Mixed modelling for QTL × environment interaction analysis

Abstract

Phenotypic data for quantitative trait loci (QTL) mapping studies are typically generated at multiple environments in order to broaden the inference space. Many aspects of the usually complex design call for a mixed modelling approach taking into account various sources of variation, e.g., incomplete blocks, a spatial error structure, genetic correlations due to the pedigree, and random environmental effects, including QTL × E interaction. Perhaps the most important source of random variation is the genetic correlation across environment, which arises when the same set of lines is tested in each environment. This correlation is likely to be positive, and ignoring it will lead to an increased rate of false positives. In this paper, we present a mixed modelling framework for QTL mapping based on complex data from multiple environments. Our main focus is on an appropriate modelling for the non-QTL part. The methodology will be illustrated using a barley data set from a BC₂F_2:5 advanced backcross trial.