Variance component models in mapping imprinted genes: Statistical theory and applications

Genomic imprinting has been thought to play an important role in seed development in flowering plants. Seed in a flowering plant normally contains diploid embryo and triploid endosperm. Empirical studies have shown that sonic economically important endosperm traits are genetically controlled by imprinted genes. However, the exact number and location of imprinted genes are largely unknown due to the lack of efficient statistical mapping methods. When an iQTL segregates in experimental line crosses, combining different line crosses with similar genetic background can improve the accuracy of iQTLs inference. To make full use of the natural information of sex-specific allelic sharing among sibpairs in line crosses, general statistical variance components frameworks are proposed to map imprinted quantitative trait loci (iQTL) for the diploid tissue and the triploid tissue, individually. Considering the special characteristics of the diploid embryo genome and triploid endosperm genome, new variance components partition methods with respect to the diploid and triploid tissues are developed. An extension to multiple QTL analysis is proposed for both diploid and triploid tissues.;A number of studies have demonstrated that multivariate traits analysis can provide more significant power and higher resolution for major gene detection in linkage analysis (Evans 2002). Furthermore, when a QTL has the pleiotropic effect on several traits, some important biologically interesting hypotheses can be performed successfully under the multivariate traits approach. It is well known that several highly correlated traits appear commonly in endosperm. So the variance components based univariate trait iQTL model is extended to bivariate traits iQTL model for mapping the parent-of-origin effect. It may expedite the process of identifying and eventually cloning genes controlling important endosperm traits.;Except for the wide application of variance components model in flowering plants, variance components analysis has been a standard means in human genetics. In brief, the genetic effect is detected by the significance of the likelihood ratio test. However, true parameters of main interest may be on the boundary of the parameter space under the null hypothesis, thus the regularity condition for declaring asymptotic chi-square distribution of the LRT statistics is not satisfied. The threshold calculation based on current methods often yields conservative hypothesis tests as discussed in a number of studies, especially in multivariate traits cases. To solve this problem, a general approximation form of the LRT under the null hypothesis of no linkage is proposed, and the chi-square mixture proportions are shown to depend on the estimated Fisher information matrix in both univariate and multivariate trait analysis.