Lineage frequency fluctuations reveal the spatial dynamics of disease transmission in SARS-CoV-2

Human mobility is a key factor in the spatial transmission of infectious diseases. Although human mobility can be quantified from survey data or cell phone data, it remains difficult to generate reliable models of the spatial-transmission dynamics from mobility data due to the complexity of the underlying epidemiological processes. For SARS-COV-2, a vast amount of lineage data is available with unprecedented spatio-temporal resolution. Here, we develop a data-driven method that uses time series neutral lineage frequencies data to infer an infectivity matrix, which describes the transmission rates between individuals in different locations. From the inferred infectivity matrix, we can determine epidemiologically relevant quantities, such as the relaxation timescale of the dynamics, the dispersal kernel associated with the disease transmission, and the spatial-spreading pattern of a new variant. Our method is also applicable to infer the couplings between non-spatial groups, such as age groups, where mobility proxies are unavailable.