Can we estimate genetic variance and breeding values for infectious disease susceptibility?

Specifically, they investigated how the precision of these estimates is affected by factors such as the amount of genetic variation in susceptibility, population structure, available data, and model formulation. The researchers also compared the performance of the GLMM with that of a conventional linear mixed model (LMM).

Genetic selection of animals for increased resistance to infectious diseases has been studied for several decades and could complement, or in some cases provide an alternative to, other disease-control measures. A common approach in animal breeding is to classify each animal according to whether it is infected or not (for example, healthy = 0 and infected = 1) and then use statistical models to estimate how much of the variation in infection status is due to genetic differences between animals.

Two commonly used approaches are linear models and threshold models. Whereas linear models treat infection status as if it were a continuous trait, threshold models assume that an animal has an underlying normally distributed liability (that is, a continuously varying level of susceptibility to infection). This liability is assumed to follow a normal distribution, with infection occurring when an animal’s liability crosses a certain threshold.

However, neither approach explicitly considers how infections spread through a group of animals. For example, an animal may remain uninfected not because it is genetically resistant, but simply because it had little contact with infected herd mates. Conversely, a genetically resistant animal may still become infected if it is repeatedly exposed to a high level of infection. By ignoring these differences in exposure and the transmission of infection between animals, conventional models can make it difficult to interpret estimated genetic parameters and breeding values in epidemiological terms. In particular, these estimates do not necessarily provide a reliable prediction of how much infection will actually decrease following genetic selection.

GLMMs based on epidemiological theory have been developed to estimate how infections are transmitted between animals. These models can also be used to estimate genetic parameters and breeding values for susceptibility to infection. “Previous studies suggest that GLMMs can estimate breeding values for susceptibility reasonably accurately, even when only a limited amount of data are available,” says Piter Bijma, one of the authors of the study. “An additional advantage is that GLMMs are already implemented in standard animal-breeding software and generally require relatively little computational time, making them potentially suitable for practical use. However, despite these advantages, the ability of GLMMs to accurately estimate genetic parameters for susceptibility had not yet been systematically validated.”

The researchers analysed longitudinal data from simulated epidemics using a GLMM implemented in ASREML software. They examined the effects of both the genetic variance in susceptibility and the transmissibility of the infection, expressed as the basic reproduction number, on the precision of genetic parameter estimates and the accuracy of EBVs. In addition, they assessed important data-related factors, including how frequently animals were observed and the size of the contact groups in which transmission occurred. Together, these analyses show how different types of disease data and experimental designs influence the reliability of genetic variance estimates and breeding values for susceptibility to infection.

The results of the study show that a GLMM applied to the repeated observation of whether individual animals are infected can provide accurate and unbiased estimates of genetic variance, as well as reliable breeding values for susceptibility to an infectious disease. Among the data requirements the researchers examined, the time between consecutive observations of infection status had the greatest effect on the accuracy of the estimates, whereas the size of the contact group had only a limited effect. The appropriate observation interval depends on how quickly animals become infected and recover from infection. Overall, the GLMM appears to be a promising and practical method for estimating genetic parameters and breeding values for susceptibility, particularly when frequent longitudinal records of individual infection status are available.

Read the full study here: Estimation of Genetic Variance and Breeding Values for Infectious Disease Susceptibility From Simulated Longitudinal Data Using Generalized Linear Mixed Models Based on Transmission Dynamics