基于LASSO算法的多QTL定位的模拟研究

Simulation Study of Mapping Multiple QTL Based on Lasso Algorithm

此篇文章中,应用一种基于LASSO的算法去同时定位多个数量性状位点(QTL)。在QTL定位方法中,迭代加权最小二乘法(IRLS)和极大似然法(ML)在参数估计准确性和检测效率方面效果几乎一致。但是在参数估计的稳健性和计算速度方面,迭代加权最小二乘法又明显优于极大似然估计。结合参数的先验分布信息,基于贝叶斯理论的极大似然估计能够分析多QTL模型。然而迭代加权最小二乘估计不能很好的检测多QTL。目前贝叶斯分析已经成为一个多QTL定位的重要的途径,但它主要缺点是计算时间过长,并缺乏简单有效的显著性检验。通过循环坐标下降的LASSO方法可将全模型的系数同时压缩并使之趋进于零,因此该方法能应用于快速同时估计整个基因组的非零遗传效应位点。在这个研究中,应用基于LASSO的算法去同时定位多QTL。模拟证明LASSO方法比迭代加权最小二乘法具有更高的估计精度和检测效率。

In this article,we applied an algorithm based on LASSO to mapping multiple quantitative trait loci（QTL）.In terms of parameter estimation and power of detection,the iteration reweight least square （IRLS）method had a very similar result compared with maximum likelihood（ML）method.But the IRLS was better than ML in terms of computing speed and robustness.Combined with priors of parameters and Bayesian theory,ML can map multiple QTL.In a single QTL model,IRLS has very limited statistical power in detecting multiple QTL.Bayesian has become an important approach to map multiple QTL.But the main shortcoming of the Bayesian analysis is slow convergence,and lack of a simple and efficient significant test.The LASSO with cyclic-coordinate descent step can shrink coefficients toward to zero simultaneously,so this method was employed to efficiently estimate non-zero genetic effect of each locus scanned over entire genome.In this research,we adopted LASSO to map multiple QTL.Simulations demonstrate that LASSO had an better performance than IRLS in terms of parameter estimation and power of detecting QTL.