Models and methods for computer simulations as a resource in plant breeding

A number of crucial decisions face the plant breeder in developing improved cultivars. Because of the impact and modern complexity of these decisions, computer simulation based on effective models can be an important resource for the breeder, particularly in providing guidance on choice of parents and decisions related to various aspects of the integrated breeding approach. Four areas related to use of computer simulation and modeling were explored as outlined in respective chapters: 1) To maximize utility, the simulation tool must be based on effective models of the genome and the process of genetic transmission through generations, the breeding process, and other ‘processes’ involved in genetic recombination, identification and production of new cultivars. Additionally, the statistical methodology employed has ramifications for predicting performance and breeding outcome. We highlighted the role of computer simulation in planning phases of crop genetic improvement, the basics of model building, statistical considerations, and key issues to be addressed. Examples of publicly available simulation software were described (features, functionalities, and assumptions) and new directions for improved/expanded approaches and tools are discussed. 2) Improvement of genome model, through accurate modeling investigation of crossover interference and additive and dominance effects were explored. Crossover interference in maize was modeled by two-pathway methods using doubled haploid data; various levels of crossover interference were found across chromosomes. To challenge the commonly invoked assumption of the infinitesimal model of genetic effects, published data from five quantitative trait loci (QTL) mapping studies were used to derive the distributions of QTL additive effects and dominance coefficients in the form of mixtures of normals. Four separate normal distributions iii with zero mean and different variances were fitted for QTL additive effects of different classes of traits. Dominance coefficients fit a single normal distribution with positive mean, indicating prevalence of additive or partial dominant gene action across many traits. 3) To enhance the predictive ability in new line development, we developed and evaluated a novel method for use in genomic selection. Genomic selection procedures have proven useful in estimating breeding value and predicting phenotype with genome-wide molecular marker information. We proposed a new nonparametric method, pRKHS, which combined the features of supervised principal component analysis and reproducing kernel Hilbert spaces regression, with versions for traits with no/low epistasis, pRKHS-NE, to high epistasis, pRKHS-E. Compared to RR-BLUP, BayesA, BayesB, and RKHS, pRKHS delivers greater predictive ability, particularly when epistasis impacts expression in the trait of interest. Beyond prediction, the new method also facilitated inferences about the extent to which epistasis influences trait expression. 4) A case study involving transgenic conversion of a target hybrid for 15 events explored optimization of parameters in version testing to facilitate a high likelihood of recovery of at least one hybrid conversion with performance equivalent to the unconverted target hybrid. We determined that by creating 5 versions of each parental conversion, then selecting 3 each based on breeding value and yield testing a total of 9 hybrid versions of the conversion facilitated a 95% probability of success. These results had implications for the trait conversion process pertaining to single event conversion and event pyramiding