多性状全基因组关联分析新方法的探索

In the past several years genome-wide association study has been widely adopted in the genetic analysis of complex traits in plants owing to short time in the construction of mapping population, high density of SNP markers, and excessive recombinant in the breeding of crop cultivar. However, almost all the current methodologies are available only for a single quantitative trait. This results in high false positive rate and low power in the detection of quantitative trait loci (QTL). As we know, multi-trait joint analysis can increase the power and precision, and distinguish pleiotropic QTL from multiple linked QTL. Based on epistatic association study for quantitative traits and multi-QTL mapping for resistance traits in crop cultivars, in this study we will investigate the algorithm of parameter estimation for multi-trait joint analysis, and its purpose is to set up the technologic platform of multi-trait genome-wide association study. Once the new method is validated by Monte Carlo simulation experiments, the corresponding software will be developed. The validated method and software will be used to carry out multi-trait genome-wide association studies for seed size and shape traits in 286 soybean cultivars. If doing so, the QTL cluster for the above traits in our previous studies may be identified. In other words, the QTL cluster is derived from one of the two situations: pleiotropic QTL and multiple linked QTL. In this project two SCI papers will be published and one software will be developed.

由于人工选择的品种群体构建时间短、SNP标记密度高和利用历史重组机会多，致使近年来关联分析在植物遗传研究中应用较为广泛。但是，目前的分析方法一般采用单性状分析，假阳性率较高，功效有待提高。研究已表明，多性状联合分析能提高遗传分析功效与精度，剖析复杂性状的一因多效。在已提出品种群体数量性状上位性关联分析和抗性性状多QTL检测新方法的前期工作基础上，本项目将研究多性状联合遗传分析的参数估计算法；进而构建多性状联合的全基因组关联分析技术平台，经Monte Carlo计算机模拟研究验证后，研制相应的计算机软件包；用于286个大豆品种群体籽粒大小与形状性状的多性状全基因组关联分析，揭示这些相关性状的遗传基础是一因多效还是基因连锁。预计发表SCI论文2篇，研制软件1套。

当前的全基因组关联分析是基于背景控制的单标记分析。由于涉及多重矫正，使每次检测的显著水平过低，导致一些重要位点丢失。为解决这一问题，提出了一种多位点快速检测的关联分析新算法FASTmrEMMA。用该算法，将多基因效应和剩余变异转换为正态离差，借助最小角估计，实现在控制其它染色体标记变异情况下对每条染色体寻找潜在关联标记，将所有潜在关联标记放入同一模型进行经验Bayes估计和似然比检验，最终鉴定出真QTN，这称为pLARmEB方法。两种方法经Monte Carlo模拟研究表明，增加了两种方法的QTN检测功效和效应估计精度，降低了假阳性率。研制了计算机软件包，以构建GWAS技术平台。分析了286个大豆品种群体籽粒大小与形状性状的QTNs，并挖掘优异等位基因。进一步实施了上述籽粒大小或形状的多性状联合分析，发现了更多的显著SNPs，揭示这些相关性状的遗传基础是一因多效还是基因连锁。发表SCI论文3篇，研制2种新方法的R软件包。