Inferring pandemic growth rates from sequence data

de Silva, Eric; Ferguson, Neil M.; Fraser, Christophe

Bayesian evolutionary analysis by sampling trees (BEAST) is a popular software package that integrates many phylogenetic and coalescent-based tools [7] . It can be used to estimate phylogenetic trees and apply the coalescent model to infer demographic changes from sequence data. A Bayesian approach is used: given the data (aligned sequences), a specified model structure (coalescent, substitution model, rate heterogeneity model) and specified prior distributions of parameters, the posterior distribution of trees and parameters is estimated. BEAST uses an efficient Markov chain Monte Carlo (MCMC) algorithm to sample the posterior distribution of phylogenetic trees and parameters. As a result of this approach, the time-varying effective population size estimated by the coalescent is averaged over many phylogenetic trees. If a non-parametric estimate based on the generalized skyline plot is implemented, then the resulting estimate is called the Bayesian skyline plot (BSP). Both piecewise constant and piecewise linear skyline models can be implemented, the latter appropriate if one knows a priori that the inferred population is growing. It is also possible to specify specific parametric models for the effective population size, such as exponential growth, which may be appropriate when analysing an emerging epidemic. We will explore both approaches here.