Yang Deguang, Han Shanshan, Jiang Dan, Yang Runqing, Fang Ming
College of Agriculture,Northeast Agricultural University,Haerbin,150030,P.R. China.
Life Science College,Heilongjiang Bayi Agricultural University,Daqing,163319,P.R. China.
Genet Res (Camb). 2015 Apr 10;97:e6. doi: 10.1017/S0016672315000014.
Bayesian shrinkage analysis estimates all QTLs effects simultaneously, which shrinks the effect of "insignificant" QTLs close to zero so that it does not need special model selection. Bayesian shrinkage estimation usually has an excellent performance on multiple QTLs mapping, but it could not give a probabilistic explanation of how often a QTLs is included in the model, also called posterior inclusion probability, which is important to assess the importance of a QTL. In this research, two methods, FitMix and SimMix, are proposed to approximate the posterior probabilities. Under the assumption of mixture distribution of the estimated QTL effect, FitMix and SimMix mathematically and intuitively fit mixture distribution, respectively. The simulation results showed that both methods gave very reasonable estimates for posterior probabilities. We also applied the two methods to map QTLs for the North American Barley Genome Mapping Project data.
贝叶斯收缩分析同时估计所有数量性状基因座（QTL）的效应，它会将“不显著”的QTL效应收缩至接近零，从而无需进行特殊的模型选择。贝叶斯收缩估计在多个QTL定位中通常具有出色的表现，但它无法给出一个QTL在模型中出现频率的概率解释，即后验包含概率，而后验包含概率对于评估QTL的重要性很重要。在本研究中，提出了两种方法，即拟合混合法（FitMix）和模拟混合法（SimMix）来近似后验概率。在估计的QTL效应的混合分布假设下，拟合混合法和模拟混合法分别从数学和直观上拟合混合分布。模拟结果表明，这两种方法对后验概率都给出了非常合理的估计。我们还将这两种方法应用于北美大麦基因组图谱项目数据的QTL定位。
