全基因组混合模型关联分析的极速回归扫描法研究

在全基因组混合模型关联分析(GWMMAS)中,利用剩余多基因遗传力代替方差比值(剩余多基因方差/误差方差),将多基因遗传力求解限制在(0,1)区间内。引入R语言Rcpp Armadillo程序包中的极速线性模型拟合函数(fast Lm Pure函数)快速估计单核苷酸多态性(SNP)效应和完整LMM最大似然值。从由GBLUP估计的性状基因组遗传力出发,逐个高通量SNP的GWMMAS约需4次全基因组回归扫描。当仅关注EMMAX法估计的大效应或高显著水准标记时,GWMMAS运行时间缩短在两次扫描之内。与采用lm函数优化剩余多基因方差比的Fa ST-LMM法相比,极速回归扫描法可成倍提高GWMMAS计算效率。计算机模拟试验证实新方法统计和计算效率,运用极速回归扫描法可高效定位与牙鲆生长性状相关基因位点。

Bare-bones regression scan for genome-wide mixed model association study

In genome-wide mixed model association study (GWMMAS), the article replaced the variance ratios (polygenic variances to residual variances) with polygenic heritabilities and then solved the polygenic heritabilities within the open interval (0,1). Single nucleotide polymorphism(SNP)effects and maximum likelihood values of factored spectrally transformed linear mixed models (FaST-LMM) could be rapidly estimated by the bare-bones linear model fitting function, fastLmPure, in the R/RcppArmadillo package. Launching from the genomic heritability of quantitative traits estimated by genomic best linear unbiased prediction (GBLUP), the single GWMMAS of the high throughput SNPs consumed the same time as approximately four genome-wide regression scans. When focusing only on SNPs having large effects or higher significance levels estimated with the EMMAX algorithm, the computing time of the GWMMAS would be reduced within two genome-wide regression scans. The procedure proposed here greatly improved compututational efficiency for GWMMAS, as compared to the FaST-LMM that optimised residual variance ratios by R/lm function. Computer simulation studies had demonstrated the statistical and computational efficiency of the new method. In addition, to applied the method to powerfully map the gene loci related to the growth traits in Japanese flounder.