用逐步回归鉴别上位性QTL互作标记的模拟研究(英文)

Simulation Studies on Identification of Interaction Markers Representing QTL Epistasis by Stepwise Regression

利用蒙特卡罗模拟对逐步回归分析在鉴别代表ＱＴＬ 上位性的互作遗传标记上的应用进行了详细研究⒚结果表明，逐步回归分析能够有效、准确地鉴别出反映ＱＴＬ上位性的互作遗传标记，并优于当前常用的双向方差分析⒚ＱＴＬ之间的紧密连锁会导致互作标记鉴别的分辨率降低，并使鉴别的效率出现扭曲⒚大样本和／或高遗传率可以从总体上提高互作标记准确鉴别的效率，而特定的互作标记鉴别的效率在很大程度上取决于它所代表的ＱＴＬ上位性的相对贡献率⒚显著水平也是影响互作标记鉴别效率的一个重要因子⒚本文建议在几种不同的显著水平下进行若干次逐步回归分析。

The usefulness of stepwise regression in identifying interaction markers that represent QTL epistasis was examined in detail by Monte Carlo simulations. It was indicated that stepwise regression method was powerful in identifying correct interaction markers, and better than two way ANOVA that is currently used for the same purpose. Close linkage tended to reduce the resolution of identifying interaction markers and distort detection power. Large sample size and/or higher heritability could generally increase the chance of correct interaction markers being identified. The chance of identifying a specific marker interaction relied heavily on the relative contribution of the QTL epistasis that the marker interaction represented. Significance level was also an important factor affecting the power of identifying correct marker interactions. Several rounds of analyses under different significance levels were suggested with additional consideration of heritability and sample size used.