Estimation of QTLs From Multi-Environment Trials.

Multi-environment trials (METs) are often used in plant breeding programs to assess the responses of a set of genotypes, and their dependence on the environment. METs are best analyzed by fitting linear mixed models using the REML (i.e. Residual, or Restricted, Maximum Likelihood) algorithm. This not only provides an efficient way of estimating the fixed and random effects, it also allows models to be fitted to describe the variance-covariance structure of the random effects. In METs this is particularly useful in the description of the genotype by environment interaction (GxE). Mixed models also provide a suitable statistical framework for the mapping of quantitative trait loci (QTLs), including extensions to QTL by environment interaction (QTLxE).

Until recently the use of these methods has been restricted to those with substantial statistical expertise. However, the authors have recently collaborated on the development of a menu-based system, in the latest release of GenStat for Windows, that makes these methods as accessible as the familiar, ordinary analysis of variance. This talk will explain the underlying philosophy of the system, and show how it provides the analysis in several simple steps. First of all, the trials from the different environments are analyzed individually to produce a table of genotype-by-environment means. In the second step, these are used to model the GxE, and identify the best model for the phenotypic data. This phenotypic model can then be used in QTL mapping, using simple and composite interval mapping, to establish a multi-QTL model to estimate QTL effects. The talk will show how data can be loaded and results saved in all the usual formats, and explain the variety of tables and graphs that can be produced to illustrate the results.