An evaluation of analysis pipelines for DNA methylation profiling using the Illumina HumanMethylation450 BeadChip platform

Abstract

The proper identification of differentially methylated CpGs is central in most epigenetic studies. The Illumina HumanMethylation450 BeadChip is widely used to quantify DNA methylation; nevertheless, the design of an appropriate analysis pipeline faces severe challenges due to the convolution of biological and technical variability and the presence of a signal bias between Infinium I and II probe design types. Despite recent attempts to investigate how to analyze DNA methylation data with such an array design, it has not been possible to perform a comprehensive comparison between different bioinformatics pipelines due to the lack of appropriate data sets having both large sample size and sufficient number of technical replicates. Here we perform such a comparative analysis, targeting the problems of reducing the technical variability, eliminating the probe design bias and reducing the batch effect by exploiting two unpublished data sets, which included technical replicates and were profiled for DNA methylation either on peripheral blood, monocytes or muscle biopsies. We evaluated the performance of different analysis pipelines and demonstrated that: (1) it is critical to correct for the probe design type, since the amplitude of the measured methylation change depends on the underlying chemistry; (2) the effect of different normalization schemes is mixed, and the most effective method in our hands were quantile normalization and Beta Mixture Quantile dilation (BMIQ); (3) it is beneficial to correct for batch effects. In conclusion, our comparative analysis using a comprehensive data set suggests an efficient pipeline for proper identification of differentially methylated CpGs using the Illumina 450K arrays.

A systematic assessment of normalization approaches for the Infinium 450K methylation platform

“…Collectively, the results based on examining the correlation and 99th QAD favor the use of BMIQ which reflects previous conclusions drawn by others. 16 However, we find the SWAN procedure to also compare favorablyalthough SWAN does not perform as well on Type I, the Type I probes are already highly reproducible and SWAN does very well in increasing correlation across duplicates in Type II probes. Remarkably, the raw, un-normalized data are already highly reproducible and the improvements offered by using BMIQ and SWAN are modest.…”

“…In downstream analyses, the difference in probe behavior can lead to differential representation of Type I and Type II probes among the top results and can adversely affect the rankings of the individual CpGs. Thus, we examined the degree of bias due to probe design type using the approach of Marabita et al 16 to consider the behavior of pairs of adjacent Type I and Type II probes that are within 200 bp of each other. Nearby probes should behave similarly, irrespective of design type such that similarity is reflective of decreased bias.…”

“…Only recently has some systematic comparisons of preprocessing procedures been conducted. 16 To further fill this critical gap in the literature, we compare the relative performance of four important normalization approaches on reproducibility using real data. Specifically, we examine the Illumina normalization and preprocessing method implemented in Illumina's GenomeStudio software (Illumina Inc.), the subset-quantile within array normalization (SWAN) method, 12 the β-mixture quantile normalization 13 (BMIQ) method, and the complete pipeline (CP) for preprocessing implemented by Touleimat and Tost.…”

Altered DNA methylation of glycolytic and lipogenic genes in liver of obese and type 2 diabetic patients

“…Samples from 17 of the subjects were subjected to both the transcriptome and methylome analysis. Methylome data were filtered for detection and analyzed according to the Beta Mixture Quantile dilation (BMIQ) method [29]. Employing a multigroup comparison, 5834 CpG loci were identified to be differentially methylated among the three groups (FDR < 0.25).…”

Epigenetic changes as prognostic predictors in endometrial carcinomas

“…The combination of quantile normalization together with BMIQ has been suggested as the most effective normalization strategy when dealing with Illumina HumanMethylation450 BeadChip data. 26 Genetic variations may affect probe hybridization; therefore, probe filtering was applied, according to Nordlund et al 27 When a CpG site was annotated to more than one gene, it was used in the average calculation for all present genes. A CpG locus was considered differentially methylated if the Db-value (between low-risk and high-risk tumors) was j0.3j and the P-value < 0.05.…”