Estimating epidemic dynamics with genomic and time series data

Abstract

AbstractAccurately estimating the prevalence and transmissibility of an infectious disease is a critical part of genetic infectious disease epidemiology. However, generating accurate estimates of these quantities, informed by both time series and sequencing data, is challenging. Birth-death processes and coalescent-based models are popular methods for modelling the transmission of infectious diseases, but they struggle with estimating the prevalence of infection.We extended our approximation of the likelihood for a point process of viral genomes and time series of case counts so it can estimate historical prevalence, and we implemented this in a BEAST2 package called Timtam. In a simulation study the approximation recovered the parameters from simulated data, even when we aggregated the point process data into a time series of daily case counts.To demonstrate how Timtam can be applied to real datasets, we estimated the reproduction number and the prevalence of infection through time during the SARS-CoV-2 outbreak onboard the Diamond Princess cruise ship using a time series of confirmed cases and sequence data. We found a greater prevalence than previously estimated and comment on how differences in the algorithms used could explain this.