Estimation of Genetic Variance and Breeding Values for Infectious Disease Susceptibility From Simulated Longitudinal...

Recent theoretical work shows that the potential of genetic selection to reduce the prevalence of infectious diseases is much larger than expected from classical quantitative genetic theory, due to indirect genetic effects that arise in the transmission process. However, to fully benefit from these indirect effects, we need to estimate genetic parameters and breeding values, which requires statistical methods tailored to the transmission process. Here, we evaluate Generalized-Linear-Mixed Models (GLMMs) implemented using software commonly used in animal breeding to estimate genetic parameters and breeding values for susceptibility of hosts to infection, using simulated data of epidemics. Longitudinal records of individuals' infection state provide information on the order of infection, as well as on the exposure dose of non-infected animals. Such information can be harnessed to estimate genetic parameters for susceptibility, and can be included in a GLMM as a so-called offset. Therefore, we used longitudinal records of individual infection state to assess the impact of sampling interval, population structure, infection characteristics, and model formulation on the estimated genetic variance and breeding values for susceptibility. The results show that a GLMM fitted to longitudinal records of individual binary infection state can produce accurate and unbiased estimates of genetic variance, as well as good prediction accuracies of breeding values for susceptibility to an infectious disease. Of the data requirements, the time interval between consecutive observations on individual infection state was the main factor affecting estimation, while group size had a limited effect. The required observation interval depends on the infection and recovery rates of individuals. The GLMM thus seems an accurate and easily implementable model to estimate genetic parameters and breeding values for susceptibility when dense longitudinal records on individual infection status are available.