For best experience please turn on javascript and use a modern browser!

Luc Coffeng (Oct 2018 – Oct 2019)

Assistant Professor at the Department of Public Health, Erasmus MC, University Medical Centre Rotterdam

My research

As an MD and epidemiologist, I work on predicting and optimising the impact of public health programmes against neglected tropical infectious diseases, using individual-based models. Most of the diseases I work on are parasitic infections that include life cycles involving multiple host species. I am particularly interested in evolution of drug resistance, the role of mobility of carriers of infection, and diagnostic challenges in evaluating progress towards elimination.

Fellowship at the IAS (October 2018 – October 2019)

Elimination of any infectious disease comes with the challenge of predicting whether or not a pathogen is still circulating in a population. Neglected tropical diseases (NTD) come with an additional challenge of complex pathogen life cycles that can only be adequately captured with complex, individual-based transmission models. Due to the complexity of these transmission models, likelihood-based inference is unfeasible. Approximated Bayesian computation based on sequential Monte Carlo is a relatively new likelihood-free technique for probabilistic inference that can be used to calibrate stochastic transmission models and perform probabilistic uncertainty analyses. However, ABC-SMC efficiency is sensitive to the curse of dimensionality and highly depends on several algorithm tuning parameters, which need to be set manually. Further, compared to more standard Markov chain Monte Carlo techniques, there is a dearth of visual indicators to evaluate ABC-SMC performance.

During my fellowship, I will work on optimising ABC-SMC performance through automated selection of algorithm tuning parameters and will develop visualisation tools for performance assessment. I will test the resulting framework on stochastic transmission models of increasing complexity to assess where and when ABC-SMC can be reliably used.