Abstract
As larger genomic data sets become available for wild study populations, the need for flexible and efficient methods to estimate and predict quantitative genetic parameters, such as the adaptive potential and measures for genetic change, increases. Even though animal and plant breeders, as well as the field of human genomics, have produced a wealth of methods, wild study systems often face challenges due to larger effective population sizes, environmental heterogeneity and higher spatio-temporal variation. Existing approaches either rely on two-step procedures, where residuals from a pre-fitted model are used as the response in a second analysis, or can become computationally inefficient as model complexity and data size increase. We therefore adapt methods from animal breeding to account for the complexity typically present in wild animal populations. The core idea is to approximate breeding values as a linear combination of principal components (PCs), where the PC effects are shrunk with Bayesian ridge regression. The result is a computationally efficient and scalable approach, denoted Bayesian principal component ridge regression (BPCRR). A case-study for a Norwegian house sparrow meta-population, as well as simulations, illustrate that the method efficiently estimates the additive genetic variance and accurately predicts breeding values. In order to assess whether BPCRR predicts informative breeding values, we also apply BPCRR to track micro-evolutionary change across time and space in the house sparrow system. To make the method accessible, we provide coded examples and data.