DNA methylation profiling for 850k reannotation and normalization evaluation [450k]. DNA methylation profiling for 850k reannotation and normalization evaluation [450k]

Illumina Infinium DNA Methylation (5mC) profiling arrays are a popular technology to measure genome-scale distribution of 5mC at low cost and high throughput, especially in cancer and other complex diseases. Following the success of the HumanMethylation450 array (450k), Illumina released the MethylationEPIC array (850k) featuring increased coverage of enhancers in addition to regulatory regions primarily covered by the 450k (i.e. promoters, gene bodies). Despite its widespread use, the analysis of 850k data remains suboptimal as it mostly still relies on Illumina’s default annotation, which underestimates enhancers and long noncoding RNAs (lncRNAs). We thus developed an approach, based on ENCODE and LNCipedia databases, that greatly improves Illumina’s default annotation of enhancers and long noncoding transcripts. Comparisons between the re-annotated 850k and its precursor, the 450k, or RRBS, another high-throughput 5mC profiling technology, revealed that the 850k covers at least three times more enhancers and lncRNAs than the other two technologies. We further investigated the reproducibility of the three technologies and applied various normalisation methods to 850k data, showing that most of them reduce variability to a level below that of RRBS. When analyzed with our new annotation and normalization pipeline, profiling for 5mC changes in breast cancer biopsies with the 850k highlighted aberrant enhancers methylation as the predominant feature, confirming previous reports. In conclusion, our study provides an updated analysis pipeline for 850k data based on a refined probe annotation and normalization that allows for the improved analysis of methylation at enhancers and long noncoding transcripts. Overall design: Total genomic DNA from HCT116 cells wild type, HCT116 cells double-knockout for DNMT1 and DNMT3B were analysed for 5mC changes using Infinium 450k

因美纳（Illumina）Infinium DNA甲基化（5-甲基胞嘧啶，5mC）分型芯片是一项广受欢迎的技术，可低成本、高通量地检测全基因组范围的5mC分布，在癌症及其他复杂疾病研究中应用尤为广泛。在人类甲基化450芯片（HumanMethylation450 array，450k）取得成功后，因美纳推出了甲基化EPIC芯片（MethylationEPIC array，850k），除覆盖450k芯片主要靶向的调控区域（即启动子、基因体）外，还增强了对增强子的覆盖度。尽管该芯片应用广泛，但850k数据的分析仍未达最优，当前多数分析仍依赖因美纳的默认注释文件，这类注释对增强子和长链非编码RNA（long noncoding RNAs，lncRNAs）的覆盖存在低估。为此，本研究基于DNA元件百科全书（ENCODE）和LNCipedia数据库开发了一套分析方法，可显著优化因美纳默认注释中对增强子及长链非编码转录本的注释效果。对重新注释后的850k芯片与其前代产品450k，以及另一款高通量5mC分型技术——简化代表性亚硫酸氢盐测序（Reduced Representation Bisulfite Sequencing，RRBS）进行比较后发现，850k芯片对增强子和lncRNA的覆盖度至少是另外两种技术的三倍。本研究进一步探究了这三种技术的重复性，并将多种标准化方法应用于850k数据，结果显示多数标准化方法可将数据变异度降至低于RRBS的水平。使用本研究开发的新型注释与标准化分析流程进行分析时，利用850k芯片检测乳腺癌活检组织的5mC变化情况，结果凸显异常增强子甲基化作为主要特征，这与此前的研究报道一致。综上，本研究基于优化的探针注释与标准化流程，为850k芯片数据提供了一套更新后的分析方案，可实现对增强子及长链非编码转录本甲基化水平的更精准分析。实验设计：采用Infinium 450k芯片，对野生型HCT116细胞以及DNA甲基转移酶1（DNMT1）和DNA甲基转移酶3B（DNMT3B）双敲除的HCT116细胞的全基因组DNA进行5mC变化分析。