metdiff

在本文中，我们报道了MeTDiff，一种用于MeRIP-Seq数据集中差异甲基化分析的新计算工具。MeTDiff允许生物学家在两种病例对照条件（例如，正常与癌症）中进行甲基化差异的转录组范围的定量比较。

在不同的模拟MeRIP-Seq数据集上评估MeTDiff以研究具有不同重复的生物学差异的影响以及偏向读数和不同程度的差异甲基化的影响。结果显示，在所有这些模拟条件下，MeTDiff可以比exomePeak更准确和稳健地鉴定差异甲基化位点。

在真实的MeRIP-Seq数据集（KO-FTO）上检查MeTDiff，结果表明MeTDiff比exomePeak获得更高的灵敏度和更高的特异性。

我们还研究MeTDiff的特性早米6在KO-FTO，KO-METTL14和KO-WTAP数据集A差异甲基化。MeTDiff的预测与每个实验的预期结果一致，因为当FTO被击倒时观察到压倒性的超DMS，而在KO-METTL14和KO-WTAP中检测到显着的低DMS。在METTL14和KO-WTAP中的低和超DMS中发现了独特的基序“AGGAG”。进一步分析揭示该基序与剪接因子的基序的紧密连接，提示米的潜在作用6甲调节剪接。最后，观察到mRNA和lncRNA中高和低DMS的不同分布，这可能表明高甲基化和低甲基化具有不同的功能。

In this article, we reported MeTDiff, a new computational tool for differential methylation analysis in MeRIP-Seq datasets. MeTDiff allows biologists to perform transcriptome-wide quantitative comparisons of methylation difference in two case-control conditions (e.g., normal vs. cancer).

MeTDiff was evaluated on different simulated MeRIP-Seq datasets to investigate the effect of biological variances with varying replicates as well as the influence of biased reads and different degree of differential methylation. The results showed that MeTDiff can more accurately and robustly identify the differential methylation sites than exomePeak in all of these simulated conditions.

MeTDiff was examined on the real MeRIP-Seq datasets (KO-FTO) and the results demonstrated that MeTDiff achieved both higher sensitivity as well as higher specificity than exomePeak.

We also examined the characteristics of MeTDiff predated m6A differential methylation in KO-FTO, KO-METTL14 and KO-WTAP datasets. MeTDiff's predictions are consistent with the expected outcome of each experiment in that overwhelming hyper-DMSs were observed when FTO was knocked down, whereas significant hypo-DMSs were detected in KO-METTL14 and KO-WTAP. A unique motif “AGGAG” was discovered in both hypo- and hyper-DMSs in METTL14 and KO-WTAP. Further analysis revealed a close connection of this motif with the motif of splicing factors, suggesting a potential role of m6A in regulating splicing. At last, distinct distributions of hyper- and hypo-DMSs in mRNA and lncRNA were observed, which may suggest that hyper- and hypo-methylation assume different functions.

 

链接：

https://ieeexplore.ieee.org/document/7052329/figures#figures