DNA methylation plays a vital role in transcription regulation. Reduced representation bisulfite sequencing (RRBS) is becoming common for analyzing genome-wide methylation profiles at the single nucleotide level. A major goal of RRBS studies is to detect differentially methylated regions (DMRs) between different biological conditions. The previous tools to predict DMRs lack consistency. Here, we simulated RRBS datasets with significant attributes of real sequencing data under a wide range of scenarios, and systematically evaluated seven DMR detection tools in terms of type I error rate, precision/recall (PR), and area under ROC curve (AUC) using different methylation levels, sequencing coverage depth, length of DMRs, read length, and sample sizes. DMRfinder, methylSig, and methylKit were our preferred tools for RRBS data analysis, in terms of their AUC and PR curves. Our comparison highlights the different applicability of DMR detection tools and provides information to guide researchers towards the advancement of sequence-based DMR analysis.

DNA 甲基化在转录调控中起着至关重要的作用。减少代表性亚硫酸氢盐测序 (RRBS) 在分析单核苷酸水平的全基因组甲基化谱方面变得越来越普遍。RRBS 研究的一个主要目标是检测不同生物条件之间的差异甲基化区域 (DMR)。之前预测 DMR 的工具缺乏一致性。在这里，我们模拟了多种场景下具有真实测序数据显着属性的RRBS数据集，并从I型错误率、准确率/召回率（PR）和ROC曲线下面积（AUC）方面系统地评估了七种DMR检测工具使用不同的甲基化水平、测序覆盖深度、DMR 长度、读取长度和样本大小。DMRfinder、methylSig 和methylKit 是我们首选的RRBS 数据分析工具，就其 AUC 和 PR 曲线而言。我们的比较突出了 DMR 检测工具的不同适用性，并提供了指导研究人员推进基于序列的 DMR 分析的信息。