A Comparative Study of Five Association Tests Based on CpG Set for Epigenome-Wide Association Studies

A novel principal component based method for identifying differentially methylated regions in Illumina Infinium MethylationEPIC BeadChip data

Differentially methylated regions (DMRs) are genomic regions with methylation patterns across multiple CpG sites that are associated with a phenotype. In this study, we proposed a Principal Component (PC) based DMR analysis method for use with data generated using the Illumina Infinium MethylationEPIC BeadChip (EPIC) array. We obtained methylation residuals by regressing the M-values of CpGs within a region on covariates, extracted PCs of the residuals, and then combined association information across PCs to obtain regional significance. Simulation-based genome-wide false positive (GFP) rates and true positive rates were estimated under a variety of conditions before determining the final version of our method, which we have named DMRPC. Then, DMRPC and another DMR method, coMethDMR, were used to perform epigenome-wide analyses of several phenotypes known to have multiple associated methylation loci (age, sex, and smoking) in a discovery and a replication cohort. Among regions that were analysed by both methods, DMRPC identified 50% more genome-wide significant age-associated DMRs than coMethDMR. The replication rate for the loci that were identified by only DMRPC was higher than the rate for those that were identified by only coMethDMR (90% for DMRPC vs. 76% for coMethDMR). Furthermore, DMRPC identified replicable associations in regions of moderate between-CpG correlation which are typically not analysed by coMethDMR. For the analyses of sex and smoking, the advantage of DMRPC was less clear. In conclusion, DMRPC is a new powerful DMR discovery tool that retains power in genomic regions with moderate correlation across CpGs.

Genetic and environment effects on structural neuroimaging endophenotype for bipolar disorder: a novel molecular approach

We investigated gene-environment effects on structural brain endophenotype in bipolar disorder (BD) using a novel method of combining polygenic risk scores with epigenetic signatures since traditional methods of examining the family history and trauma effects have significant limitations. The study enrolled 119 subjects, including 55 BD spectrum (BDS) subjects diagnosed with BD or major depressive disorder (MDD) with subthreshold BD symptoms and 64 non-BDS subjects comprising 32 MDD subjects without BD symptoms and 32 healthy subjects. The blood samples underwent genome-wide genotyping and methylation quantification. We derived polygenic risk score (PRS) and methylation profile score (MPS) as weighted summations of risk single nucleotide polymorphisms and methylation probes, respectively, which were considered as molecular measures of genetic and environmental risks for BD. Linear regression was used to relate PRS, MPS, and their interaction to 44 brain structure measures quantified from magnetic resonance imaging (MRI) on 47 BDS subjects, and the results were compared with those based on family history and childhood trauma. After multiplicity corrections using false discovery rate (FDR), MPS was found to be negatively associated with the volume of the medial geniculate thalamus (FDR = 0.059, partial R² = 0.208). Family history, trauma scale, and PRS were not associated with any brain measures. PRS and MPS show significant interactions on whole putamen (FDR = 0.09, partial R² = 0.337). No significant gene-environment interactions were identified for the family history and trauma scale. PRS and MPS generally explained greater proportions of variances of the brain measures (range of partial R² = [0.008, 0.337]) than the clinical risk factors (range = [0.004, 0.228]).

Associations between an integrated component of maternal glycemic regulation in pregnancy and cord blood DNA methylation

Background: Previous studies suggest that fetal programming to hyperglycemia in pregnancy is due to modulation of DNA methylation (DNAm), but they have been limited in their maternal glycemic characterization. Methods: In the Gen3G study, we used a principal component analysis to integrate multiple glucose and insulin values measured during the second trimester oral glucose tolerance test. We investigated associations between principal components and cord blood DNAm levels in an epigenome-wide analysis among 430 mother-child pairs. Results: The first principal component was robustly associated with lower DNAm at cg26974062 (TXNIP; p = 9.9 × 10^-9) in cord blood. TXNIP is a well-known DNAm marker for type 2 diabetes in adults. Conclusion: We hypothesize that abnormal glucose metabolism in pregnancy may program dysregulation of TXNIP across the life course.

PathwayPCA: an R/Bioconductor Package for Pathway Based Integrative Analysis of Multi-Omics Data

The authors present pathwayPCA, an R/Bioconductor package for integrative pathway analysis that utilizes modern statistical methodology, including supervised and adaptive, elastic-net, sparse principal component analysis. pathwayPCA can be applied to continuous, binary, and survival outcomes in studies with multiple covariates and/or interaction effects. It outperforms several alternative methods at identifying disease-associated pathways in integrative analysis using both simulated and real datasets. In addition, several case studies are provided to illustrate pathwayPCA analysis with gene selection, estimating, and visualizing sample-specific pathway activities, identifying sex-specific pathway effects in kidney cancer, and building integrative models for predicting patient prognosis. pathwayPCA is an open-source R package, freely available through the Bioconductor repository. pathwayPCA is expected to be a useful tool for empowering the wider scientific community to analyze and interpret the wealth of available proteomics data, along with other types of molecular data recently made available by Clinical Proteomic Tumor Analysis Consortium and other large consortiums.

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.

Integrative Omics Approach to Identifying Genes Associated With Atrial Fibrillation

Rationale: GWAS (Genome-Wide Association Studies) have identified hundreds of genetic loci associated with atrial fibrillation (AF). However, these loci explain only a small proportion of AF heritability. Objective: To develop an approach to identify additional AF-related genes by integrating multiple omics data. Methods and Results: Three types of omics data were integrated: (1) summary statistics from the AFGen 2017 GWAS; (2) a whole blood EWAS (Epigenome-Wide Association Study) of AF; and (3) a whole blood TWAS (Transcriptome-Wide Association Study) of AF. The variant-level GWAS results were collapsed into gene-level associations using fast set-based association analysis. The CpG-level EWAS results were also collapsed into gene-level associations by an adapted SNP-set Kernel Association Test approach. Both GWAS and EWAS gene-based associations were then meta-analyzed with TWAS using a fixed-effects model weighted by the sample size of each data set. A tissue-specific network was subsequently constructed using the NetWAS (Network-Wide Association Study). The identified genes were then compared with the AFGen 2018 GWAS that contained more than triple the number of AF cases compared with AFGen 2017 GWAS. We observed that the multiomics approach identified many more relevant AF-related genes than using AFGen 2018 GWAS alone (1931 versus 206 genes). Many of these genes are involved in the development and regulation of heart- and muscle-related biological processes. Moreover, the gene set identified by multiomics approach explained much more AF variance than those identified by GWAS alone (10.4% versus 3.5%). Conclusions: We developed a strategy to integrate multiple omics data to identify AF-related genes. Our integrative approach may be useful to improve the power of traditional GWAS, which might be particularly useful for rare traits and diseases with limited sample size.

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

Epigenome-wide association studies (EWASs) have become increasingly popular for studying DNA methylation (DNAm) variations in complex diseases. The Illumina methylation arrays provide an economical, high-throughput and comprehensive platform for measuring methylation status in EWASs. A number of software tools have been developed for identifying disease-associated differentially methylated regions (DMRs) in the epigenome. However, in practice, we found these tools typically had multiple parameter settings that needed to be specified and the performance of the software tools under different parameters was often unclear. To help users better understand and choose optimal parameter settings when using DNAm analysis tools, we conducted a comprehensive evaluation of 4 popular DMR analysis tools under 60 different parameter settings. In addition to evaluating power, precision, area under precision-recall curve, Matthews correlation coefficient, F1 score and type I error rate, we also compared several additional characteristics of the analysis results, including the size of the DMRs, overlap between the methods and execution time. The results showed that none of the software tools performed best under their default parameter settings, and power varied widely when parameters were changed. Overall, the precision of these software tools were good. In contrast, all methods lacked power when effect size was consistent but small. Across all simulation scenarios, comb-p consistently had the best sensitivity as well as good control of false-positive rate.

Comparison of novel and existing methods for detecting differentially methylated regions

BACKGROUND

Single-probe analyses in epigenome-wide association studies (EWAS) have identified associations between DNA methylation and many phenotypes, but do not take into account information from neighboring probes. Methods to detect differentially methylated regions (DMRs) (clusters of neighboring probes associated with a phenotype) may provide more power to detect associations between DNA methylation and diseases or phenotypes of interest.

RESULTS

We proposed a novel approach, GlobalP, and perform comparisons with 3 methods-DMRcate, Bumphunter, and comb-p-to identify DMRs associated with log triglycerides (TGs) in real GAW20 data before and after fenofibrate treatment. We applied these methods to the summary statistics from an EWAS performed on the methylation data. Comb-p, DMRcate, and GlobalP detected very similar DMRs near the gene CPT1A on chromosome 11 in both the pre- and posttreatment data. In addition, GlobalP detected 2 DMRs before fenofibrate treatment in the genes ETV6 and ABCG1. Bumphunter identified several DMRs on chromosomes 1 and 20, which did not overlap with DMRs detected by other methods.

CONCLUSIONS

Our novel method detected the same DMR identified by two existing methods and detected two additional DMRs not identified by any of the existing methods we compared.

Integrative methylation score to identify epigenetic modifications associated with lipid changes resulting from fenofibrate treatment in families

Epigenome-wide association studies (EWAS) have traditionally focused on the association test of single epigenetic markers with complex traits. However, it is possible that multiple cytosine-phosphate-guanine (CpG) sites at the same locus could jointly exert their effects on human traits. Therefore, a region-based test that combines multiple markers could be more powerful. We used 2 different region-based tests to investigate the association between changes in DNA methylation and drug response, including the median methylation level test (MMLT) and sequence kernel association test (SKAT). No genes were found to be significantly associated with the drug response (for triglycerides, the false discovery rate ranged from 0.855 to 0.999; for high-density lipoprotein cholesterol, and the false discovery rate ranged from 0.584 to 0.915). Further evidence is needed to explore potential application of gene-level methylation association analysis.

Application of novel and existing methods to identify genes with evidence of epigenetic association: results from GAW20

BACKGROUND

The rise in popularity and accessibility of DNA methylation data to evaluate epigenetic associations with disease has led to numerous methodological questions. As part of GAW20, our working group of 8 research groups focused on gene searching methods.

RESULTS

Although the methods were varied, we identified 3 main themes within our group. First, many groups tackled the question of how best to use pedigree information in downstream analyses, finding that (a) the use of kinship matrices is common practice, (b) ascertainment corrections may be necessary, and (c) pedigree information may be useful for identifying parent-of-origin effects. Second, many groups also considered multimarker versus single-marker tests. Multimarker tests had modestly improved power versus single-marker methods on simulated data, and on real data identified additional associations that were not identified with single-marker methods, including identification of a gene with a strong biological interpretation. Finally, some of the groups explored methods to combine single-nucleotide polymorphism (SNP) and DNA methylation into a single association analysis.

CONCLUSIONS

A causal inference method showed promise at discovering new mechanisms of SNP activity; gene-based methods of summarizing SNP and DNA methylation data also showed promise. Even though numerous questions still remain in the analysis of DNA methylation data, our discussions at GAW20 suggest some emerging best practices.