Modelling And Simulation Of CIMMYT Maize Breeding Programs

Conventional plant breeding largely depends on phenotypic selection and breeder's experience,therefore the breeding efficiency is low and predictions are inaccurate. Along with the fast developmentin molecular biology and biotechnology, a large amount of biological data is available for geneticstudies of important breeding traits in plants, which in turn allows conducting genotypic selection in thebreeding process. The optimization of breeding strategies can help improve the breeding efficieny, whilethe efficient use of known gene and genetic information can improve the predictability in breeding.Computer simulation as a tool has been applied successfully in many special plant breeding studies.QuLine is a simulation breeding program for developing inbred lines, which has been used to comparetwo selection strategies, to improve the preceision of parental selection using known gene information,to predict cross performance, to study the effects on selection of dominance and epistasis, to optimizemarker assisted selection to pyramid multiple genes efficiently, and to apply design approach inbreeding. QuHybrid is a computer software that simulates breeding programs for developing hybrids.While retaining most QuLine functionalities, QuHybrid is able to make testcross and hybridperformance predictions; it is thus possible to simulate the hybrid breeding program. In the presentstudy, hrbrid breeding program simulation tool QuHybrid was applied to model and simulate CIMMYTmaize breeding programs, to simulate and optimize breeding strategies for improving pro-Vitamin Acontent in maize, and to compare of the efficiency of different testers on maximizing genetic gain in ahybrid maize breeding program.The main results obtained in this study are summarized below.1. Documentation of CIMMYT maize breeding programs: The history of CIMMYT GlobalMaize Program (GMP), maize breeding mega-environments, subprograms of maize breeding programand their objectives, maize germplasm, and maize breeding methodology for developing inbred linesand hybrids, implementating DH, MARS, and GS were introduced. This information is necessary whendefining the input files for running simulation tool QuHybrid. Breeding flows for developing inbredlines and hybrids, implementating DH, MARS, and GS can be used as examples in QuHybridsimulation experiments.2. Simulation and optimization of breeding strategies for improving pro-Vitamin A contentin maize: Biofortification for pro vitamin A content (pVAC) of modern maize inbreds and hybrids is afeasible way to deal with vitamin A deficiency in rural areas in developing countries. The objective ofthis study was to evaluate the probability of success of breeding strategies when transferring the highpVAC present in donors to elite modern adapted lines. For this purpose, a genetic model including4additive major genes was built based on previous genetic studies,20breeding scales and9differentselection schemes including phenotypic selection (PS) and marker assisted selection (MAS) weresimulated and compared. Results indicated that the current breeding scale needs to be doubled so as tohave a probability higher than90%of achieving the HarvestPlus target on pVAC in one breeding cycle.MAS for simultaneously selecting all pVAC genes and a combined scheme for selecting two majorpVAC genes by MAS followed by ultra performance liquid chromatography (UPLC) screening for theremaining genetic variation on pVAC were identified as being most effective and cost-efficient. In addition, the size of the early generation population and the number of crosses had minor effects onbreeding efficiency under a fixed resource, when60or more crosses are made. Breeding on a largerscale was more efficient both genetically and economically. The assumpution of pVAC were controlledby four additive genes is releative simple. When target traits were controlled by more genes and morecomplicated effects are involved, it can be expected that the relative gain of MAS compared with PSwill be reduced. So the conclusion from pVAC cannot be extended to other traits, whose genetic modelsare not similar to the model of pVAC. However, the approach presented in this study could be used as ageneral way for quantifying probability of success and comparing different breeding schemes in otherbreeding programs.3. Comparison of different testers on maximizing genetic gain in maize hybrid breeding:Three testers commonly used in most maize hybrid breeding programs were considerd, i.e., an inbredline tester with the best per se performance, an inbred line tester with the worst per se performance, anda single cross hybrid fromed by these two inbred lines. Breeding efficiencies of these3testers werecompared in12genetic models, and the best tester was identified for each genetic model. Resultsindicated that the three testers resulted in different breeding efficiencies in different genetic models. It isfeasiable to use per se performance of inbred line, i.e., the gene frequency of favorable alleles, as acriterion to identify the best tester, but the resuts varied in different genetic models. Which tester is thebest depends upon the genetic models of target traits. In our simulated genetic models, efficiency of thesingle cross tester was in the middle compared with the other two testers. It may be reasonable to use asingle cross as tester in selecting elite maize hybrids. In QuHybrid, breeding methods can be comparedby the line per se genetic gain, testcross genetic gain, and performance of all potential F1hybridsbetween the final selected inbred lines and lines in the other heterotic group. The three criteria wereused in this study to compare the efficiency of different testers, which may also be used as a generalway for comparing different breeding schemes in other hybrid breeding programs.Many simulation experiments have been designed and conducted to solve complex questions inplant breeding and quantitative genetics. Some simulations not only answered the questions that breederconcerned but also found out some facts which breeders did not realize, which indiactes the usefulnessof breeding simulation. QU-GENE can handle more complicated models, and QuHybrid can simulatealmost all breeding activities in a hybrid breeding program. The combination of these two simulationprograms can solve most, if not all, questions existed in hybrid breeding program. In CIMMYT,investigations using the QuLine and QuHybrids tools are going on for the efficient use of highthrought-put genotyping data in wheat and maize breeding, comparison of various genotype tophenotype prediction models, optimization of the genomic selection in wheat and maize, the efficientbreeding methods to introduce exotic favorable alleles to CIMMYT elite breeding materials.