulated scenario output
Example model fit to data

This supports decision-focused outputs such as projected impact of interventions, uncertainty bounds, and comparison of policy options.

Figures from Imai et al. (2023).

🔬 Use in research and policy

The stack is used across academic research, public health analysis, and teaching. Examples include:

  • Infectious disease modelling studies published in leading journals
  • Analyses supporting the UK COVID-19 response and government decision-making
  • Collaborations with public health partners, including the UK Health Security Agency
  • Teaching at BSc and MSc level and international short courses in infectious disease modelling

The tools are used across multiple UK institutions and internationally, with active community contributions through open development on GitHub.

📄 Related publications

📚 Resources

Documentation and tutorials

For additional information, examples, and tutorials, please see:

Team and institutional context