Genetic Architecture of Domestication and other Complex Traits in Maize

Research on domestication informs our understanding of the genetic architectures of important traits and can guide efforts to identify causal genetic variants of crop. Numerous morphological traits have changed during domestication from teosinte to maize. However, standing variation still exists for several domestication-related traits in modern maize. The genetic architecture for standing variation remaining in maize for these domestication-related traits is unknown. In Chapter 2, the maize Nested Association Mapping population (NAM) and a diversity panel were used to estimate the proportion of variation due to polygenic, small-effect QTL versus larger effect variants; compare the genomic positions of larger effect variants to the known locations of domestication; and partition the genetic variance using variance components analysis methods. Additive polygenic models explained most of the genotypic variation for domestication-related traits; no large effect loci were detected for any trait. Previously defined improvement sweep regions were associated with more trait variation than expected based on the proportion of the genome they represent. Small effect polygenic variants (enriched in improvement sweep regions of the genome) are responsible for most of the standing variation for domestication-related traits in maize. Linear mixed models are widely used in humans, animals, and plants to conduct genome-wide association studies (GWAS). Different from human datasets, experimental units for plants are typically multiple-plant plots of families or lines that are replicated across environments. This structure leads to computational challenges to conducting a genome scan on plot-level data. Two-stage methods have been proposed to reduce the complexity and increase the computational speed for GWAS. However, the appropriate dependent variable to use in the second stage analysis and how to handle unbalanced datasets are not clear. In Chapter 3, I developed a weighted two-stage analysis to reduce bias and improve power of GWAS while maintaining the computational efficiency of two-stage analyses. Power and false discovery rate of one-stage, different two-stage models and new weighted analysis are compared by simulation based on real marker data of a diverse panel of maize inbred lines. Only weighted two-stage GWAS has power and false discovery rate similar to the one-stage analysis for severely unbalanced data in simulation. The weighted two-stage analysis method was implemented in a free open source software TASSEL. Genetic diversity reduced severely due to selection and bottleneck effect during domestication of maize from teosinte. However, the gene pool of teosinte might harbor agriculturally beneficial alleles. Due to their linkage with a genomic background that is unadapted, measuring the potential benefit of teosinte alleles is challenging. A population called the ZeaSynthetic population was developed as a bridge to investigate teosinte specific alleles in a common maize background. Genotypes of 1846 parents were obtained by genotyping by sequencing and phenotypes of 923 pairs of S1 and S0 full-sib progeny families were measured across six environments. Due to unusual population structure and experimental design, there is no ready to use software available for analysis. In Chapter 4, I used a linear mixed model to estimate additive and dominance effects at each SNP and used simulation to evaluate power and bias of genetic effect estimates from this model. Simulation results showed that the linear mixed model is a reasonable model with low bias estimating genetic effects and high power detecting QTLs. Analysis of the real data showed that loci with rare alleles and loci in lower recombination regions of the genome tend to have larger additive and dominance effects. These results suggest that recessive deleterious alleles tend to be concentrated in lower recombination regions, and that favorable alleles for agriculture are more likely to be at higher starting frequencies and found at loci in higher recombination regions