Fundamental identifiability limits in molecular epidemiology

摘要: Viral phylogenies provide crucial information on the spread of infectious diseases, and many studies fit mathematical models to phylogenetic data estimate epidemiological parameters such as effective reproduction ratio (Re) over time. Such phylodynamic inferences often complement or even substitute for conventional surveillance data, particularly when sampling is poor delayed. It remains generally unknown, however, how robust are, especially there uncertainty regarding pathogen prevalence intensity. Here we use recently developed techniques fully characterize that can possibly be extracted from serially collected viral in context commonly used birth-death-sampling model. We show any candidate scenario, exists a myriad alternative, markedly different yet plausible "congruent" scenarios cannot distinguished using alone, no matter large dataset. In absence strong constraints rate priors across entire study period, neither maximum-likelihood fitting nor Bayesian inference reliably reconstruct true dynamics alone; rather, estimators only converge "congruence class" dynamics. propose concrete feasible strategies making more data.