Studies On Approaches For Mapping QTLs With Complicated Epistasis And Their Interactions With Environment And Methods For Heterosis Prediction

Epistasis and genotype X environment interaction exist universally and have been testified by a lot of researches. These effects play an important role in inheritance, evolution and heterosis in plants. However, most of current methods for QTL mapping cannot simultaneously analyze epistases and their interaction with environments. The methods proposed by Wang et al (1999) could analyze additive effects, additive by additive epistatic effects and their interaction with environments, but could not estimate dominance effects and epistatic effects related to dominance. Considering the characteristic of common mapping populations and basing on the principles of composite interval mapping through mixed linear model suggested by Zhu (1998), Zhu and Weir (1998), a novel strategy for QTL mapping was proposed in this study, which is adapt to analyze F, population, especially permanent F2 population. In the present study, a new approach for heterosis prediction was also proposed. Monte Carlo simulations were applied to compare the reliability of new methods systematically. The experiment data collected in 1999 and 2000 through permanent F, population of rice were analyzed by computer software developed in this study. The main contents and conclusions of this study are as follows:1. Genetic models were constructed for QTL mapping by two-dimensional searching. Corresponding analysis methods were also proposed, which could estimate additive effects, dominance effects, epistatic effects of additive by additive, additive by dominance, dominance by additive, dominance by dominance, and could predict their interaction effects with environments. Formula for prediction of general heterosis and interaction heterosis of single-cross offspring based on QTL effects were deduced.2. Monte Carlo simulations were conducted to study the new approaches of QTL mapping, the results indicated that general least squares (GLS) method, which was widely applied in mixed linear model, could unbiasedly estimate all genetic main effects, including additive effects, dominance effects and epistatic effects of additive by additive, additive by dominance, dominance by additive, dominance by dominance. The interaction effects between genetic main effects and environments could also be predicted unbiasedly by linear unbiased prediction (LUP). The heterosis prediction based on QTL effects was also unbiased.3. The estimation of additive and dominance effects through QTL analysis by one-dimensional search while ignoring epistases showed similar accuracy to that by two-dimensional search by including epistases. Existence of epistases could decrease the precision for the estimation of additive and dominance effects. The estimation of genetic main effects would be biased if ignoring the interaction effects of QTL X environment (QE). Joint analysis of multiple environments including QE interaction effects could increase the precision for the estimation of genetic main effects.4. Using recombinant inbred line (RJL) of rice derived from (Zhenshang 97BX Minghui 63), permanent F2 population was constructed by random mating among RIL lines. Present study focused on the investigation and analysis of two traits, heading date and kilo-grain weight.5. Twenty-one and seventeen QTLs were identified in QTL mapping analysis of heading date and kilo-grain weight respectively. The results oft-test indicated that few genetic main effects of QTLs found were significant, among which additive by additive epistatic effects and additive effects were most significant for QTLs controlling heading date, additive effects were most significant for QTLs controlling kilo-grain weight. The direction of the effect values indicated that these QTLs distributed in Zhenshang 97B and Minghui 63 in repulsion phase. Jackknife was applied to predict and test QE interaction effects, but only few of significant QE interaction effects were found.6. The results of heterosis prediction based on QTL effects showed that heading date had negative general...