DERP uses recursive neural networks to solve inference problems in phylodynamics. In particular, it provides a way to estimate key aspects of an outbreak (e.g. the reproduction number and prevalence of infection) using a phylogeny reconstructed from genomic data.

Overview of DERP
  • Code to simulate data phylogenies from epidemics.
  • Code to train a neural network to predict epidemic properties from a phylogeny.
  • (Coming soon) Code to assist with applying the trained neural network.

Bibliography

  • Zarebski AE, Williams T, and du Plessis L (2026) Amortized Phylodynamic Inference with Neural Bayes Estimators and Recursive Neural Networks. arXiv. (Link)

If you are interested in deep learning with PyTorch, you might also like my blog.

Project maintained by Alexander E. Zarebski