📐 Mathematical Modeling
Agent-based models, differential equations, and phylodynamics to understand and predict infectious disease dynamics — from biomolecular scales to population-level transmission.
Computational Epidemiology
Our modeling program develops mathematical and computational frameworks to simulate, understand, and predict the transmission of infectious diseases. We integrate individual-based models, compartmental ODE/PDE systems, phylodynamics, and deep learning to address pressing public health questions in Burundi and beyond.
Modeling Approaches
A spectrum of methods from mechanistic ODE models to AI-driven forecasting
Compartmental Models (ODE/PDE)
SIR, SEIR, and extended compartmental frameworks using ordinary and partial differential equations to capture disease dynamics at population level. Fitted to surveillance data for parameter estimation and scenario projections.
Agent-Based Models (ABM)
Individual-level stochastic simulations (SimpactCyan, custom frameworks) that capture heterogeneous behaviours, sexual networks, and within-host dynamics. Calibrated to multi-data sources including surveys, phylogenies, and genomic data.
Phylodynamic Models
Integrating phylogenetic trees with epidemiological models to reconstruct transmission networks, infer age-mixing patterns, and estimate population-level parameters from genetic data.
Climate-Driven & Spatial Models
Incorporating temperature, rainfall, NDVI, and topographic data into transmission models. Generates risk maps using the basic reproduction number R₀ as a spatial quantitative measure under climate change scenarios.
Deep Learning & Forecasting
Neural network models (LSTM, Transformer-based architectures) trained on epidemiological time series for malaria prediction. AI-powered surveillance network for Mpox detection and response (AI4Mpox, 2025).
Multiscale & Within-Host Models
Bridging biomolecular dynamics (parasite–vector interactions) with population-level transmission via multiscale ODE and ABM frameworks. Current focus: Iso Lomso fellowship on malaria parasite biomolecular dynamics.
Bayesian & Statistical Inference
Semi-parametric models with error-prone covariates, Lie group methods for ODE solving, yield curve parametric models, and Bayesian calibration of individual-based models to observed data.
Multiscale Modeling Framework
Linking biomolecular to population levels for a complete picture of disease dynamics
Molecular Scale
Parasite & pathogen biomolecular dynamics
Within-Host
Host immune response & vector interactions
Individual
Agent-based behaviour & contact networks
Population
Transmission dynamics & R₀ across space
Modeling Research Projects
From published results to ongoing investigations
Malaria trend and control interventions in Burundi (2000–2021)
Review of interventions over two decades. Found more than a twofold increase in malaria cases since 2000, reaching 843,000 per million inhabitants in 2019 despite significant scale-up of health facilities and testing. Highlighted gaps in RDT accuracy and high asymptomatic proportions.
Climate change effects on malaria transmission dynamics in Burundi
Climate-driven SEIR model incorporating temperature and rainfall. Produced risk maps using R₀ under future climate scenarios. Key finding: southwestern regions (Rumonge, Bururi) and western provinces (Bubanza, Bujumbura Rural) are projected to become the highest-risk areas, reversing current patterns.
Predicting malaria dynamics using deep learning models
Neural network forecasting models applied to malaria time series data in Burundi. Demonstrated superior predictive accuracy over classical statistical methods for outbreak anticipation and resource planning.
HIV transmission network determinants via calibrated agent-based models
Combined sexual behavioural survey data, phylodynamics, and SimpactCyan ABM into a unified framework for HIV prevention research. Inferred age-mixing patterns and quantified uncertainty in transmission network reconstruction from phylogenetic trees.
Ebola transmission dynamics: will future outbreaks become cyclic?
Mathematical model exploring cyclicity of Ebola outbreaks using DRC outbreak data. Investigates whether climatic, ecological, and behavioral drivers could sustain endemic cycles rather than episodic outbreaks.
Multiscale modeling of malaria parasite biomolecular dynamics
Current STIAS fellowship project. Developing agent-based and multiscale ODE frameworks to understand biomolecular interactions between malaria parasites and mosquito vectors, linking within-host dynamics to population-level transmission.
AI4Mpox: AI-powered surveillance and response network
Funded by a 50,000 CAD grant. Developing an AI-powered surveillance system integrating epidemiological, genomic, and climate data for early detection and response to Mpox outbreaks in Central and East Africa.
Diseases We Model
Infectious diseases at the core of our mathematical modeling program
Malaria
Climate-driven, spatial, deep learning & multiscale
HIV / AIDS
Agent-based, phylodynamic & network models
Tuberculosis
ODE compartmental & Lie group methods
Ebola
Cyclicity & outbreak trajectory modeling
COVID-19
Phylogenetic & epidemiological dynamics
Zoonoses
Spillover risk mapping & outbreak modeling
Mpox
AI surveillance & genomic-linked models
NCDs
Burden projection & risk factor modeling
Tools & Software
Open-source and custom platforms driving our computational work
SimpactCyan
Open-source ABM for HIV with R & Python interfaces (co-developed)
R / Python
Statistical modelling, bioinformatics pipelines, and data analysis
QGIS
Spatial epidemiology, risk mapping, and geographic visualisation
BEAST / IQ-TREE
Phylogenetic inference and molecular clock modeling
TensorFlow / Keras
Deep learning for malaria forecasting and AI surveillance
MATLAB / Julia
Numerical ODE/PDE solvers for compartmental model fitting