Inferring the effects of mutations on SARS-CoV-2 transmission

Much work has been devoted to inferring the fitness effects of mutations in evolving populations using models from population genetics. In parallel, many different models of disease spread have been developed in the field of epidemiology. But comparatively little work has been devoted to models of disease spread that are conducive to inferring the effects of mutations on disease transmission. Here we develop a model for disease spread in a localized population that accounts for both the possibility of a long tailed distribution for the number of people infected by a single individual, and that pathogen genetic variation may affect the probability of transmission, i.e., transmission fitness. Using this model, we develop a method of inferring the fitness for different mutations from time series data. We verify the model against simulations of disease spread that include super spreaders, as well as against standard susceptible-infected-recovered (SIR) models, and show that in both cases fitness is accurately recovered. Applied to SARS-CoV-2 data, we find a few groups of linked mutations that appear to strongly affect the viral transmission rate.