Comparison of different normalization assumptions for analyses of DNA methylation data from the cancer genome

Data processing choices can affect findings in differential methylation analyses: an investigation using data from the LIMIT RCT

Objective

A wide array of methods exist for processing and analysing DNA methylation data. We aimed to perform a systematic comparison of the behaviour of these methods, using cord blood DNAm from the LIMIT RCT, in relation to detecting hypothesised effects of interest (intervention and pre-pregnancy maternal BMI) as well as effects known to be spurious, and known to be present.

Methods

DNAm data, from 645 cord blood samples analysed using Illumina 450K BeadChip arrays, were normalised using three different methods (with probe filtering undertaken pre- or post- normalisation). Batch effects were handled with a supervised algorithm, an unsupervised algorithm, or adjustment in the analysis model. Analysis was undertaken with and without adjustment for estimated cell type proportions. The effects estimated included intervention and BMI (effects of interest in the original study), infant sex and randomly assigned groups. Data processing and analysis methods were compared in relation to number and identity of differentially methylated probes, rankings of probes by p value and log-fold-change, and distributions of p values and log-fold-change estimates.

Results

There were differences corresponding to each of the processing and analysis choices. Importantly, some combinations of data processing choices resulted in a substantial number of spurious 'significant' findings. We recommend greater emphasis on replication and greater use of sensitivity analyses.

Interaction between ACOT7 and LncRNA NMRAL2P via Methylation Regulates Gastric Cancer Progression

PURPOSE

Gastric cancer (GC) has a very poor prognosis when diagnosed at a late stage. Acyl-CoA thioesterase 7 (ACOT7) is a major isoform of the acyl coenzyme family that catalyzes the hydrolysis of fatty acyl-CoAs into unesterified free fatty acid and coenzyme A. The purpose of this study was to investigate the expression levels of ACOT7 in GC and mechanisms related therewith.

MATERIALS AND METHODS

Screening of systematic biology studies revealed ACOT7 as a key gene in GC, as well as involvement of the long non-coding RNA NMRAL2P in ACOT7 expression. In this study, GC tissues and adjacent tissue samples were obtained from 10 GC patients at the Department of Gastrointestinal Surgery. GES1 and SGC-7901 cells were collected and treated to silence ACOT7 and overexpress NMRAL2P. The expressions of ACOT7 and NMRAL2P were detected by real-time quantitative PCR and Western blot. Additionally, cell proliferation, apoptosis, migration, and invasion were examined.

RESULTS

ACOT7 was upregulated in gastric tumor tissues and GC cell lines. ACOT7 gene silencing induced a less malignant phenotype and was closely correlated to reduced cell proliferation and migration, altered cell cycle, and increased apoptosis. Furthermore, NMRAL2P was downregulated in tumor tissues and GC cell lines. NMRAL2P overexpression induced a more malignant phenotype and significantly inhibited the expression of ACOT7. Importantly, NMRAL2P indirectly methylated ACOT7 by binding to DNMT3b, thereby suppressing ACOT7 expression.

CONCLUSION

NMRAL2P activation suppresses ACOT7 expression in GC. Thus, ACOT7 could be a promising target for the treatment of GC.

A cross-package Bioconductor workflow for analysing methylation array data

Methylation in the human genome is known to be associated with development and disease. The Illumina Infinium methylation arrays are by far the most common way to interrogate methylation across the human genome. This paper provides a Bioconductor workflow using multiple packages for the analysis of methylation array data. Specifically, we demonstrate the steps involved in a typical differential methylation analysis pipeline including: quality control, filtering, normalization, data exploration and statistical testing for probe-wise differential methylation. We further outline other analyses such as differential methylation of regions, differential variability analysis, estimating cell type composition and gene ontology testing. Finally, we provide some examples of how to visualise methylation array data.

A cross-package Bioconductor workflow for analysing methylation array data

Methylation in the human genome is known to be associated with development and disease. The Illumina Infinium methylation arrays are by far the most common way to interrogate methylation across the human genome. This paper provides a Bioconductor workflow using multiple packages for the analysis of methylation array data. Specifically, we demonstrate the steps involved in a typical differential methylation analysis pipeline including: quality control, filtering, normalization, data exploration and statistical testing for probe-wise differential methylation. We further outline other analyses such as differential methylation of regions, differential variability analysis, estimating cell type composition and gene ontology testing. Finally, we provide some examples of how to visualise methylation array data.

A cross-package Bioconductor workflow for analysing methylation array data

Methylation in the human genome is known to be associated with development and disease. The Illumina Infinium methylation arrays are by far the most common way to interrogate methylation across the human genome. This paper provides a Bioconductor workflow using multiple packages for the analysis of methylation array data. Specifically, we demonstrate the steps involved in a typical differential methylation analysis pipeline including: quality control, filtering, normalization, data exploration and statistical testing for probe-wise differential methylation. We further outline other analyses such as differential methylation of regions, differential variability analysis, estimating cell type composition and gene ontology testing. Finally, we provide some examples of how to visualise methylation array data.

Between-array normalization for 450K data

The Illumina Infinium HumanMethylation450 BeadChip is frequently used in epigenetic research. Besides quantile normalization there is currently no standard method to normalize the data between arrays. We describe some properties of the data generated by this platform and present a normalization method based on local regression. We compare the performance of this method with other commonly used approaches in three benchmarks (correlation between 21 pairs of technical replicates, detection of differential methylation and correlation of methylation levels for smoking-associated CpG sites with smoking behavior of 655 participants of an epidemiological study). Results indicate that the proposed method improves reproducibility, whereas some commonly used methods can have adverse effects.

Identification of functionally methylated regions based on discriminant analysis through integrating methylation and gene expression data

DNA methylation is essential not only in cellular differentiation but also in diseases.

Large-scale characterization of DNA methylation changes in human gastric carcinomas with and without metastasis

PURPOSE

Metastasis is the leading cause of death for gastric carcinoma. An epigenetic biomarker panel for predicting gastric carcinoma metastasis could have significant clinical impact on the care of patients with gastric carcinoma. The main purpose of this study is to characterize the methylation differences between gastric carcinomas with and without metastasis.

EXPERIMENTAL DESIGN

Genome-wide DNA methylation profiles between 4 metastatic and 4 nonmetastatic gastric carcinomas and their surgical margins (SM) were analyzed using methylated-CpG island amplification with microarray. The methylation states of 73 candidate genes were further analyzed in patients with gastric carcinoma in a discovery cohort (n=108) using denatured high performance liquid chromatography, bisulfite-sequencing, and MethyLight. The predictive values of potential metastasis-methylation biomarkers were validated in cohorts of patients with gastric carcinoma in China (n=330), Japan (n=129), and Korea (n=153).

RESULTS

The gastric carcinoma genome showed significantly higher proportions of hypomethylation in the promoter and exon-1 regions, as well as increased hypermethylation of intragenic fragments when compared with SMs. Significant differential methylation was validated in the CpG islands of 15 genes (P<0.05) and confirmed using bisulfite sequencing. These genes included BMP3, BNIP3, CDKN2A, ECEL1, ELK1, GFRA1, HOXD10, KCNH1, PSMD10, PTPRT, SIGIRR, SRF, TBX5, TFPI2, and ZNF382. Methylation changes of GFRA1, SRF, and ZNF382 resulted in up- or downregulation of their transcription. Most importantly, the prevalence of GFRA1, SRF, and ZNF382 methylation alterations was consistently and coordinately associated with gastric carcinoma metastasis and the patients' overall survival throughout discovery and validation cohorts in China, Japan, and Korea.

CONCLUSION

Methylation changes of GFRA1, SRF, and ZNF382 may be a potential biomarker set for prediction of gastric carcinoma metastasis.

Deciphering global signal features of high-throughput array data from cancers

Normalization of array data relies on the assumption that most genes are not altered, which means that the signals for different samples should be scaled to have similar median or average values. However, accumulating evidence suggests that gene expression could be widely up-regulated in cancers. Our previous results and subsequent findings have shown that violation of the assumption led to erroneous interpretation of microarray data. To decipher the global signal features of microarray data from cancer samples, we empirically evaluated a large collection of gene and miRNA expression profiles and copy-number variation arrays. Our results showed that, at the transcriptomic level, genes and miRNAs are widely over-expressed in a large proportion of cancers. In contrast, at the genomic level, global raw signal intensities for methylation and copy number variation show negligible differences between cancer and normal samples. These results force us to re-evaluate the proper use of normalization procedures under different experimental conditions and for different array platforms.

Batch effects and pathway analysis: two potential perils in cancer studies involving DNA methylation array analysis

BACKGROUND

DNA methylation microarrays have become an increasingly popular means of studying the role of epigenetics in cancer, although the methods used to analyze these arrays are still being developed and existing methods are not always widely disseminated among microarray users.

METHODS

We investigated two problems likely to confront DNA methylation microarray users: (i) batch effects and (ii) the use of widely available pathway analysis software to analyze results. First, DNA taken from individuals exposed to low and high levels of drinking water arsenic were plated twice on Illumina's Infinium 450 K HumanMethylation Array, once in order of exposure and again following randomization. Second, we conducted simulations in which random CpG sites were drawn from the 450 K array and subjected to pathway analysis using Ingenuity's IPA software.

RESULTS

The majority of differentially methylated CpG sites identified in Run One were due to batch effects; few sites were also identified in Run Two. In addition, the pathway analysis software reported many significant associations between our data, randomly drawn from the 450 K array, and various diseases and biological functions.

CONCLUSIONS

These analyses illustrate the pitfalls of not properly controlling for chip-specific batch effects as well as using pathway analysis software created for gene expression arrays to analyze DNA methylation array data.

IMPACT

We present evidence that (i) chip-specific effects can simulate plausible differential methylation results and (ii) popular pathway analysis software developed for expression arrays can yield spurious results when used in tandem with methylation microarrays.

Analysing and interpreting DNA methylation data

DNA methylation is an epigenetic mark that has suspected regulatory roles in a broad range of biological processes and diseases. The technology is now available for studying DNA methylation genome-wide, at a high resolution and in a large number of samples. This Review discusses relevant concepts, computational methods and software tools for analysing and interpreting DNA methylation data. It focuses not only on the bioinformatic challenges of large epigenome-mapping projects and epigenome-wide association studies but also highlights software tools that make genome-wide DNA methylation mapping more accessible for laboratories with limited bioinformatics experience.