Optimal weighting of information in marker-assisted selection

In crosses between inbred lines linear regression can be used to estimate marker effects; these marker effects then allow marker-assisted selection (MAS) for quantitative traits. Weighting of marker and phenotypic information in MAS requires estimation of genetic variance associated with the markers: the usual estimators are biased, resulting in too much weight being placed on marker information relative to phenotypic information. In this paper we develop a cross-validation method to remove this bias, and show by simulation that response to selection using this method is almost as high as that achieved using optimal weighting of marker and phenotypic information.