家畜全基因组关联分析中的贝叶斯方法  

Bayesian methods for genome-wide association analyses in domestic animals

作　　者：梅步俊[1,2] 王志华[3] 

机构地区：[1]河套学院农学系,内蒙古巴彦淖尔015000 [2]爱荷华州立大学动物科学系,爱荷华州埃姆斯50010 [3]河套学院土木工程系,内蒙古巴彦淖尔015000

出　　处：《黑龙江畜牧兽医》2016年第11期38-44,293,共8页Heilongjiang Animal Science And veterinary Medicine

基　　金：国家自然科学基金项目(31460594);国家留学基金委项目(201308155140);河套学院教学研究项目(HTXYJZ14005)

摘　　要：为了研究贝叶斯统计在家畜全基因组关联分析(GWAS)中的应用,试验采用了贝叶斯多元回归(BMR)方法,主要讨论其理论基础、处理SNPs间局部高相关性和后验Ⅰ型错误率(PER)在控制GWAS中假阳性(FP)问题中的应用,并基于模拟数据来比较该方法的性质。结果表明:1)即使预测变量间存在由连锁不平衡(LD)导致的潜在相关性,BMR也可以成功地应用在全基因组预测问题中;2)在真阳性检测和低假阳性方面,本方法仍然具有较为优良的性质。说明贝叶斯多元回归方法可以得到优良的推断结果,但是QTL加性效应估计值可能出现与模拟数据"真"值正负号相反的情况。To study the application of Bayesian statistics in the genome - wide association analyses （GWAS） in domestic animals, Bayesian mul- tiple regression （BMR） was used in the test. The main contribution of this study was to discuss theoretical basis of BMR, its ability to handle partially high correlation between single nucleotide polymorphisms （SNPs）, and the application of posterior type I error rate （PER） for control- ling false positives （FP） in the GWAS. Comparisons are based on the simulated datasets to assess the properties of this method. The results showed that even though the predictor variables were potentially dependent due to inkage disequilibrium （LD） between SNPs, BMR method was able to successfully use for whole - genomic prediction; the method still had more excellent properties in the aspects of true positive detection and low false positive. The results indicate that using BMR method can get the excellent inference result, but the estimated values of additive effect of QTLs may appear the opposite case in the positive and negative signs of the ＂true＂ values of the simulated data.

关 键 词：贝叶斯多元回归 全基因组关联分析 多重比较 动物育种 计算生物学 

分 类 号：Q78[生物学—分子生物学]