Developing predictive models of influenza evolution

Understanding the predictability of evolution is a classic problem in biology, with particular relevance for rapidly evolving pathogens like influenza. Influenza vaccines must be designed to hit a "moving target," as the virus constantly evolves to escape human immune responses. We developed a novel approach to this problem by combining mathematical methods from statistical physics with epidemiological modeling. In this talk, I'll describe our approach to estimating how different mutations affect the transmissibility of the virus, and how we can use these estimates to forecast future evolution. I'll also discuss how we've applied this approach to model influenza evolution in real data. Despite a shifting immune environment, we can generate successful predictions for both the composition of the viral population in the near future and the success or loss of individual mutations.