Mixed Linear Model Approach For Dissecting Genetic Architecture Of Complex Traits And Software Development

Crop seeds, as primary source of human food, animal feed and industrial raw materials, are mainly comprised of endosperm and embryo, which is triploid and diploid respectively. Most agronomic traits, including crop seed traits, are complex traits, which are affected not only by individual loci, but also by gene-gene interactions and gene-environment (GE) interactions. In addition, as advance in high-throughput sequencing technology, genome wide association study has been a standard approach for dissecting genetic architecture of human disease and complex agronomic traits. However, most genome wide association studies are based on additive model and focusing on single trait analysis.In this dissertation, we developed new statistical methods for detecting gene-gene interaction and GE interaction for both seed traits linkage mapping and multiple traits association mapping. Monte Carlo simulation and real data analysis demonstrated the unbiasedness and robustness of the proposed methods. The main contents of the dissertation are as follows,The first chapter introduces the recent development in linkage analysis and genome wide association analysis as well as corresponding statistical software. In addition, we also introduce mixed linear model approach for hypothesis test and parameter estimation.The second chapter introduces a newly developed statistical method and computational algorithm for crop seed traits mapping based on mixed linear models. Seeds of flowering plants develop from double fertilization and play a vital role in reproduction and supplying food for human and animals. Multiple genetic systems, e.g., maternal, embryo and/or endosperm genomes, are together involved in the genetic variation of seed traits. Understanding the genetic mechanism of seed traits is a major challenge because of its complex mechanism of multiple genetic systems, especially for the epistasis within or between different genomes and their interaction with environments. According to the genetic features of seed characters, two statistical models were proposed for mapping QTLs with epistasis and GE interactions underlying endosperm and embryo traits of interest, respectively. Our models integrating maternal and offspring genomes into one mapping framework, can accurately analyze the maternal additive and dominant effects, the endosperm/embryo additive and dominant effects, the epistatic effects of two loci in one or two different genomes, as well as interaction effects of each genetic component of QTL with environment. The mapping population can be generated either from double back-cross of immortalized F2 (IF2) to the two parents, from random-cross of IF2, or from selfing of IF2 population. Extensive simulations under different heritabilities and model parameters were performed to investigate the statistical properties of the models. A set of real data on cottonseed was used as an example to demonstrate our methods. A software package, QTLNetwork-Seed-1.0.exe, was developed for QTL analysis of seed traits.The third chapter introduces a newly developed statistical method and computer algorithm of joint association analysis for multiple traits based on multivariate linear mixed models. With the development of high-throughput sequencing technologies, genome-wide association studies (GWAS) have been a standard approach to discover the genetic determinants of complex traits. However, a major concern in GWAS is that not only individuals but also genetic loci are correlated. Linear mixed model approach has emerged as a flexible and efficient method to account for the population structure. Furthermore, since many complex disease syndromes consist of a large number of highly related, rather than independent, clinical or molecular phenotypes, it is natural to identify genetic variants that are associated with multiple traits simultaneously. Moreover, it is known that the genetic factors may not simply accumulate but could also interact with other loci or environment. However, most GWAS performed to date have been focusing on testing additive effects for associations with individual traits. To combat these challenges, we extended a multivariate linear mixed model approach, which incorporated epistasis and GE interaction effects for mapping genetic causal variants that are phenotypic specific and common for multiple correlated traits. The approach considers both within trait variance and between trait variance simultaneously. Extensive simulations under different residual correlation, heritability, and model parameters supported that our approach provides increased power and precision to detect pleiotropic loci that affect more than one trait. The application on rice and maize data also showed the robustness of our method. A software package, JAMT (Joint Analysis for Multiple Traits), was developed for multiple traits association analysis.