Inferring pandemic growth rates from sequence data

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

Strictly, the ordinate axis of BSPs produced by BEAST is a compound value of N e . t, where t is the average generation time. In reality, the generation time varies from one infection to the next and so it is not straightforward to precisely evaluate N e at a given time (some refer to this compound value as the genetic diversity); this is not the case for the simulations in this study where the generation time was fixed to t ¼ 2.6 days (and we use this value to show N e in all plots). The interpretation of the absolute value of the estimated effective population size, N e , and what relationship this quantity has with the actual number of individuals infected is not obvious. While the effective population size is significantly less than the number of individuals infected towards the end of the simulation (this is not the case early on), this number increases with sampling density. In most studies, N e is assumed to be the ideal (Wright -Fisher) population size that gives the same coalescent rate seen in the imputed phylogeny. That N e is invariably less than the true population size is often attributed to population structure or other heterogeneity, such as in infectiousness. In addition, in this study, we quote the median posterior N e estimates, which tend to be significantly smaller than the mean. Recent studies [34] conclude that the coalescence rate is a function of the prevalence, through sampling effects, as well as the incidence, although for a short generation time disease such as influenza, the difference between these is limited (effectively just a factor of the mean generation time). This study has focused more on the relative rate of change of N e estimates than the meaning of the absolute values of such estimates, however.