A comparison of feature selection and classification methods in DNA methylation studies using the Illumina Infinium platform

Methods in DNA methylation array dataset analysis: A review

Understanding the intricate relationships between gene expression levels and epigenetic modifications in a genome is crucial to comprehending the pathogenic mechanisms of many diseases. With the advancement of DNA Methylome Profiling techniques, the emphasis on identifying Differentially Methylated Regions (DMRs/DMGs) has become crucial for biomarker discovery, offering new insights into the etiology of illnesses. This review surveys the current state of computational tools/algorithms for the analysis of microarray-based DNA methylation profiling datasets, focusing on key concepts underlying the diagnostic/prognostic CpG site extraction. It addresses methodological frameworks, algorithms, and pipelines employed by various authors, serving as a roadmap to address challenges and understand changing trends in the methodologies for analyzing array-based DNA methylation profiling datasets derived from diseased genomes. Additionally, it highlights the importance of integrating gene expression and methylation datasets for accurate biomarker identification, explores prognostic prediction models, and discusses molecular subtyping for disease classification. The review also emphasizes the contributions of machine learning, neural networks, and data mining to enhance diagnostic workflow development, thereby improving accuracy, precision, and robustness.

Epigenome-wide DNA methylation profiling in comparison between pathological and physiological hypertrophy of human cardiomyocytes

Background: Physiological and pathological stimuli result in distinct forms of cardiac hypertrophy, but the molecular regulation comparing the two, especially at the DNA methylation level, is not well understood. Methods: We conducted an in vitro study using human cardiomyocytes exposed to angiotensin II (AngII) and insulin-like growth factor 1 (IGF-1) to mimic pathologically and physiologically hypertrophic heart models, respectively. Whole genome DNA methylation patterns were profiled by the Infinium human MethylationEPIC platform with >850 K DNA methylation loci. Two external datasets were used for comparisons and qRT-PCR was performed for examining expression of associated genes of those identified DNA methylation loci. Results: We detected 194 loci that are significantly differentially methylated after AngII treatment, and 206 significant loci after IGF-1 treatment. Mapping the significant loci to genes, we identified 158 genes corresponding to AngII treatment and 175 genes to IGF-1 treatment. Using the gene-set enrichment analysis, the PI3K-Akt signaling pathway was identified to be significantly enriched for both AngII and IGF-1 treatment. The Hippo signaling pathway was enriched after IGF-1 treatment, but not for AngII treatment. CDK6 and RPTOR are components of the PI3K-Akt pathway but have different DNA methylation patterns in response to AngII and IGF-1. qRT-PCR confirmed the different gene expressions of CDK6 and PRTOR. Conclusion: Our study is pioneering in profiling epigenome DNA methylation changes in adult human cardiomyocytes under distinct stress conditions: pathological (AngII) and physiological (IGF-1). The identified DNA methylation loci, genes, and pathways might have the potential to distinguish between pathological and physiological cardiac hypertrophy.

Machine learning in the identification of prognostic DNA methylation biomarkers among patients with cancer: A systematic review of epigenome-wide studies

BACKGROUND

DNA methylation biomarkers have great potential in improving prognostic classification systems for patients with cancer. Machine learning (ML)-based analytic techniques might help overcome the challenges of analyzing high-dimensional data in relatively small sample sizes. This systematic review summarizes the current use of ML-based methods in epigenome-wide studies for the identification of DNA methylation signatures associated with cancer prognosis.

METHODS

We searched three electronic databases including PubMed, EMBASE, and Web of Science for articles published until 2 January 2023. ML-based methods and workflows used to identify DNA methylation signatures associated with cancer prognosis were extracted and summarized. Two authors independently assessed the methodological quality of included studies by a seven-item checklist adapted from 'A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies (PROBAST)' and from the 'Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK). Different ML methods and workflows used in included studies were summarized and visualized by a sunburst chart, a bubble chart, and Sankey diagrams, respectively.

RESULTS

Eighty-three studies were included in this review. Three major types of ML-based workflows were identified. 1) unsupervised clustering, 2) supervised feature selection, and 3) deep learning-based feature transformation. For the three workflows, the most frequently used ML techniques were consensus clustering, least absolute shrinkage and selection operator (LASSO), and autoencoder, respectively. The systematic review revealed that the performance of these approaches has not been adequately evaluated yet and that methodological and reporting flaws were common in the identified studies using ML techniques.

CONCLUSIONS

There is great heterogeneity in ML-based methodological strategies used by epigenome-wide studies to identify DNA methylation markers associated with cancer prognosis. In theory, most existing workflows could not handle the high multi-collinearity and potentially non-linearity interactions in epigenome-wide DNA methylation data. Benchmarking studies are needed to compare the relative performance of various approaches for specific cancer types. Adherence to relevant methodological and reporting guidelines are urgently needed.

A meta-analysis of immune-cell fractions at high resolution reveals novel associations with common phenotypes and health outcomes

BACKGROUND

Changes in cell-type composition of tissues are associated with a wide range of diseases and environmental risk factors and may be causally implicated in disease development and progression. However, these shifts in cell-type fractions are often of a low magnitude, or involve similar cell subtypes, making their reliable identification challenging. DNA methylation profiling in a tissue like blood is a promising approach to discover shifts in cell-type abundance, yet studies have only been performed at a relatively low cellular resolution and in isolation, limiting their power to detect shifts in tissue composition.

METHODS

Here we derive a DNA methylation reference matrix for 12 immune-cell types in human blood and extensively validate it with flow-cytometric count data and in whole-genome bisulfite sequencing data of sorted cells. Using this reference matrix, we perform a directional Stouffer and fixed effects meta-analysis comprising 23,053 blood samples from 22 different cohorts, to comprehensively map associations between the 12 immune-cell fractions and common phenotypes. In a separate cohort of 4386 blood samples, we assess associations between immune-cell fractions and health outcomes.

RESULTS

Our meta-analysis reveals many associations of cell-type fractions with age, sex, smoking and obesity, many of which we validate with single-cell RNA sequencing. We discover that naïve and regulatory T-cell subsets are higher in women compared to men, while the reverse is true for monocyte, natural killer, basophil, and eosinophil fractions. Decreased natural killer counts associated with smoking, obesity, and stress levels, while an increased count correlates with exercise and sleep. Analysis of health outcomes revealed that increased naïve CD4 + T-cell and N-cell fractions associated with a reduced risk of all-cause mortality independently of all major epidemiological risk factors and baseline co-morbidity. A machine learning predictor built only with immune-cell fractions achieved a C-index value for all-cause mortality of 0.69 (95%CI 0.67-0.72), which increased to 0.83 (0.80-0.86) upon inclusion of epidemiological risk factors and baseline co-morbidity.

CONCLUSIONS

This work contributes an extensively validated high-resolution DNAm reference matrix for blood, which is made freely available, and uses it to generate a comprehensive map of associations between immune-cell fractions and common phenotypes, including health outcomes.

Benchmarking omics-based prediction of asthma development in children

BACKGROUND

Asthma is a heterogeneous disease with high morbidity. Advancement in high-throughput multi-omics approaches has enabled the collection of molecular assessments at different layers, providing a complementary perspective of complex diseases. Numerous computational methods have been developed for the omics-based patient classification or disease outcome prediction. Yet, a systematic benchmarking of those methods using various combinations of omics data for the prediction of asthma development is still lacking.

OBJECTIVE

We aimed to investigate the computational methods in disease status prediction using multi-omics data.

METHOD

We systematically benchmarked 18 computational methods using all the 63 combinations of six omics data (GWAS, miRNA, mRNA, microbiome, metabolome, DNA methylation) collected in The Vitamin D Antenatal Asthma Reduction Trial (VDAART) cohort. We evaluated each method using standard performance metrics for each of the 63 omics combinations.

RESULTS

Our results indicate that overall Logistic Regression, Multi-Layer Perceptron, and MOGONET display superior performance, and the combination of transcriptional, genomic and microbiome data achieves the best prediction. Moreover, we find that including the clinical data can further improve the prediction performance for some but not all the omics combinations.

CONCLUSIONS

Specific omics combinations can reach the optimal prediction of asthma development in children. And certain computational methods showed superior performance than other methods.

A computational solution for bolstering reliability of epigenetic clocks: Implications for clinical trials and longitudinal tracking

Epigenetic clocks are widely used aging biomarkers calculated from DNA methylation data, but this data can be surprisingly unreliable. Here we show technical noise produces deviations up to 9 years between replicates for six prominent epigenetic clocks, limiting their utility. We present a computational solution to bolster reliability, calculating principal components from CpG-level data as input for biological age prediction. Our retrained principal-component versions of six clocks show agreement between most replicates within 1.5 years, improved detection of clock associations and intervention effects, and reliable longitudinal trajectories in vivo and in vitro. This method entails only one additional step compared to traditional clocks, requires no replicates or prior knowledge of CpG reliabilities for training, and can be applied to any existing or future epigenetic biomarker. The high reliability of principal component-based clocks is critical for applications to personalized medicine, longitudinal tracking, in vitro studies, and clinical trials of aging interventions.

Inference of age-associated transcription factor regulatory activity changes in single cells

Transcription factors (TFs) control cell identity and function. How their activity is altered during healthy aging is critical for an improved understanding of aging and disease risk, yet relatively little is known about such changes at cell-type resolution. Here we present and validate a TF activity estimation method for single cells from the hematopoietic system that is based on TF regulons, and apply it to a mouse single-cell RNA-sequencing atlas, to infer age-associated differentiation activity changes in the immune cells of different organs. This revealed an age-associated signature of macrophage dedifferentiation, which is shared across tissue types, and aggravated in tumor-associated macrophages. By extending the analysis to all major cell types, we reveal cell-type and tissue-type-independent age-associated alterations to regulatory factors controlling antigen processing, inflammation, collagen processing and circadian rhythm, that are implicated in age-related diseases. Finally, our study highlights the limitations of using TF expression to infer age-associated changes, underscoring the need to use regulatory activity inference methods.

Novel epigenetic network biomarkers for early detection of esophageal cancer

BACKGROUND

Early detection of esophageal cancer is critical to improve survival. Whilst studies have identified biomarkers, their interpretation and validity is often confounded by cell-type heterogeneity.

RESULTS

Here we applied systems-epigenomic and cell-type deconvolution algorithms to a discovery set encompassing RNA-Seq and DNA methylation data from esophageal adenocarcinoma (EAC) patients and matched normal-adjacent tissue, in order to identify robust biomarkers, free from the confounding effect posed by cell-type heterogeneity. We identify 12 gene-modules that are epigenetically deregulated in EAC, and are able to validate all 12 modules in 4 independent EAC cohorts. We demonstrate that the epigenetic deregulation is present in the epithelial compartment of EAC-tissue. Using single-cell RNA-Seq data we show that one of these modules, a proto-cadherin module centered around CTNND2, is inactivated in Barrett's Esophagus, a precursor lesion to EAC. By measuring DNA methylation in saliva from EAC cases and controls, we identify a chemokine module centered around CCL20, whose methylation patterns in saliva correlate with EAC status.

CONCLUSIONS

Given our observations that a CCL20 chemokine network is overactivated in EAC tissue and saliva from EAC patients, and that in independent studies CCL20 has been found to be overactivated in EAC tissue infected with the bacterium F. nucleatum, a bacterium that normally inhabits the oral cavity, our results highlight the possibility of using DNAm measurements in saliva as a proxy for changes occurring in the esophageal epithelium. Both the CTNND2/CCL20 modules represent novel promising network biomarkers for EAC that merit further investigation.

CpG Island Methylator Phenotype-A Hope for the Future or a Road to Nowhere?

The CpG island methylator phenotype (CIMP) can be regarded as the most notable emanation of epigenetic instability in cancer. Since its discovery in the late 1990s, CIMP has been extensively studied, mainly in colorectal cancers (CRC) and gliomas. Consequently, knowledge on molecular and pathological characteristics of CIMP in CRC and other tumour types has rapidly expanded. Concordant and widespread hypermethylation of multiple CpG islands observed in CIMP in multiple cancers raised hopes for future epigenetically based diagnostics and treatments of solid tumours. However, studies on CIMP in solid tumours were hampered by a lack of generalisability and reproducibility of epigenetic markers. Moreover, CIMP was not a satisfactory marker in predicting clinical outcomes. The idea of targeting epigenetic abnormalities such as CIMP for cancer therapy has not been implemented for solid tumours, either. Twenty-one years after its discovery, we aim to cover both the fundamental and new aspects of CIMP and its future application as a diagnostic marker and target in anticancer therapies.

Predicting Chronological Age from DNA Methylation Data: A Machine Learning Approach for Small Datasets and Limited Predictors

Recent research studies using epigenetic data have been exploring whether it is possible to estimate how old someone is using only their DNA. This application stems from the strong correlation that has been observed in humans between the methylation status of certain DNA loci and chronological age. While genome-wide methylation sequencing has been the most prominent approach in epigenetics research, recent studies have shown that targeted sequencing of a limited number of loci can be successfully used for the estimation of chronological age from DNA samples, even when using small datasets. Following this shift, the need to investigate further into the appropriate statistics behind the predictive models used for DNA methylation-based prediction has been identified in multiple studies. This chapter will look into an example of basic data manipulation and modeling that can be applied to small DNA methylation datasets (100-400 samples) produced through targeted methylation sequencing for a small number of predictors (10-25 methylation sites). Data manipulation will focus on converting the obtained methylation values for the different predictors to a statistically meaningful dataset, followed by a basic introduction into importing such datasets in R, as well as randomizing and splitting into appropriate training and test sets for modeling. Finally, a basic introduction to R modeling will be outlined, starting with feature selection algorithms and continuing with a simple modeling example (linear model) as well as a more complex algorithm (Support Vector Machine).

Combining current knowledge on DNA methylation-based age estimation towards the development of a superior forensic DNA intelligence tool

The estimation of chronological age from biological fluids has been an important quest for forensic scientists worldwide, with recent approaches exploiting the variability of DNA methylation patterns with age in order to develop the next generation of forensic 'DNA intelligence' tools for this application. Drawing from the conclusions of previous work utilising massively parallel sequencing (MPS) for this analysis, this work introduces a DNA methylation-based age estimation method for blood that exhibits the best combination of prediction accuracy and sensitivity reported to date. Statistical evaluation of markers from 51 studies using microarray data from over 4000 individuals, followed by validation using in-house generated MPS data, revealed a final set of 11 markers with the greatest potential for accurate age estimation from minimal DNA material. Utilising an algorithm based on support vector machines, the proposed model achieved an average error (MAE) of 3.3 years, with this level of accuracy retained down to 5 ng of starting DNA input (~ 1 ng PCR input). The accuracy of the model was retained (MAE = 3.8 years) in a separate test set of 88 samples of Spanish origin, while predictions for donors of greater forensic interest (< 55 years of age) displayed even higher accuracy (MAE = 2.6 years). Finally, no sex-related bias was observed for this model, while there were also no signs of variation observed between control and disease-associated populations for schizophrenia, rheumatoid arthritis, frontal temporal dementia and progressive supranuclear palsy in microarray data relating to the 11 markers.

Identification of inflammation-related DNA methylation biomarkers in periodontitis patients based on weighted co-expression analysis

Evidence from past research has shown that DNA methylation plays a key role in the pathogenesis of periodontitis, regulating gene expression levels and thereby affecting the occurrence of various diseases. Three sample sets of methylation data and gene expression data were downloaded from Gene Expression Omnibus (GEO) database. A diagnostic classifier is established based on gene expression data and CpG methylation data. Abnormal expression of immune-related pathways and methyltransferase-related genes in patients with periodontitis was detected. A total of 8,029 differentially expressed CpG (DMP) was annotated to the promoter region of 4,940 genes, of which 295 immune genes were significantly enriched. The CpG sites of 23 differentially co-expressed immune gene promoter regions were identified, and 13 CpG were generally hypermethylated in healthy group samples, while some were methylated in most patients. Five CpGs were screened as robust periodontitis biomarkers. The accuracy in the training data set, the two external verification data sets, and in the transcriptome was 95.5%, 80% and 78.3%, and 82.6%, respectively. This study provided new features for the diagnosis of periodontitis, and contributed to the personalized treatment of periodontitis.

Filtering High-Dimensional Methylation Marks With Extremely Small Sample Size: An Application to Gastric Cancer Data

DNA methylations in critical regions are highly involved in cancer pathogenesis and drug response. However, to identify causal methylations out of a large number of potential polymorphic DNA methylation sites is challenging. This high-dimensional data brings two obstacles: first, many established statistical models are not scalable to so many features; second, multiple-test and overfitting become serious. To this end, a method to quickly filter candidate sites to narrow down targets for downstream analyses is urgently needed. BACkPAy is a pre-screening Bayesian approach to detect biological meaningful patterns of potential differential methylation levels with small sample size. BACkPAy prioritizes potentially important biomarkers by the Bayesian false discovery rate (FDR) approach. It filters non-informative sites (i.e., non-differential) with flat methylation pattern levels across experimental conditions. In this work, we applied BACkPAy to a genome-wide methylation dataset with three tissue types and each type contains three gastric cancer samples. We also applied LIMMA (Linear Models for Microarray and RNA-Seq Data) to compare its results with what we achieved by BACkPAy. Then, Cox proportional hazards regression models were utilized to visualize prognostics significant markers with The Cancer Genome Atlas (TCGA) data for survival analysis. Using BACkPAy, we identified eight biological meaningful patterns/groups of differential probes from the DNA methylation dataset. Using TCGA data, we also identified five prognostic genes (i.e., predictive to the progression of gastric cancer) that contain some differential methylation probes, whereas no significant results was identified using the Benjamin-Hochberg FDR in LIMMA. We showed the importance of using BACkPAy for the analysis of DNA methylation data with extremely small sample size in gastric cancer. We revealed that RDH13, CLDN11, TMTC1, UCHL1, and FOXP2 can serve as predictive biomarkers for gastric cancer treatment and the promoter methylation level of these five genes in serum could have prognostic and diagnostic functions in gastric cancer patients.

DNA Methylation Change Profiling of Colorectal Disease: Screening towards Clinical Use

Colon cancer remains one of the leading causes of cancer-related deaths worldwide. Transformation of colon epithelial cells into invasive adenocarcinomas has been well known to be due to the accumulation of multiple genetic and epigenetic changes. In the past decade, the etiology of inflammatory bowel disease (IBD) which is characterized by chronic inflammation of the intestinal mucosa, was only partially explained by genetic studies providing susceptibility loci, but recently epigenetic studies have provided critical evidences affecting IBD pathogenesis. Over the past decade, A deep understanding of epigenetics along with technological advances have led to identifying numerous genes that are regulated by promoter DNA hypermethylation in colorectal diseases. Recent advances in our understanding of the role of DNA methylation in colorectal diseases could improve a multitude of powerful DNA methylation-based biomarkers, particularly for use as diagnosis, prognosis, and prediction for therapeutic approaches. This review focuses on the emerging potential for translational research of epigenetic alterations into clinical utility as molecular biomarkers. Moreover, this review discusses recent progress regarding the identification of unknown hypermethylated genes in colon cancers and IBD, as well as their possible role in clinical practice, which will have important clinical significance, particularly in the era of the personalized medicine.

Nonlinear ridge regression improves cell-type-specific differential expression analysis

BACKGROUND

Epigenome-wide association studies (EWAS) and differential gene expression analyses are generally performed on tissue samples, which consist of multiple cell types. Cell-type-specific effects of a trait, such as disease, on the omics expression are of interest but difficult or costly to measure experimentally. By measuring omics data for the bulk tissue, cell type composition of a sample can be inferred statistically. Subsequently, cell-type-specific effects are estimated by linear regression that includes terms representing the interaction between the cell type proportions and the trait. This approach involves two issues, scaling and multicollinearity.

RESULTS

First, although cell composition is analyzed in linear scale, differential methylation/expression is analyzed suitably in the logit/log scale. To simultaneously analyze two scales, we applied nonlinear regression. Second, we show that the interaction terms are highly collinear, which is obstructive to ordinary regression. To cope with the multicollinearity, we applied ridge regularization. In simulated data, nonlinear ridge regression attained well-balanced sensitivity, specificity and precision. Marginal model attained the lowest precision and highest sensitivity and was the only algorithm to detect weak signal in real data.

CONCLUSION

Nonlinear ridge regression performed cell-type-specific association test on bulk omics data with well-balanced performance. The omicwas package for R implements nonlinear ridge regression for cell-type-specific EWAS, differential gene expression and QTL analyses. The software is freely available from https://github.com/fumi-github/omicwas.

Whole genome methylation and transcriptome analyses to identify risk for cerebral palsy (CP) in extremely low gestational age neonates (ELGAN)

Preterm birth remains the leading identifiable risk factor for cerebral palsy (CP), a devastating form of motor impairment due to developmental brain injury occurring around the time of birth. We performed genome wide methylation and whole transcriptome analyses to elucidate the early pathogenesis of CP in extremely low gestational age neonates (ELGANs). We evaluated peripheral blood cell specimens collected during a randomized trial of erythropoietin for neuroprotection in the ELGAN (PENUT Trial, NCT# 01378273). DNA methylation data were generated from 94 PENUT subjects (n = 47 CP vs. n = 47 Control) on day 1 and 14 of life. Gene expression data were generated from a subset of 56 subjects. Only one differentially methylated region was identified for the day 1 to 14 change between CP versus no CP, without evidence for differential gene expression of the associated gene RNA Pseudouridine Synthase Domain Containing 2. iPathwayGuide meta-analyses identified a relevant upregulation of JAK1 expression in the setting of decreased methylation that was observed in control subjects but not CP subjects. Evaluation of whole transcriptome data identified several top pathways of potential clinical relevance including thermogenesis, ferroptossis, ribosomal activity and other neurodegenerative conditions that differentiated CP from controls.

A methodological pipeline to generate an epigenetic marker of prenatal exposure to air pollution indicators

A barrier in the children's environmental health field has been the lack of early-warning systems to identify risks of childhood illness and developmental disorders. We aimed to develop a methodology to identify an accessible biomarker measured in a small amount of blood to distinguish newborns at elevated risk from a toxic prenatal exposure, using air pollutants as a case study. Because air pollutants are associated with altered DNA methylation, we developed a pipeline using DNA methylation signatures measured in umbilical cord blood, which could be used as predictors of prenatal exposure. We used air pollution indicators, including modelled trimester-specific and pregnancy average NO₂ and PM_2.5, and DNA methylation signatures from Illumina arrays measured in two New York City-based longitudinal birth cohorts from the Columbia Centre for Children's Environmental Health. We developed a screening plus three-part pipeline that incorporates selection, testing, and validation to identify whether DNA methylation can be used to predict exposure to prenatal air pollution indicators, NO₂ and PM_2.5. Applying this pipeline, we found that cord blood DNA methylation could be used to predict high vs. low average pregnancy NO₂ (AUC = 0.60, 95% CI: 0.52-0.68, with validation AUC = 0.60). Similar results were found for high vs. low third trimester NO₂. In this proof of concept study using air pollutants as an example, we provide an approach (with a generalizable analytic pipeline) that can be used for prediction of prenatal exposure to contaminants. This approach has potential to identify children at risk of developmental disorders and illness resulting from prenatal exposure.

Identification, Heritability, and Relation With Gene Expression of Novel DNA Methylation Loci for Blood Pressure

We conducted an epigenome-wide association study meta-analysis on blood pressure (BP) in 4820 individuals of European and African ancestry aged 14 to 69. Genome-wide DNA methylation data from peripheral leukocytes were obtained using the Infinium Human Methylation 450k BeadChip. The epigenome-wide association study meta-analysis identified 39 BP-related CpG sites with P<1×10-5. In silico replication in the CHARGE consortium of 17 010 individuals validated 16 of these CpG sites. Out of the 16 CpG sites, 13 showed novel association with BP. Conversely, out of the 126 CpG sites identified as being associated (P<1×10-7) with BP in the CHARGE consortium, 21 were replicated in the current study. Methylation levels of all the 34 CpG sites that were cross-validated by the current study and the CHARGE consortium were heritable and 6 showed association with gene expression. Furthermore, 9 CpG sites also showed association with BP with P<0.05 and consistent direction of the effect in the meta-analysis of the Finnish Twin Cohort (199 twin pairs and 4 singletons; 61% monozygous) and the Netherlands Twin Register (266 twin pairs and 62 singletons; 84% monozygous). Bivariate quantitative genetic modeling of the twin data showed that a majority of the phenotypic correlations between methylation levels of these CpG sites and BP could be explained by shared unique environmental rather than genetic factors, with 100% of the correlations of systolic BP with cg19693031 (TXNIP) and cg00716257 (JDP2) determined by environmental effects acting on both systolic BP and methylation levels.

Induction and recovery of CpG site specific methylation changes in human bronchial cells after long-term exposure to carbon nanotubes and asbestos

INTRODUCTION

Inhalation of asbestos induces lung cancer via different cellular mechanisms. Together with the increased production of carbon nanotubes (CNTs) grows the concern about adverse effects on the lungs given the similarities with asbestos. While it has been established that CNT and asbestos induce epigenetic alterations, it is currently not known whether alterations at epigenetic level remain stable after withdrawal of the exposure. Identification of DNA methylation changes after a low dose of CNT and asbestos exposure and recovery can be useful to determine the fibre/particle toxicity and adverse outcome.

METHODS

Human bronchial epithelial cells (16HBE) were treated with a low and non-cytotoxic dose (0.25 µg/ml) of multi-walled carbon nanotubes (MWCNTs-NM400) or single-walled carbon nanotubes (SWCNTs-SRM2483) and 0.05 µg/ml amosite (brown) asbestos for the course of four weeks (sub-chronic exposure). After this treatment, the cells were further incubated (without particle/fibre) for two weeks, allowing recovery from the exposure (recovery period). Nuclear depositions of the CNTs were assessed using femtosecond pulsed laser microscopy in a label-free manner. DNA methylation alterations were analysed using microarrays that assess more than 850 thousand CpG sites in the whole genome.

RESULTS

At non-cytotoxic doses, CNTs were noted to be incorporated with in the nucleus after a four weeks period. Exposure to MWCNTs induced a single hypomethylation at a CpG site and gene promoter region. No change in DNA methylation was observed after the recovery period for MWCNTs. Exposure to SWCNTs or amosite induced hypermethylation at CpG sites after sub-chronic exposure which may involve in 'transcription factor activity' and 'sequence-specific DNA binding' gene ontologies. After the recovery period, hypermethylation and hypomethylation were noted for both SWCNTs and amosite. Hippocalcinlike 1 (HPCAL1), protease serine 3 (PRSS3), kallikrein-related peptidase 3 (KLK3), kruppel like factor 3 (KLF3) genes were hypermethylated at different time points in either SWCNT-exposed or amosite-exposed cells.

CONCLUSION

These results suggest that the specific SWCNT (SRM2483) and amosite fibres studied induce hypo- or hypermethylation on CpG sites in DNA after very low-dose exposure and recovery period. This effect was not seen for the studied MWCNT (NM400).

The effects of bariatric surgery on clinical profile, DNA methylation, and ageing in severely obese patients

BACKGROUND

Severe obesity is a growing, worldwide burden and conventional therapies including radical change of diet and/or increased physical activity have limited results. Bariatric surgery has been proposed as an alternative therapy showing promising results. It leads to substantial weight loss and improvement of comorbidities such as type 2 diabetes. Increased adiposity is associated with changes in epigenetic profile, including DNA methylation. We investigated the effect of bariatric surgery on clinical profile, DNA methylation, and biological age estimated using Horvath's epigenetic clock.

RESULTS

To determine the impact of bariatric surgery and subsequent weight loss on clinical traits, a cohort of 40 severely obese individuals (BMI = 30-73 kg/m²) was examined at the time of surgery and at three follow-up visits, i.e., 3, 6, and 12 months after surgery. The majority of the individuals were women (65%) and the mean age at surgery was 45.1 ± 8.1 years. We observed a significant decrease over time in BMI, fasting glucose, HbA1c, HOMA-IR, insulin, total cholesterol, triglycerides, LDL and free fatty acids levels, and a significant small increase in HDL levels (all p values < 0.05). Epigenome-wide association analysis revealed 4857 differentially methylated CpG sites 12 months after surgery (at Bonferroni-corrected p value < 1.09 × 10^-7). Including BMI change in the model decreased the number of significantly differentially methylated CpG sites by 51%. Gene set enrichment analysis identified overrepresentation of multiple processes including regulation of transcription, RNA metabolic, and biosynthetic processes in the cell. Bariatric surgery in severely obese patients resulted in a decrease in both biological age and epigenetic age acceleration (EAA) (mean = - 0.92, p value = 0.039).

CONCLUSIONS

Our study shows that bariatric surgery leads to substantial BMI decrease and improvement of clinical outcomes observed 12 months after surgery. These changes explained part of the association between bariatric surgery and DNA methylation. We also observed a small, but significant improvement of biological age. These epigenetic changes may be modifiable by environmental lifestyle factors and could be used as potential biomarkers for obesity and in the future for obesity related comorbidities.

Machine learning workflows to estimate class probabilities for precision cancer diagnostics on DNA methylation microarray data

DNA methylation data-based precision cancer diagnostics is emerging as the state of the art for molecular tumor classification. Standards for choosing statistical methods with regard to well-calibrated probability estimates for these typically highly multiclass classification tasks are still lacking. To support this choice, we evaluated well-established machine learning (ML) classifiers including random forests (RFs), elastic net (ELNET), support vector machines (SVMs) and boosted trees in combination with post-processing algorithms and developed ML workflows that allow for unbiased class probability (CP) estimation. Calibrators included ridge-penalized multinomial logistic regression (MR) and Platt scaling by fitting logistic regression (LR) and Firth's penalized LR. We compared these workflows on a recently published brain tumor 450k DNA methylation cohort of 2,801 samples with 91 diagnostic categories using a 5 × 5-fold nested cross-validation scheme and demonstrated their generalizability on external data from The Cancer Genome Atlas. ELNET was the top stand-alone classifier with the best calibration profiles. The best overall two-stage workflow was MR-calibrated SVM with linear kernels closely followed by ridge-calibrated tuned RF. For calibration, MR was the most effective regardless of the primary classifier. The protocols developed as a result of these comparisons provide valuable guidance on choosing ML workflows and their tuning to generate well-calibrated CP estimates for precision diagnostics using DNA methylation data. Computation times vary depending on the ML algorithm from <15 min to 5 d using multi-core desktop PCs. Detailed scripts in the open-source R language are freely available on GitHub, targeting users with intermediate experience in bioinformatics and statistics and using R with Bioconductor extensions.

Identification of diagnostic biomarkers in patients with gestational diabetes mellitus based on transcriptome gene expression and methylation correlation analysis

BACKGROUND

Gestational diabetes mellitus (GDM) has a high prevalence in the period of pregnancy. However, the lack of gold standards in current screening and diagnostic methods posed the biggest limitation. Regulation of gene expression caused by DNA methylation plays an important role in metabolic diseases. In this study, we aimed to screen GDM diagnostic markers, and establish a diagnostic model for predicting GDM.

METHODS

First, we acquired data of DNA methylation and gene expression in GDM samples (N = 41) and normal samples (N = 41) from the Gene Expression Omnibus (GEO) database. After pre-processing the data, linear models were used to identify differentially expressed genes (DEGs). Then we performed pathway enrichment analysis to extract relationships among genes from pathways, construct pathway networks, and further analyzed the relationship between gene expression and methylation of promoter regions. We screened for genes which are significantly negatively correlated with methylation and established mRNA-mRNA-CpGs network. The network topology was further analyzed to screen hub genes which were recognized as robust GDM biomarkers. Finally, the samples were randomly divided into training set (N = 28) and internal verification set (N = 27), and the support vector machine (SVM) ten-fold cross-validation method was used to establish a diagnostic classifier, which verified on internal and external data sets.

RESULTS

In this study, we identified 465 significant DEGs. Functional enrichment analysis revealed that these genes were associated with Type I diabetes mellitus and immunization. And we constructed an interactional network including 1091 genes by using the regulatory relationships of all 30 enriched pathways. 184 epigenetics regulated genes were screened by analyzing the relationship between gene expression and promoter regions' methylation in the network. Moreover, the accuracy rate in the training data set was increased up to 96.3, and 82.1% in the internal validation set, and 97.3% in external validation data sets after establishing diagnostic classifiers which were performed by analyzing the gene expression profiles of obtained 10 hub genes from this network, combined with SVM.

CONCLUSIONS

This study provided new features for the diagnosis of GDM and may contribute to the diagnosis and personalized treatment of GDM.

TOAST: improving reference-free cell composition estimation by cross-cell type differential analysis

In the analysis of high-throughput data from complex samples, cell composition is an important factor that needs to be accounted for. Except for a limited number of tissues with known pure cell type profiles, a majority of genomics and epigenetics data relies on the "reference-free deconvolution" methods to estimate cell composition. We develop a novel computational method to improve reference-free deconvolution, which iteratively searches for cell type-specific features and performs composition estimation. Simulation studies and applications to six real datasets including both DNA methylation and gene expression data demonstrate favorable performance of the proposed method. TOAST is available at https://bioconductor.org/packages/TOAST .

Sensitive Periods for the Effect of Childhood Adversity on DNA Methylation: Results From a Prospective, Longitudinal Study

BACKGROUND

Exposure to early-life adversity is known to predict DNA methylation (DNAm) patterns that may be related to psychiatric risk. However, few studies have investigated whether adversity has time-dependent effects based on the age at exposure.

METHODS

Using a two-stage structured life course modeling approach, we tested the hypothesis that there are sensitive periods when adversity induces greater DNAm changes. We tested this hypothesis in relation to two alternatives: an accumulation hypothesis, in which the effect of adversity increases with the number of occasions exposed, regardless of timing; and a recency model, in which the effect of adversity is stronger for more proximal events. Data came from the Accessible Resource for Integrated Epigenomic Studies, a subsample of mother-child pairs from the Avon Longitudinal Study of Parents and Children (n = 691-774).

RESULTS

After covariate adjustment and multiple testing correction, we identified 38 CpG sites that were differentially methylated at 7 years of age following exposure to adversity. Most loci (n = 35) were predicted by the timing of adversity, namely exposures before 3 years of age. Neither the accumulation nor recency of the adversity explained considerable variability in DNAm. A standard epigenome-wide association study of lifetime exposure (vs. no exposure) failed to detect these associations.

CONCLUSIONS

The developmental timing of adversity explains more variability in DNAm than the accumulation or recency of exposure. Very early childhood appears to be a sensitive period when exposure to adversity predicts differential DNAm patterns. Classification of individuals as exposed versus unexposed to early-life adversity may dilute observed effects.

Statistical method evaluation for differentially methylated CpGs in base resolution next-generation DNA sequencing data

High-throughput bisulfite methylation sequencing such as reduced representation bisulfite sequencing (RRBS), Agilent SureSelect Human Methyl-Seq (Methyl-seq) or whole-genome bisulfite sequencing is commonly used for base resolution methylome research. These data are represented either by the ratio of methylated cytosine versus total coverage at a CpG site or numbers of methylated and unmethylated cytosines. Multiple statistical methods can be used to detect differentially methylated CpGs (DMCs) between conditions, and these methods are often the base for the next step of differentially methylated region identification. The ratio data have a flexibility of fitting to many linear models, but the raw count data take consideration of coverage information. There is an array of options in each datatype for DMC detection; however, it is not clear which is an optimal statistical method. In this study, we systematically evaluated four statistic methods on methylation ratio data and four methods on count-based data and compared their performances with regard to type I error control, sensitivity and specificity of DMC detection and computational resource demands using real RRBS data along with simulation. Our results show that the ratio-based tests are generally more conservative (less sensitive) than the count-based tests. However, some count-based methods have high false-positive rates and should be avoided. The beta-binomial model gives a good balance between sensitivity and specificity and is preferred method. Selection of methods in different settings, signal versus noise and sample size estimation are also discussed.

DNA methylation signatures as biomarkers of prior environmental exposures

Purpose of review

This review demonstrates the growing body of evidence connecting DNA methylation to prior exposure. It highlights the potential to use DNA methylation patterns as a feasible, stable, and accurate biomarker of past exposure, opening new opportunities for environmental and gene-environment interaction studies among existing banked samples.

Recent findings

We present the evidence for association between past exposure, including prenatal exposures, and DNA methylation measured at a later time in the life course. We demonstrate the potential utility of DNA methylation-based biomarkers of past exposure using results from multiple studies of smoking as an example. Multiple studies show the ability to accurately predict prenatal smoking exposure based on DNA methylation measured at birth, in childhood, and even adulthood. Separate sets of DNA methylation loci have been used to predict past personal smoking exposure (postnatal) as well. Further, it appears that these two types of exposures, prenatal and previous personal exposure, can be isolated from each other. There is also a suggestion that quantitative methylation scores may be useful for estimating dose. We highlight the remaining needs for rigor in methylation biomarker development including analytic challenges as well as the need for development across multiple developmental windows, multiple tissue types, and multiple ancestries.

Summary

If fully developed, DNA methylation-based biomarkers can dramatically shift our ability to carry out environmental and genetic-environmental epidemiology using existing biobanks, opening up unprecedented opportunities for environmental health.

Methylation-Based Classification of Cervical Squamous Cell Carcinoma into Two New Subclasses Differing in Immune-Related Gene Expression

Cervical cancer is traditionally classified into two major histological subtypes, cervical squamous cell carcinoma (CSCC) and cervical adenocarcinoma (CA). However, heterogeneity exists among patients, comprising possible subpopulations with distinct molecular profiles. We applied consensus clustering to 307 methylation samples with cervical cancer from The Cancer Genome Atlas (TCGA). Fisher’s exact test was used to perform transcription factors (TFs) and genomic region enrichment. Gene expression profiles were downloaded from TCGA to assess expression differences. Immune cell fraction was calculated to quantify the immune cells infiltration. Putative neo-epitopes were predicted from somatic mutations. Three subclasses were identified: Class 1 correlating with the CA subtype and Classes 2 and 3 dividing the CSCC subtype into two subclasses. We found the hypomethylated probes in Class 3 exhibited strong enrichment in promoter region as compared with Class 2. Five TFs significantly enriched in the hypomethylated promoters and their highly expressed target genes in Class 3 functionally involved in the immune pathway. Gene function analysis revealed that immune-related genes were significantly increased in Class 3, and a higher level of immune cell infiltration was estimated. High expression of 24 immune genes exhibited a better overall survival and correlated with neo-epitope burden. Additionally, we found only two immune-related driver genes, CARD11 and JAK3, to be significantly increased in Class 3. Our analyses provide a classification of the largest CSCC subtype into two new subclasses, revealing they harbored differences in immune-related gene expression.

Epigenetic control of gene regulation during development and disease: A view from the retina

Complex biological processes, such as organogenesis and homeostasis, are stringently regulated by genetic programs that are fine-tuned by epigenetic factors to establish cell fates and/or to respond to the microenvironment. Gene regulatory networks that guide cell differentiation and function are modulated and stabilized by modifications to DNA, RNA and proteins. In this review, we focus on two key epigenetic changes - DNA methylation and histone modifications - and discuss their contribution to retinal development, aging and disease, especially in the context of age-related macular degeneration (AMD) and diabetic retinopathy. We highlight less-studied roles of DNA methylation and provide the RNA expression profiles of epigenetic enzymes in human and mouse retina in comparison to other tissues. We also review computational tools and emergent technologies to profile, analyze and integrate epigenetic information. We suggest implementation of editing tools and single-cell technologies to trace and perturb the epigenome for delineating its role in transcriptional regulation. Finally, we present our thoughts on exciting avenues for exploring epigenome in retinal metabolism, disease modeling, and regeneration.

Global and gene-specific DNA methylation effects of different asbestos fibres on human bronchial epithelial cells

Inhalation exposure to asbestos is associated with lung and pleural diseases in humans and remains a major public health issue worldwide. Human bronchial epithelial cells (16HBE) were exposed to UICC amosite, crocidolite and chrysotile. Cytotoxicity, genotoxicity, global DNA methylation on cytosine residues (using LC-MS/MS) were investigated at different doses (2.5-100 μg/ml). Gene-specific DNA methylation alterations at the whole genome were investigated using a microarray that interrogates >450 thousand CpG sites. Subsequently, gene functional analyses (KEGG pathway, Gene Ontology and functional classification) were performed on genes with differentially methylated gene promoters. At non-cytotoxic doses, global DNA methylation was altered after 24 h exposure to amosite and crocidolite (>2.5 μg/ml). Exposure to amosite and crocidolite (amphibole type asbestos) induced both hypomethylation and hypermethylation at single CpG site and gene promoter levels whereas exposure to chrysotile (serpentine type asbestos) induced hypomethylation at the gene promoter level. Gene functional classification analyses revealed that all types of asbestos fibres induce alterations on GO-clusters i.e. on regulation of Rho-protein signal transduction, nucleus, (e.g. homeobox genes), ATP-binding function and extracellular region (e.g. WNT-group of genes). These differentially methylated genes might contribute to asbestos-related diseases in bronchial cells.

A framework for analyzing DNA methylation data from Illumina Infinium HumanMethylation450 BeadChip

Background

DNA methylation has been identified to be widely associated to complex diseases. Among biological platforms to profile DNA methylation in human, the Illumina Infinium HumanMethylation450 BeadChip (450K) has been accepted as one of the most efficient technologies. However, challenges exist in analysis of DNA methylation data generated by this technology due to widespread biases.

Results

Here we proposed a generalized framework for evaluating data analysis methods for Illumina 450K array. This framework considers the following steps towards a successful analysis: importing data, quality control, within-array normalization, correcting type bias, detecting differentially methylated probes or regions and biological interpretation.

Conclusions

We evaluated five methods using three real datasets, and proposed outperform methods for the Illumina 450K array data analysis. Minfi and methylumi are optimal choice when analyzing small dataset. BMIQ and RCP are proper to correcting type bias and the normalized result of them can be used to discover DMPs. R package missMethyl is suitable for GO term enrichment analysis and biological interpretation.

pETM: a penalized Exponential Tilt Model for analysis of correlated high-dimensional DNA methylation data

Motivation

DNA methylation plays an important role in many biological processes and cancer progression. Recent studies have found that there are also differences in methylation variations in different groups other than differences in methylation means. Several methods have been developed that consider both mean and variance signals in order to improve statistical power of detecting differentially methylated loci. Moreover, as methylation levels of neighboring CpG sites are known to be strongly correlated, methods that incorporate correlations have also been developed. We previously developed a network-based penalized logistic regression for correlated methylation data, but only focusing on mean signals. We have also developed a generalized exponential tilt model that captures both mean and variance signals but only examining one CpG site at a time.

Results

In this article, we proposed a penalized Exponential Tilt Model (pETM) using network-based regularization that captures both mean and variance signals in DNA methylation data and takes into account the correlations among nearby CpG sites. By combining the strength of the two models we previously developed, we demonstrated the superior power and better performance of the pETM method through simulations and the applications to the 450K DNA methylation array data of the four breast invasive carcinoma cancer subtypes from The Cancer Genome Atlas (TCGA) project. The developed pETM method identifies many cancer-related methylation loci that were missed by our previously developed method that considers correlations among nearby methylation loci but not variance signals.

Availability and Implementation

The R package 'pETM' is publicly available through CRAN: http://cran.r-project.org .

Contact

sw2206@columbia.edu.

Supplementary information

Supplementary data are available at Bioinformatics online.

Epigenetic regulation of placental gene expression in transcriptional subtypes of preeclampsia

Background

Preeclampsia (PE) is a heterogeneous, hypertensive disorder of pregnancy, with no robust biomarkers or effective treatments. We hypothesized that this heterogeneity is due to the existence of multiple subtypes of PE and, in support of this hypothesis, we recently identified five clusters of placentas within a large gene expression microarray dataset (N = 330), of which four (clusters 1, 2, 3, and 5) contained a substantial number of PE samples. However, while transcriptional analysis of placentas can subtype patients, we propose that the addition of epigenetic information could discern gene regulatory mechanisms behind the distinct PE pathologies, as well as identify clinically useful potential biomarkers.

Results

We subjected 48 of our samples from transcriptional clusters 1, 2, 3, and 5 to Infinium HumanMethylation450 arrays. Samples belonging to transcriptional clusters 1–3 still showed visible relationships to each other by methylation, but cluster 5, with known chromosomal abnormalities, no longer formed a cohesive group. Within transcriptional clusters 2 and 3, controlling for fetal sex and gestational age in the identification of differentially methylated sites, compared to the healthier cluster 1, dramatically reduced the number of significant sites, but increased the percentage that demonstrated a strong linear correlation with gene expression (from 5% and 2% to 9% and 8%, respectively). Locations exhibiting a positive relationship between methylation and gene expression were most frequently found in CpG open sea enhancer regions within the gene body, while those with a significant negative correlation were often annotated to the promoter in a CpG shore region. Integrated transcriptome and epigenome analysis revealed modifications in TGF-beta signaling, cell adhesion, oxidative phosphorylation, and metabolism pathways in cluster 2 placentas, and aberrations in antigen presentation, allograft rejection, and cytokine-cytokine receptor interaction in cluster 3 samples.

Conclusions

Overall, we have established DNA methylation alterations underlying a portion of the transcriptional development of “canonical” PE in cluster 2 and “immunological” PE in cluster 3. However, a significant number of the observed methylation changes were not associated with corresponding changes in gene expression, and vice versa, indicating that alternate methods of gene regulation will need to be explored to fully comprehend these PE subtypes.

Electronic supplementary material

The online version of this article (10.1186/s13148-018-0463-6) contains supplementary material, which is available to authorized users.

DNA methylome analysis reveals distinct epigenetic patterns of ascending aortic dissection and bicuspid aortic valve

Aims

Epigenetics may mediate the effects of environmental risk factors on disease, including heart disease. Thus, measuring the DNA methylome offers the opportunity to identify novel disease biomarkers and novel insights into disease mechanisms. The DNA methylation landscape of ascending aortic dissection (AD) and bicuspid aortic valve (BAV) with aortic aneurysmal dilatation remain uncharacterized. The present study aimed to explore the genome-wide DNA methylation landscape underpinning these two diseases.

Methods and results

We used Illumina 450k DNA methylation beadarrays to analyse 21 ascending aorta samples, including 10 cases with AD, 5 with BAV and 6 healthy controls. We adjusted for intra-sample cellular heterogeneity, providing the first unbiased genome-wide exploration of the DNA methylation landscape underpinning these two diseases. We discover that both diseases are characterized by loss of DNA methylation at non-CpG sites. We validate this non-CpG hypomethylation signature with pyrosequencing. In contrast to non-CpGs, AD and BAV exhibit distinct DNA methylation landscapes at CpG sites, with BAV characterized mainly by hypermethylation of EZH2 targets. In the case of AD, integrative DNA methylation gene expression analysis reveals that AD is characterized by a dedifferentiated smooth muscle cell phenotype. Our integrative analysis further reveals hypomethylation associated overexpression of RARA in AD, a pattern which is also seen in cells exposed to smoke toxins.

Conclusion

Our data supports a model in which increased cellular proliferation in AD and BAV underpins loss of methylation at non-CpG sites. Our data further supports a model, in which AD is associated with an inflammatory vascular remodeling process, possibly mediated by the epigenome and linked to environmental risk factors such as smoking.

Occupational exposure to pesticides is associated with differential DNA methylation

Objectives

Occupational pesticide exposure is associated with a wide range of diseases, including lung diseases, but it is largely unknown how pesticides influence airway disease pathogenesis. A potential mechanism might be through epigenetic mechanisms, like DNA methylation. Therefore, we assessed associations between occupational exposure to pesticides and genome-wide DNA methylation sites.

Methods

1561 subjects of LifeLines were included with either no (n=1392), low (n=108) or high (n=61) exposure to any type of pesticides (estimated based on current or last held job). Blood DNA methylation levels were measured using Illumina 450K arrays. Associations between pesticide exposure and 420 938 methylation sites (CpGs) were assessed using robust linear regression adjusted for appropriate confounders. In addition, we performed genome-wide stratified and interaction analyses by gender, smoking and airway obstruction status, and assessed associations between gene expression and methylation for genome-wide significant CpGs (n=2802).

Results

In total for all analyses, high pesticide exposure was genome-wide significantly (false discovery rate P<0.05) associated with differential DNA methylation of 31 CpGs annotated to 29 genes. Twenty of these CpGs were found in subjects with airway obstruction. Several of the identified genes, for example, RYR1, ALLC, PTPRN2, LRRC3B, PAX2 and VTRNA2-1, are genes previously linked to either pesticide exposure or lung-related diseases. Seven out of 31 CpGs were associated with gene expression levels.

Conclusions

We show for the first time that occupational exposure to pesticides is genome-wide associated with differential DNA methylation. Further research should reveal whether this differential methylation plays a role in the airway disease pathogenesis induced by pesticides.

MethCNA: a database for integrating genomic and epigenomic data in human cancer

Background

The integration of DNA methylation and copy number alteration data promises to provide valuable insight into the underlying molecular mechanisms responsible for cancer initiation and progression. However, the generation and processing of these datasets are costly and time-consuming if carried out separately. The Illumina Infinium HumanMethylation450 BeadChip, initially designed for the evaluation of DNA methylation levels, allows copy number variant calling using bioinformatics tools.

Results

A substantial amount of Infinium HumanMethylation450 data across various cancer types has been accumulated in recent years and is a valuable resource for large-scale data analysis. Here we present MethCNA, a comprehensive database for genomic and epigenomic data integration in human cancer. In the current release, MethCNA contains about 10,000 tumor samples representing 37 cancer types. All raw array data were collected from The Cancer Genome Atlas and NCBI Gene Expression Omnibus database and analyzed using a pipeline that integrated multiple computational resources and tools. The normalized copy number aberration data and DNA methylation alterations were obtained. We provide a user-friendly web-interface for data mining and visualization.

Conclusions

The Illumina Infinium HumanMethylation450 BeadChip enables the interrogation and integration of both genomic and epigenomic data from exactly the same DNA specimen, and thus can aid in distinguishing driver from passenger mutations in cancer. We expect MethCNA will enable researchers to explore DNA methylation and copy number alteration patterns, identify key oncogenic drivers in cancer, and assist in the development of targeted therapies. MethCNA is publicly available online at http://cgma.scu.edu.cn/MethCNA.

Electronic supplementary material

The online version of this article (10.1186/s12864-018-4525-0) contains supplementary material, which is available to authorized users.

Differences in MWCNT- and SWCNT-induced DNA methylation alterations in association with the nuclear deposition

BACKGROUND

Subtle DNA methylation alterations mediated by carbon nanotubes (CNTs) exposure might contribute to pathogenesis and disease susceptibility. It is known that both multi-walled carbon nanotubes (MWCNTs) and single-walled carbon nanotubes (SWCNTs) interact with nucleus. Such, nuclear-CNT interaction may affect the DNA methylation effects. In order to understand the epigenetic toxicity, in particular DNA methylation alterations, of SWCNTs and short MWCNTs, we performed global/genome-wide, gene-specific DNA methylation and RNA-expression analyses after exposing human bronchial epithelial cells (16HBE14o- cell line). In addition, the presence of CNTs on/in the cell nucleus was evaluated in a label-free way using femtosecond pulsed laser microscopy.

RESULTS

Generally, a higher number of SWCNTs, compared to MWCNTs, was deposited at both the cellular and nuclear level after exposure. Nonetheless, both CNT types were in physical contact with the nuclei. While particle type dependency was noticed for the identified genome-wide and gene-specific alterations, no global DNA methylation alteration on 5-methylcytosine (5-mC) sites was observed for both CNTs. After exposure to MWCNTs, 2398 genes were hypomethylated (at gene promoters), and after exposure to SWCNTs, 589 CpG sites (located on 501 genes) were either hypo- (N = 493 CpG sites) or hypermethylated (N = 96 CpG sites). Cells exposed to MWCNTs exhibited a better correlation between gene promoter methylation and gene expression alterations. Differentially methylated and expressed genes induced changes (MWCNTs > SWCNTs) at different cellular pathways, such as p53 signalling, DNA damage repair and cell cycle. On the other hand, SWCNT exposure showed hypermethylation on functionally important genes, such as SKI proto-oncogene (SKI), glutathione S-transferase pi 1 (GTSP1) and shroom family member 2 (SHROOM2) and neurofibromatosis type I (NF1), which the latter is both hypermethylated and downregulated.

CONCLUSION

After exposure to both types of CNTs, epigenetic alterations may contribute to toxic or repair response. Moreover, our results suggest that the observed differences in the epigenetic response depend on particle type and differential CNT-nucleus interactions.

DNA methylation markers associated with type 2 diabetes, fasting glucose and HbA1c levels: a systematic review and replication in a case-control sample of the Lifelines study

AIMS/HYPOTHESIS

Epigenetic mechanisms may play an important role in the aetiology of type 2 diabetes. Recent epigenome-wide association studies (EWASs) identified several DNA methylation markers associated with type 2 diabetes, fasting glucose and HbA_1c levels. Here we present a systematic review of these studies and attempt to replicate the CpG sites (CpGs) with the most significant associations from these EWASs in a case-control sample of the Lifelines study.

METHODS

We performed a systematic literature search in PubMed and EMBASE for EWASs to test the association between DNA methylation and type 2 diabetes and/or glycaemic traits and reviewed the search results. For replication purposes we selected 100 unique CpGs identified in peripheral blood, pancreas, adipose tissue and liver from 15 EWASs, using study-specific Bonferroni-corrected significance thresholds. Methylation data (Illumina 450K array) in whole blood from 100 type 2 diabetic individuals and 100 control individuals from the Lifelines study were available. Multivariate linear models were used to examine the associations of the specific CpGs with type 2 diabetes and glycaemic traits.

RESULTS

From the 52 CpGs identified in blood and selected for replication, 15 CpGs showed nominally significant associations with type 2 diabetes in the Lifelines sample (p < 0.05). The results for five CpGs (in ABCG1, LOXL2, TXNIP, SLC1A5 and SREBF1) remained significant after a stringent multiple-testing correction (changes in methylation from -3% up to 3.6%, p < 0.0009). All associations were directionally consistent with the original EWAS results. None of the selected CpGs from the tissue-specific EWASs were replicated in our methylation data from whole blood. We were also unable to replicate any of the CpGs associated with HbA_1c levels in the healthy control individuals of our sample, while two CpGs (in ABCG1 and CCDC57) for fasting glucose were replicated at a nominal significance level (p < 0.05).

CONCLUSIONS/INTERPRETATION

A number of differentially methylated CpGs reported to be associated with type 2 diabetes in the EWAS literature were replicated in blood and show promise for clinical use as disease biomarkers. However, more prospective studies are needed to support the robustness of these findings.

Considerations for Design and Analysis of DNA Methylation Studies

The number of epigenetic studies is exponentially increasing. There is anticipation that DNA methylation may close gaps in our understanding of disease etiology, and how certain risk factors affect health and disease, but also that it has potential as a biomarker for disease. Human DNA methylation studies require careful considerations for design and analysis including population and tissue selection, population stratification, cell heterogeneity, confounding, temporality, sample size, appropriate statistical analysis, and validation of results. In this chapter, we discuss relevant aspects for the design of DNA methylation studies and delineate essential steps for their analysis. Specifically, we summarize methods used to extricate biologic signals from technical noise, and statistical approaches to capture meaningful variability based on the research hypothesis.

Statistical and integrative system-level analysis of DNA methylation data

Epigenetics plays a key role in cellular development and function. Alterations to the epigenome are thought to capture and mediate the effects of genetic and environmental risk factors on complex disease. Currently, DNA methylation is the only epigenetic mark that can be measured reliably and genome-wide in large numbers of samples. This Review discusses some of the key statistical challenges and algorithms associated with drawing inferences from DNA methylation data, including cell-type heterogeneity, feature selection, reverse causation and system-level analyses that require integration with other data types such as gene expression, genotype, transcription factor binding and other epigenetic information.

Machine learning for epigenetics and future medical applications

Understanding epigenetic processes holds immense promise for medical applications. Advances in Machine Learning (ML) are critical to realize this promise. Previous studies used epigenetic data sets associated with the germline transmission of epigenetic transgenerational inheritance of disease and novel ML approaches to predict genome-wide locations of critical epimutations. A combination of Active Learning (ACL) and Imbalanced Class Learning (ICL) was used to address past problems with ML to develop a more efficient feature selection process and address the imbalance problem in all genomic data sets. The power of this novel ML approach and our ability to predict epigenetic phenomena and associated disease is suggested. The current approach requires extensive computation of features over the genome. A promising new approach is to introduce Deep Learning (DL) for the generation and simultaneous computation of novel genomic features tuned to the classification task. This approach can be used with any genomic or biological data set applied to medicine. The application of molecular epigenetic data in advanced machine learning analysis to medicine is the focus of this review.

DNA Methylation Score as a Biomarker in Newborns for Sustained Maternal Smoking during Pregnancy

BACKGROUND

Maternal smoking during pregnancy, especially when sustained, leads to numerous adverse health outcomes in offspring. Pregnant women disproportionately underreport smoking and smokers tend to have lower follow-up rates to repeat questionnaires. Missing, incomplete, or inaccurate data on presence and duration of smoking in pregnancy impairs identification of novel health effects and limits adjustment for smoking in studies of other pregnancy exposures. An objective biomarker in newborns of maternal smoking during pregnancy would be valuable.

OBJECTIVES

We developed a biomarker of sustained maternal smoking in pregnancy using common DNA methylation platforms.

METHODS

Using a dimension reduction method, we developed and tested a numeric score in newborns to reflect sustained maternal smoking in pregnancy from data on cotinine, a short-term smoking biomarker measured mid-pregnancy, and Illumina450K cord blood DNA methylation from newborns in the Norwegian Mother and Child Cohort Study (MoBa).

RESULTS

This score reliably predicted smoking status in the training set (n = 1,057; accuracy = 96%, sensitivity = 80%, specificity = 98%). Sensitivity (58%) was predictably lower in the much smaller test set (n = 221), but accuracy (91%) and specificity (97%) remained high. Reduced birth weight, a well-known effect of maternal smoking, was as strongly related to the score as to cotinine. A three-site score had lower, but acceptable, performance (accuracy_train = 82%, accuracy_test = 83%).

CONCLUSIONS

Our smoking methylation score represents a promising novel biomarker of sustained maternal smoking during pregnancy easily calculated with Illumina450K or IlluminaEPIC data. It may help identify novel health impacts and improve adjustment for smoking when studying other risk factors with more subtle effects.

Methylome-wide association study of whole blood DNA in the Norfolk Island isolate identifies robust loci associated with age

Epigenetic regulation of various genomic functions, including gene expression, provide mechanisms whereby an organism can dynamically respond to changes in its environment and modify gene expression accordingly. One epigenetic mechanism implicated in human aging and age-related disorders is DNA methylation. Isolated populations such as Norfolk Island (NI) should be advantageous for the identification of epigenetic factors related to aging due to reduced genetic and environmental variation. Here we conducted a methylome-wide association study of age using whole blood DNA in 24 healthy female individuals from the NI genetic isolate (aged 24-47 years). We analysed 450K methylation array data using a machine learning approach (GLMnet) to identify age-associated CpGs. We identified 497 CpG sites, mapping to 422 genes, associated with age, with 11 sites previously associated with age. The strongest associations identified were for a single CpG site in MYOF and an extended region within the promoter of DDO. These hits were validated in curated public data from 2316 blood samples (MARMAL-AID). This study is the first to report robust age associations for MYOF and DDO, both of which have plausible functional roles in aging. This study also illustrates the value of genetic isolates to reveal new associations with epigenome-level data.

Tools and Strategies for Analysis of Genome-Wide and Gene-Specific DNA Methylation Patterns

DNA methylation is a stable epigenetic mechanism that has important roles in the normal function of a cell and therefore also in disease etiology. Accurate measurements of normal and altered DNA methylation patterns are important to understand its role in regulating gene expression and cell phenotype. Remarkable progress has been made over the last decade in developing methodologies to investigate DNA methylation. The availability of next-generation sequencing has enabled the profiling of methylation marks at an unprecedented scale. Several methods that were previously used to profile locus-specific methylation have now been upgraded to a genome-wide scale using high-throughput sequencing or array platforms. However, because there are so many techniques available, researchers are faced with the challenge of assessing the potential merits or limitations of each technique and selecting the appropriate method for their analysis. In this review we discuss the strengths and weaknesses of genome-wide and gene-specific analysis tools for interrogating DNA methylation. We particularly focus on the design and analysis strategies involved. This review will provide a guideline for selecting the appropriate methods and tools for large-scale and locus-specific DNA methylation analysis.

Epigenetic effects of carbon nanotubes in human monocytic cells

Carbon nanotubes (CNTs) are fibrous carbon-based nanomaterials with a potential to cause carcinogenesis in humans. Alterations in DNA methylation on cytosine-phosphate-guanidine (CpG) sites are potential markers of exposure-induced carcinogenesis. This study examined cytotoxicity, genotoxicity and DNA methylation alterations on human monocytic cells (THP-1) after incubation with single-walled CNTs (SWCNTs) and multi-walled CNTs (MWCNTs). Higher cytotoxicity and genotoxicity were observed after incubation with SWCNTs than incubation with MWCNTs. At the selected concentrations (25 and 100 µg/ml), DNA methylation alterations were studied. Liquid chromatography-mass spectrometry (LC-MS/MS) was used to assess global DNA methylation, and Illumina 450K microarrays were used to assess methylation of single CpG sites. Next, we assessed gene promoter-specific methylation levels. We observed no global methylation or hydroxymethylation alterations, but on gene-specific level, distinct clustering of CNT-treated samples were noted. Collectively, CNTs induced gene promoter-specific altered methylation and those 1127 different genes were identified to be hypomethylated. Differentially methylated genes were involved in several signalling cascade pathways, vascular endothelial growth factor and platelet activation pathways. Moreover, possible contribution of the epigenetic alterations to monocyte differentiation and mixed M1/M2 macrophage polarisation were discussed.

Methylation of SOCS3 is inversely associated with metabolic syndrome in an epigenome-wide association study of obesity

Epigenetic mechanisms, including DNA methylation, mediate the interaction between gene and environment and may play an important role in the obesity epidemic. We assessed the relationship between DNA methylation and obesity in peripheral blood mononuclear cells (PBMCs) at 485,000 CpG sites across the genome in family members (8-90 y of age) using a discovery cohort (192 individuals) and a validation cohort (1,052 individuals) of Northern European ancestry. After Bonferroni-correction (P_α=0.05 = 1.31 × 10⁻⁷) for genome-wide significance, we identified 3 loci, cg18181703 (SOCS3), cg04502490 (ZNF771), and cg02988947 (LIMD2), where methylation status was associated with body mass index percentile (BMI%), a clinical index for obesity in children, adolescents, and adults. These sites were also associated with multiple metabolic syndrome (MetS) traits, including central obesity, fat depots, insulin responsiveness, and plasma lipids. The SOCS3 methylation locus was also associated with the clinical definition of MetS. In the validation cohort, SOCS3 methylation status was found to be inversely associated with BMI% (P = 1.75 × 10⁻⁶), waist to height ratio (P = 4.18 × 10⁻⁷), triglycerides (P = 4.01 × 10⁻⁴), and MetS (P = 4.01 × 10⁻⁷), and positively correlated with HDL-c (P = 4.57 × 10⁻⁸). Functional analysis in a sub cohort (333 individuals) demonstrated SOCS3 methylation and gene expression in PBMCs were inversely correlated (P = 2.93 × 10⁻⁴) and expression of SOCS3 was positively correlated with status of MetS (P = 0.012). We conclude that epigenetic modulation of SOCS3, a gene involved in leptin and insulin signaling, may play an important role in obesity and MetS.

Rapid and specific hypomethylation of enhancers in endothelial cells during adaptation to cell culturing

Epigenetics, including DNA methylation, is one way for a cell to respond to the surrounding environment. Traditionally, DNA methylation has been perceived as a quite stable modification; however, lately, there have been reports of a more dynamic CpG methylation that can be affected by, for example, long-term culturing. We recently reported that methylation in the enhancer of the gene encoding the key fibrinolytic enzyme tissue-type plasminogen activator (t-PA) was rapidly erased during cell culturing. In the present study we used sub-culturing of human umbilical vein endothelial cells (HUVECs) as a model of environmental challenge to examine how fast genome-wide methylation changes can arise. To assess genome-wide DNA methylation, the Infinium HumanMethylation450 BeadChip was used on primary, passage 0, and passage 4 HUVECs. Almost 2% of the analyzed sites changed methylation status to passage 4, predominantly displaying hypomethylation. Sites annotated as enhancers were overrepresented among the differentially methylated sites (DMSs). We further showed that half of the corresponding genes concomitantly altered their expression, most of them increasing in expression. Interestingly, the stroke-related gene HDAC9 increased its expression several hundredfold. This study reveals a rapid hypomethylation of CpG sites in enhancer elements during the early stages of cell culturing. As many methods for methylation analysis are biased toward CpG rich promoter regions, we suggest that such methods may not always be appropriate for the study of methylation dynamics. In addition, we found that significant changes in expression arose in genes with enhancer DMSs. HDAC9 displayed the most prominent increase in expression, indicating, for the first time, that dynamic enhancer methylation may be central in regulating this important stroke-associated gene.

A Class-Information-Based Sparse Component Analysis Method to Identify Differentially Expressed Genes on RNA-Seq Data

With the development of deep sequencing technologies, many RNA-Seq data have been generated. Researchers have proposed many methods based on the sparse theory to identify the differentially expressed genes from these data. In order to improve the performance of sparse principal component analysis, in this paper, we propose a novel class-information-based sparse component analysis (CISCA) method which introduces the class information via a total scatter matrix. First, CISCA normalizes the RNA-Seq data by using a Poisson model to obtain their differential sections. Second, the total scatter matrix is gotten by combining the between-class and within-class scatter matrices. Third, we decompose the total scatter matrix by using singular value decomposition and construct a new data matrix by using singular values and left singular vectors. Then, aiming at obtaining sparse components, CISCA decomposes the constructed data matrix by solving an optimization problem with sparse constraints on loading vectors. Finally, the differentially expressed genes are identified by using the sparse loading vectors. The results on simulation and real RNA-Seq data demonstrate that our method is effective and suitable for analyzing these data.

Cancer Biomarkers from Genome-Scale DNA Methylation: Comparison of Evolutionary and Semantic Analysis Methods

DNA methylation profiling exploits microarray technologies, thus yielding a wealth of high-volume data. Here, an intelligent framework is applied, encompassing epidemiological genome-scale DNA methylation data produced from the Illumina’s Infinium Human Methylation 450K Bead Chip platform, in an effort to correlate interesting methylation patterns with cancer predisposition and, in particular, breast cancer and B-cell lymphoma. Feature selection and classification are employed in order to select, from an initial set of ~480,000 methylation measurements at CpG sites, predictive cancer epigenetic biomarkers and assess their classification power for discriminating healthy versus cancer related classes. Feature selection exploits evolutionary algorithms or a graph-theoretic methodology which makes use of the semantics information included in the Gene Ontology (GO) tree. The selected features, corresponding to methylation of CpG sites, attained moderate-to-high classification accuracies when imported to a series of classifiers evaluated by resampling or blindfold validation. The semantics-driven selection revealed sets of CpG sites performing similarly with evolutionary selection in the classification tasks. However, gene enrichment and pathway analysis showed that it additionally provides more descriptive sets of GO terms and KEGG pathways regarding the cancer phenotypes studied here. Results support the expediency of this methodology regarding its application in epidemiological studies.