J. Wang, M. van Ginkel, R. Trethowan, CIMMYT, Mexico D. Podlich, I. DeLacy and M. Cooper, School of Land and Food Sciences, The University of Queensland, Australia

Introduction

The major objective of plant breeding programs is to develop new genotypes that are genetically superior to those currently available for a specific target mega-environment (ME) or a target population of environments (TPE). To achieve this objective, plant breeders employ a range of crossing and selection methods. For example, at CIMMYT, the most frequently used method from the 1940s till the early 1980s was pedigree selection; modified pedigree/bulk selection started being used in the early 1980s. Today the selected bulk method is being used on certain populations in CIMMYT's bread wheat breeding program.

Generally speaking, quantitative genetics provides much of the framework for designing and analyzing selection methods used within breeding programs. However, assumptions in quantitative genetics are usually made to render some theories mathematically or statistically tractable. Some assumptions can be easily tested or satisfied by experimental designs. Others could never be true for example, assumptions of no linkage and no genotype by environment interaction (GEI). Still other assumptions are difficult to test for example, the existence of epistasis. Therefore, many predictions made in plant breeding programs are based on a relatively simple genotype by environment system.

Computer simulation provides us with a tool to investigate the implications of relaxing some of these assumptions and the effect this would have on the conduct of a breeding program.

The CIMMYT Wheat Breeding Simulation Project is jointly supported by GRDC and the University of Queensland, Australia, and CIMMYT. The aims of this project are to:

• Design a simulation module based on QU-GENE software that will identify opportunities to further improve the efficiency of the CIMMYT wheat breeding and dissemination programs;
• Characterize the target population of environments (TPE) in client countries, including those in Australia, that are relevant to CIMMYT wheat breeding objectives and procedures, and store them in ICIS, and
• Develop a QU-GENE/ICIS software and data exchange interface to enable the use of the genotype and environment characterization information held in ICIS for modelling CIMMYT and Australian wheat breeding strategies using QU-GENE.

Steps of a simulation project

1. Documentation of the CIMMYT wheat breeding program

The initial step toward breeding simulation is to document CIMMYT's wheat breeding programs and expound their operations and activities in a quantitative and breeding/genetic fashion. This detailed description is used for designing simulation software and should include:

• Constitution of entries in the crossing block: elite CIMMYT germplasm, major released cultivars, advanced lines from wide crosses, pathology, etc.
• Parental selection process for crossing and type of crosses (e.g., simple cross, top cross, and backcross).
• Germplasm flow from crossing blocks to yield trials and from there to International Screening Nurseries and Yield Trials (Figures 1 and 2).
• Breeding traits, among cross or family selection intensity, within cross or family selection intensity, sown-grain weight and population size, and harvest method in each generation.

2. Definition of a genotype by environment system

The underlying basis for simulation must be a genotype by environment system. The genes, their locations on the chromosomes, and their frequencies in breeding populations constitute the genetic component of the system. For simulation we only consider those loci with two or more alternative alleles. Some genes have been located on the chromosomes; however, most genes have not, especially for most agronomic and economic traits. For this purpose, we will make educated guesses on the number of genes and temporarily assign these genes on the linkage map. Then we will use historical data such as genetic gains and the magnitude of genotype by environment variation to test the assumption of gene number.

The number of environments in the target mega-environment and their frequencies constitute the environmental component of the system; gene effects under different environments are the interaction part of the system. For simulation, the following information should be specified:

• Genes for traits and gene linkage map, gene frequencies in crossing blocks, and genetic effects (additive, dominance, and epistasis).
• Constitution of the mega-environments.
• Adaptation landscape model for genotype by environment interaction: E (N:K), landscape representation of genetic adaptation (Figure 3).

  Figure 3. Landscape representation of genotypic adaptation in environments: E(N:K)

   

• E: number of environments

• N: number of genes

• K: level of epistasis

3. Ways of comparing different selection strategies in plant breeding (Figure 4)

• Genetic advance or gain
• Number of lines above predetermined checks
• Frequencies of favorable alleles
• Mean and variance of the final selected population
• Gene diversity of the final selected population

Figure 4. A searching process of a selection strategy on the adaptation landscape

4. Development and validation of the simulation module, and implementation of simulation experiments.

Plant breeding issues to be determined by simulation

Many issues in plant breeding can be studied by simulation and field experiments. A few examples:

• Comparison of pedigree selection, modified pedigree/bulk, and selected bulk methodologies that have been used in CIMMYT's wheat breeding programs (Table 1).
• Balance the number of crosses and the size of segregating populations (Table 1).
• Suitable selection intensity for each generation: high selection intensity in early generations or in late generations for a specific trait (Table 1)?
• Effectiveness of different selection sites and their order/sequence in shuttle breeding (Table 1).
• Comparison of simple, top, back, and double crosses in regard to 1) introducing genes from the donor parent and 2) retaining genes from the adapted parent.
• Correlation between parents and their offspring: Can F1 hybrids predict the performance of their advanced lines?
• Ways to better accommodate genotype by environment interaction and epistasis in plant breeding.
• Effective distance between markers and linked genes, and in which generation to apply marker assisted selection (MAS).
• Comparison of breeding/selection/evaluation methodologies to develop germplasm with wide and/or specific adaptation expressing stable yields.

© CIMMYT April 2001

 Kronstad Symposium Poster List | Wheat Program | Wheat Research Results