中等稀疏条件下基因交互作用的两步Bayes方法

植物遗传与基因组学研究表明许多重要的农艺性状有影响的基因位点不是稀疏的,受到大量微效基因的影响,并且还存在基因交互项的影响.本文基于重要油料作物油菜的花期数据,研究中等稀疏条件下的基因选择问题,提出了一种两步Bayes模型选择方法.考虑基因间的交互作用,模型的维数急剧增长,加上数据结构特别,通常的变量选择方法效果不好....

Two-step Bayesian method for detecting gene-gene interactions with moderate sparsity

Based on flowering time trait in Brassica napus,we propose a two-step Bayesian analysis method for model selection under moderate sparsity.The phenotype is quantitative and the genotype takes two values 1 and-1.By considering the gene-gene interaction,the model dimensionality increases dramatically so that the existing variable selection methods cannot handle this problem.In this paper,we propose a two-step selection method.First,we apply the fused Kolmogorov filter to screen out some obviously unimportant variables.Then,based on the remaining variables,we add interaction effect to the model.To overcome the heavy computation burden of the Bayesian method,we add indicator variables to every candidate effects and choose the active effect via posterior distribution of the indicator variable.The simulation results show that our proposed method performs better delivering higher prediction accuracy than those without the indicator function.Furthermore,we apply the proposed model to analyze a flowering time data in the Brassica napus population in multiple environments.