论文部分内容阅读
利用高密度分子标记,在基因组水平上预测育种值已经在动植物遗传育种研究中得到应用,但是高密度标记也带来许多统计和计算上的问题.为了解决这些基因组选择中的问题,产生了很多不同的方法,包括RR-BLUP,GBLUP,Bayes A,Bayes B,Bayes Cπ和Bayesian LASSO等.本文将这些方法用于一组小麦数据集的分析,同时模拟了不同数目QTL和不同遗传率情况下各种方法分析结果的差异.研究结果表明:在确定基因组选择方法时,要充分考虑所研究性状的遗传结构.如果确认某种性状由较少的大效应QTL控制时,各种方法预测能力的差异较大,应选择Bayes Cπ.如果QTL数目中等,各种方法预测能力的绝对差异较小,但是仍然发现Bayes A优于其他方法.如果性状由大量的微效基因决定,各种方法之间几乎找不到显著的差异,不过此时无论是在模拟分析还是在小麦实际产量的预测中,RR-BLUP都略优于其他方法,说明在这种情况下RR-BLUP是有效的方法.
The use of high-density molecular markers to predict breeding values at the genome level has been used in animal and plant genetics and breeding studies, but high-density labeling also brings about many statistical and computational problems.In order to solve these problems in genome selection, Many different methods, such as RR-BLUP, GBLUP, Bayes A, Bayes B, Bayes Cπ and Bayesian LASSO, etc. These methods are applied to the analysis of a set of wheat datasets and simulate different numbers of QTLs and different hereditary rates The results show that the genetic structure of the traits studied should be fully taken into account when determining the genomic selection method, and that the ability to predict a given method is less predictive when it is determined that a trait is controlled by fewer large-effect QTLs Should be chosen Bayes Cπ. Bayes A is still found to outperform other methods if the QTLs are medium in magnitude and the absolute difference in predictive power of each method is small. If traits are determined by a large number of microsatellite genes, Almost no significant difference was found between the two methods, but RR-BLUP was slightly better than the other methods both in the simulation analysis and in the forecast of actual wheat production, It shows that RR-BLUP is a valid method in this situation.