An evaluation of supervised methods for identifying differentially methylated regions in Illumina methylation arrays

Deconstructing the epigenomic architecture of human neurodegeneration

The past 10 years have seen a rapid advance in our ability to profile the epigenome from human pathologic material, opening up new study designs to investigate the role of epigenomic features in human disease. Moderate to large scale studies have now been conducted in the target tissue of neurodegenerative diseases, the brain, and, through the use of rigorous statistical methodologies, have laid a foundation of validated observations and successful study designs that inform our perspective on the role of the epigenome in these diseases, generate new hypotheses, and guide our path forward for a second generation of studies. It is clear that sampling the epigenome is not redundant with other "omic" profiling of the same tissue and that it can serve as an important vehicle for the integration of the effect of multiple environmental exposures on risk of disease. In some cases, change in the epigenome may thus have a causal impact on disease, but we now have evidence that such changes may also mediate some of the effect of tau proteinopathy and that other changes may moderate the impact of genetic risk factors. Thus, the epigenome may be involved at multiple different stages of the sequence of events that leads to human neurodegeneration, and we review the study designs that may begin to guide the development of a more comprehensive perspective on the aging brain's epigenome.

Detecting differentially methylated regions with multiple distinct associations

Aim: We evaluated five methods for detecting differentially methylated regions (DMRs): DMRcate, comb-p, seqlm, GlobalP and dmrff. Materials & methods: We used a simulation study and real data analysis to evaluate performance. Additionally, we evaluated the use of an ancestry-matched reference cohort to estimate correlations between CpG sites in cord blood. Results: Several methods had inflated Type I error, which increased at more stringent significant levels. In power simulations with 1-2 causal CpG sites with the same direction of effect, dmrff was consistently among the most powerful methods. Conclusion: This study illustrates the need for more thorough simulation studies when evaluating novel methods. More work must be done to develop methods with well-controlled Type I error that do not require individual-level data.

Epigenome-wide meta-analysis of DNA methylation differences in prefrontal cortex implicates the immune processes in Alzheimer's disease

DNA methylation differences in Alzheimer's disease (AD) have been reported. Here, we conducted a meta-analysis of more than 1000 prefrontal cortex brain samples to prioritize the most consistent methylation differences in multiple cohorts. Using a uniform analysis pipeline, we identified 3751 CpGs and 119 differentially methylated regions (DMRs) significantly associated with Braak stage. Our analysis identified differentially methylated genes such as MAMSTR, AGAP2, and AZU1. The most significant DMR identified is located on the MAMSTR gene, which encodes a cofactor that stimulates MEF2C. Notably, MEF2C cooperates with another transcription factor, PU.1, a central hub in the AD gene network. Our enrichment analysis highlighted the potential roles of the immune system and polycomb repressive complex 2 in pathological AD. These results may help facilitate future mechanistic and biomarker discovery studies in AD.

A Linear Regression and Deep Learning Approach for Detecting Reliable Genetic Alterations in Cancer Using DNA Methylation and Gene Expression Data

DNA methylation change has been useful for cancer biomarker discovery, classification, and potential treatment development. So far, existing methods use either differentially methylated CpG sites or combined CpG sites, namely differentially methylated regions, that can be mapped to genes. However, such methylation signal mapping has limitations. To address these limitations, in this study, we introduced a combinatorial framework using linear regression, differential expression, deep learning method for accurate biological interpretation of DNA methylation through integrating DNA methylation data and corresponding TCGA gene expression data. We demonstrated it for uterine cervical cancer. First, we pre-filtered outliers from the data set and then determined the predicted gene expression value from the pre-filtered methylation data through linear regression. We identified differentially expressed genes (DEGs) by Empirical Bayes test using Limma. Then we applied a deep learning method, "nnet" to classify the cervical cancer label of those DEGs to determine all classification metrics including accuracy and area under curve (AUC) through 10-fold cross validation. We applied our approach to uterine cervical cancer DNA methylation dataset (NCBI accession ID: GSE30760, 27,578 features covering 63 tumor and 152 matched normal samples). After linear regression and differential expression analysis, we obtained 6287 DEGs with false discovery rate (FDR) <0.001. After performing deep learning analysis, we obtained average classification accuracy 90.69% (±1.97%) of the uterine cervical cancerous labels. This performance is better than that of other peer methods. We performed in-degree and out-degree hub gene network analysis using Cytoscape. We reported five top in-degree genes (PAIP2, GRWD1, VPS4B, CRADD and LLPH) and five top out-degree genes (MRPL35, FAM177A1, STAT4, ASPSCR1 and FABP7). After that, we performed KEGG pathway and Gene Ontology enrichment analysis of DEGs using tool WebGestalt(WEB-based Gene SeT AnaLysis Toolkit). In summary, our proposed framework that integrated linear regression, differential expression, deep learning provides a robust approach to better interpret DNA methylation analysis and gene expression data in disease study.

Large epigenome-wide association study of childhood ADHD identifies peripheral DNA methylation associated with disease and polygenic risk burden

Epigenetic variation in peripheral tissues is being widely studied as a molecular biomarker of complex disease and disease-related exposures. To date, few studies have examined differences in DNA methylation associated with attention-deficit hyperactivity disorder (ADHD). In this study, we profiled genetic and methylomic variation across the genome in saliva samples from children (age 7-12 years) with clinically established ADHD (N = 391) and nonpsychiatric controls (N = 213). We tested for differentially methylated positions (DMPs) associated with both ADHD diagnosis and ADHD polygenic risk score, by using linear regression models including smoking, medication effects, and other potential confounders in our statistical models. Our results support previously reported associations between ADHD and DNA methylation levels at sites annotated to VIPR2, and identify several novel disease-associated DMPs (p < 1e-5), although none of them were genome-wide significant. The two top-ranked, ADHD-associated DMPs (cg17478313 annotated to SLC7A8 and cg21609804 annotated to MARK2) are also significantly associated with nearby SNPs (p = 1.2e-46 and p = 2.07e-59), providing evidence that disease-associated DMPs are under genetic control. We also report a genome-wide significant association between ADHD polygenic risk and variable DNA methylation at a site annotated to the promoter of GART and SON (p = 6.71E-8). Finally, we show that ADHD-associated SNPs colocalize with SNPs associated with methylation levels in saliva. This is the first large-scale study of DNA methylation in children with ADHD. Our results represent novel epigenetic biomarkers for ADHD that may be useful for patient stratification, reinforce the importance of genetic effects on DNA methylation, and provide plausible molecular mechanisms for ADHD risk variants.

coMethDMR: accurate identification of co-methylated and differentially methylated regions in epigenome-wide association studies with continuous phenotypes

Recent technology has made it possible to measure DNA methylation profiles in a cost-effective and comprehensive genome-wide manner using array-based technology for epigenome-wide association studies. However, identifying differentially methylated regions (DMRs) remains a challenging task because of the complexities in DNA methylation data. Supervised methods typically focus on the regions that contain consecutive highly significantly differentially methylated CpGs in the genome, but may lack power for detecting small but consistent changes when few CpGs pass stringent significance threshold after multiple comparison. Unsupervised methods group CpGs based on genomic annotations first and then test them against phenotype, but may lack specificity because the regional boundaries of methylation are often not well defined. We present coMethDMR, a flexible, powerful, and accurate tool for identifying DMRs. Instead of testing all CpGs within a genomic region, coMethDMR carries out an additional step that selects co-methylated sub-regions first. Next, coMethDMR tests association between methylation levels within the sub-region and phenotype via a random coefficient mixed effects model that models both variations between CpG sites within the region and differential methylation simultaneously. coMethDMR offers well-controlled Type I error rate, improved specificity, focused testing of targeted genomic regions, and is available as an open-source R package.

Epigenome-Wide Association Study for All-Cause Mortality in a Cardiovascular Cohort Identifies Differential Methylation in Castor Zinc Finger 1 (CASZ1)

Background DNA methylation is implicated in many chronic diseases and may contribute to mortality. Therefore, we conducted an epigenome-wide association study (EWAS) for all-cause mortality with whole-transcriptome data in a cardiovascular cohort (CATHGEN [Catheterization Genetics]). Methods and Results Cases were participants with mortality≥7 days postcatheterization whereas controls were alive with≥2 years of follow-up. The Illumina Human Methylation 450K and EPIC arrays (Illumina, San Diego, CA) were used for the discovery and validation sets, respectively. A linear model approach with empirical Bayes estimators adjusted for confounders was used to assess difference in methylation (Δβ). In the discovery set (55 cases, 49 controls), 25 629 (6.5%) probes were differently methylated (P<0.05). In the validation set (108 cases, 108 controls), 3 probes were differentially methylated with a false discovery rate-adjusted P<0.10: cg08215811 (SLC4A9; log2 fold change=-0.14); cg17845532 (MATK; fold change=-0.26); and cg17944110 (castor zinc finger 1 [CASZ1]; FC=0.26; P<0.0001; false discovery rate-adjusted P=0.046-0.080). Meta-analysis identified 6 probes (false discovery rate-adjusted P<0.05): the 3 above, cg20428720 (intergenic), cg17647904 (NCOR2), and cg23198793 (CAPN3). Messenger RNA expression of 2 MATK isoforms was lower in cases (fold change=-0.24 [P=0.007] and fold change=-0.61 [P=0.009]). The CASZ1, NCOR2, and CAPN3 transcripts did not show differential expression (P>0.05); the SLC4A9 transcript did not pass quality control. The cg17944110 probe is located within a potential regulatory element; expression of predicted targets (using GeneHancer) of the regulatory element, UBIAD1 (P=0.01) and CLSTN1 (P=0.03), were lower in cases. Conclusions We identified 6 novel methylation sites associated with all-cause mortality. Methylation in CASZ1 may serve as a regulatory element associated with mortality in cardiovascular patients. Larger studies are necessary to confirm these observations.

Serum dioxin and DNA methylation in the sperm of operation ranch hand veterans exposed to Agent Orange

BACKGROUND

Exposure to the herbicide Agent Orange during the Vietnam War was widespread and is associated with numerous adverse health outcomes. A continuing concern of veterans is the possibility that exposure to the dioxin-containing herbicide might induce adverse reproductive outcomes. We sought to assess whether exposure to Agent Orange in Vietnam was associated with changes in DNA methylation in sperm in a subset of Vietnam veterans who participated in the Air Force Health Study (AFHS).

METHODS

We studied 37 members of the AFHS chosen to have no, low, medium or high exposure to Agent Orange, based upon serum dioxin levels obtained during a series of examinations. DNA from stored semen was extracted and DNA methylation assessed on the Illumina 450 K platform.

RESULTS

Initial epigenome-wide analysis returned no loci that survived control for false discovery. However, the TEAD3 gene had four different CpG sites that showed loss of DNA methylation associated with dioxin exposure. Analysis assessing regional DNA methylation changes revealed 36 gene regions, including the region of the imprinted gene H19 to have altered DNA methylation associated with high exposure compared to the low exposure group. Additional comparison of our data with sperm DNA methylation data from Russian boys exposed to dioxin found an additional 5 loci that were altered in both studies and exhibited a consistent direction of association.

CONCLUSIONS

Studying a small number of sperm samples from veterans enrolled in the AFHS, we did not find evidence of significant epigenome-wide alterations associated with exposure to Agent Orange. However, additional analysis showed that the H19 gene region is altered in the sperm of Agent Orange-exposed Ranch Hand veterans. Our study also replicated several findings of a prior study of dioxin-exposed Russian boys. These results provide additional candidate loci for further investigation and may have implications for the reproductive health of dioxin-exposed individuals.

Epigenome-wide association study of depression symptomatology in elderly monozygotic twins

Depression is a severe and debilitating mental disorder diagnosed by evaluation of affective, cognitive and physical depression symptoms. Severity of these symptoms strongly impacts individual's quality of life and is influenced by a combination of genetic and environmental factors. One of the molecular mechanisms allowing for an interplay between these factors is DNA methylation, an epigenetic modification playing a pivotal role in regulation of brain functioning across lifespan. The aim of this study was to investigate if there are DNA methylation signatures associated with depression symptomatology in order to identify molecular mechanisms contributing to pathophysiology of depression. We performed an epigenome-wide association study (EWAS) of continuous depression symptomatology score measured in a cohort of 724 monozygotic Danish twins (346 males, 378 females). Through EWAS analyses adjusted for sex, age, flow-cytometry based blood cell composition, and twin relatedness structure in the data we identified depression symptomatology score to be associated with blood DNA methylation levels in promoter regions of neuropsin (KLK8, p-value = 4.7 × 10^-7) and DAZ associated protein 2 (DAZAP2, p-value = 3.13 × 10^-8) genes. Other top associated probes were located in gene bodies of MAD1L1 (p-value = 5.16 × 10^-6), SLC29A2 (p-value = 6.15 × 10^-6) and AKT1 (p-value = 4.47 × 10^-6), all genes associated before with development of depression. Additionally, the following three measures (a) DNAmAge (calculated with Horvath and Hannum epigenetic clock estimators) adjusted for chronological age, (b) difference between DNAmAge and chronological age, and (c) DNAmAge acceleration were not associated with depression symptomatology score in our cohort. In conclusion, our data suggests that depression symptomatology score is associated with DNA methylation levels of genes implicated in response to stress, depressive-like behaviors, and recurrent depression in patients, but not with global DNA methylation changes across the genome.

Epigenome-wide exploratory study of monozygotic twins suggests differentially methylated regions to associate with hand grip strength

Hand grip strength is a measure of muscular strength and is used to study age-related loss of physical capacity. In order to explore the biological mechanisms that influence hand grip strength variation, an epigenome-wide association study (EWAS) of hand grip strength in 672 middle-aged and elderly monozygotic twins (age 55–90 years) was performed, using both individual and twin pair level analyses, the latter controlling the influence of genetic variation. Moreover, as measurements of hand grip strength performed over 8 years were available in the elderly twins (age 73–90 at intake), a longitudinal EWAS was conducted for this subsample. No genome-wide significant CpG sites or pathways were found, however two of the suggestive top CpG sites were mapped to the COL6A1 and CACNA1B genes, known to be related to muscular dysfunction. By investigating genomic regions using the comb-p algorithm, several differentially methylated regions in regulatory domains were identified as significantly associated to hand grip strength, and pathway analyses of these regions revealed significant pathways related to the immune system, autoimmune disorders, including diabetes type 1 and viral myocarditis, as well as negative regulation of cell differentiation. The genes contributing to the immunological pathways were HLA-B, HLA-C, HLA-DMA, HLA-DPB1, MYH10, ERAP1 and IRF8, while the genes implicated in the negative regulation of cell differentiation were IRF8, CEBPD, ID2 and BRCA1. In conclusion, this exploratory study suggests hand grip strength to associate with differentially methylated regions enriched in immunological and cell differentiation pathways, and hence merits further investigations.