Statistical approaches in plant breeding: Maximising the use of the genetic information

Abstract

Two critical stages of a plant improvement programme are (1) the creation of genetic variation and (2) discriminating this variation through progeny evaluation. In each selection stage, a large number of genotypes are evaluated from which a few incrementally superior individuals are either progressed to the next selection stage or released as commercial cultivars. As progeny testing is expensive, and with many genotypes to screen, it is essential to optimise the experimental design and plot technique and statistical analyses of genetic tests. In the early stages of selection, genotypes are often planted in trials in small, partly replicated, single-row plots due to limited sources of planting material and land requirements for field testing due to the large number of entries to test. Such trials are then subject to variation arising from spatial variability and interplot competition, which reduces the accuracy and confidence of identifying elite genotypes. Spatial variability and interplot competition may seriously affect the estimates of genetic merit and, hence, reduce genetic progress unless they are accounted for in either the trial design or analysis process. Although there are many analysis methods to individually model spatial variability or interplot competition, there are only a few studies that jointly account for both sources of bias. An approach which partitions spatial variability into global trend and extraneous variation and allows for both genotypic and residual level competition is described in this chapter. As the selection cycle progresses, with only the superior individuals advanced to the next stage, genotypes may be tested across a number of trial locations and possibly over several years. Combining data from these many different sources and levels of imbalance to estimate genetic parameters is possible through a linear mixed model that uses residual maximum likelihood to estimate variance components and best linear unbiased prediction to estimate the random effects. There are many different multiplicative methods available to analyse data from multi-environment trials. The advantages and disadvantages of these methods are described in this chapter. One such method, the factor analytic approach, is widely used throughout plant breeding programs as it can partition spatial variability in local and global trends and extraneous variation and account for heterogeneity of residual variance simultaneously. Such methods described in this chapter provide options for tropical crop plant breeders to improve and optimise the design and analysis of genetic experiments.