Li Xuan, Qiu Weiliang, Morrow Jarrett, DeMeo Dawn L, Weiss Scott T, Fu Yuejiao, Wang Xiaogang
Department of Mathematics and Statistics, York University, 4700 Keele Street, Toronto, ON, M3J1P3, Canada.
Channing Division of Network Medicine, Brigham and Women's Hospital, Harvard Medical School, 181 Longwood Avenue, Boston, MA, 02115, United States of America.
PLoS One. 2015 Dec 18;10(12):e0145295. doi: 10.1371/journal.pone.0145295. eCollection 2015.
Variable DNA methylation has been associated with cancers and complex diseases. Researchers have identified many DNA methylation markers that have different mean methylation levels between diseased subjects and normal subjects. Recently, researchers found that DNA methylation markers with different variabilities between subject groups could also have biological meaning. In this article, we aimed to help researchers choose the right test of equal variance in DNA methylation data analysis. We performed systematic simulation studies and a real data analysis to compare the performances of 7 equal-variance tests, including 2 tests recently proposed in the DNA methylation analysis literature. Our results showed that the Brown-Forsythe test and trimmed-mean-based Levene's test had good performance in testing for equality of variance in our simulation studies and real data analyses. Our results also showed that outlier profiles could be biologically very important.
可变的DNA甲基化与癌症和复杂疾病有关。研究人员已经鉴定出许多DNA甲基化标志物，其在患病个体和正常个体之间具有不同的平均甲基化水平。最近，研究人员发现，在不同组个体之间具有不同变异性的DNA甲基化标志物也可能具有生物学意义。在本文中，我们旨在帮助研究人员在DNA甲基化数据分析中选择正确的方差齐性检验。我们进行了系统的模拟研究和实际数据分析，以比较7种方差齐性检验的性能，其中包括DNA甲基化分析文献中最近提出的2种检验。我们的结果表明，在我们的模拟研究和实际数据分析中，Brown-Forsythe检验和基于截尾均值的Levene检验在检验方差齐性方面表现良好。我们的结果还表明，异常值分布在生物学上可能非常重要。
