Search

Background: Changing cell-type proportions can confound studies of differential gene expression or DNA methylation (DNAm) from peripheral blood mononuclear cells (PBMCs). We examined how cell-type proportions derived from the transcriptome versus the methylome (DNAm) influence estimates of differentially expressed genes (DEGs) and differentially methylated positions (DMPs)., Methods: Transcriptome and DNAm data were obtained from PBMC RNA and DNA of Kenyan children (n = 8) before, during, and 6 weeks following uncomplicated malaria. DEGs and DMPs between time points were detected using cell-type adjusted modeling with Cibersortx or IDOL, respectively., Results: Most major cell types and principal components had moderate to high correlation between the two deconvolution methods (r = 0.60-0.96). Estimates of cell-type proportions and DEGs or DMPs were largely unaffected by the method, with the greatest discrepancy in the estimation of neutrophils., Conclusion: Variation in cell-type proportions is captured similarly by both transcriptomic and methylome deconvolution methods for most major cell types., (© 2024. The Author(s).)

Background: Health outcomes among children born prematurely are known to be sexually dimorphic, with male infants often more affected, yet the mechanism behind this observation is not clear. CpG methylation levels in the placenta and blood also differ by sex and are associated with adverse health outcomes. We contrasted CpG methylation levels in the placenta and neonatal blood (n = 358) from the Extremely Low Gestational Age Newborn (ELGAN) cohort based on the EPIC array, which assays over 850,000 CpG sites across the epigenome. Sex-specific epigenome-wide association analyses were conducted for the placenta and neonatal blood samples independently, and the results were compared to determine tissue-specific differences between the methylation patterns in males and females. All models were adjusted for cell type heterogeneity. Enrichment pathway analysis was performed to identify the biological functions of genes related to the sexually dimorphic CpG sites., Results: Approximately 11,500 CpG sites were differentially methylated in relation to sex. Of these, 5949 were placenta-specific and 5361 were blood-specific, with only 233 CpG sites overlapping in both tissues. For placenta-specific CpG sites, 90% were hypermethylated in males. For blood-specific CpG sites, 95% were hypermethylated in females. In the placenta, keratinocyte differentiation biological pathways were enriched among the differentially methylated genes. No enrichment pathways were observed for blood., Conclusions: Distinct methylation patterns were observed between male and female children born extremely premature, and keratinocyte differentiation pathways were enriched in the placenta. These findings provide new insights into the epigenetic mechanisms underlying sexually dimorphic health outcomes among extremely premature infants., (© 2023. BioMed Central Ltd., part of Springer Nature.)

Background: Children born extremely preterm are at heightened risk for intellectual and social impairment, including Autism Spectrum Disorder (ASD). There is increasing evidence for a key role of the placenta in prenatal developmental programming, suggesting that the placenta may, in part, contribute to origins of neurodevelopmental outcomes., Methods: We examined associations between placental transcriptomic and epigenomic profiles and assessed their ability to predict intellectual and social impairment at age 10 years in 379 children from the Extremely Low Gestational Age Newborn (ELGAN) cohort. Assessment of intellectual ability (IQ) and social function was completed with the Differential Ability Scales-II and Social Responsiveness Scale (SRS), respectively. Examining IQ and SRS allows for studying ASD risk beyond the diagnostic criteria, as IQ and SRS are continuous measures strongly correlated with ASD. Genome-wide mRNA, CpG methylation and miRNA were assayeds with the Illumina Hiseq 2500, HTG EdgeSeq miRNA Whole Transcriptome Assay, and Illumina EPIC/850 K array, respectively. We conducted genome-wide differential analyses of placental mRNA, miRNA, and CpG methylation data. These molecular features were then integrated for a predictive analysis of IQ and SRS outcomes using kernel aggregation regression. We lastly examined associations between ASD and the multi-omic-predicted component of IQ and SRS., Results: Genes with important roles in neurodevelopment and placental tissue organization were associated with intellectual and social impairment. Kernel aggregations of placental multi-omics strongly predicted intellectual and social function, explaining approximately 8% and 12% of variance in SRS and IQ scores via cross-validation, respectively. Predicted in-sample SRS and IQ showed significant positive and negative associations with ASD case-control status., Limitations: The ELGAN cohort comprises children born pre-term, and generalization may be affected by unmeasured confounders associated with low gestational age. We conducted external validation of predictive models, though the sample size (N = 49) and the scope of the available out-sample placental dataset are limited. Further validation of the models is merited., Conclusions: Aggregating information from biomarkers within and among molecular data types improves prediction of complex traits like social and intellectual ability in children born extremely preterm, suggesting that traits within the placenta-brain axis may be omnigenic.

Background: A low level of methylation at cg05575921 in the aryl-hydrocarbon receptor repressor (AHRR) gene is robustly associated with smoking, and some studies have observed associations between cg05575921 methylation and increased lung cancer risk and mortality. To prospectively examine whether decreased methylation at cg05575921 may identify high risk subpopulations for lung cancer screening among heavy smokers, and mortality in cases, we evaluated associations between cg05575921 methylation and lung cancer risk and mortality, by histotype, in heavy smokers., Methods: The β-Carotene and Retinol Efficacy Trial (CARET) included enrollees ages 45-69 with ≥ 20 pack-year smoking histories and/or occupational asbestos exposure. A subset of CARET participants had cg05575921 methylation available from HumanMethylationEPIC assays of blood collected on average 4.3 years prior to lung cancer diagnosis in cases. Cg05575921 methylation β-values were treated continuously for a 10% methylation decrease and as quintiles, where quintile 1 (Q1, referent) represents high methylation and Q5, low methylation. We used conditional logistic regression models to examine lung cancer risk overall and by histotype in a nested case-control study including 316 lung cancer cases (diagnosed through 2005) and 316 lung cancer-free controls matched on age (±5 years), sex, race/ethnicity, enrollment year, current/former smoking, asbestos exposure, and follow-up time. Mortality analyses included 372 lung cancer cases diagnosed between 1985 and 2013 with available methylation data. We used Cox proportional hazards models to examine mortality overall and by histotype., Results: Decreased cg05575921 methylation was strongly associated with smoking, even in our population of heavy smokers. We did not observe associations between decreased pre-diagnosis cg05575921 methylation and increased lung cancer risk, overall or by histotype. We observed linear increasing trends for lung cancer-specific mortality across decreasing cg05575921 methylation quintiles for adenocarcinoma and small cell carcinoma (P-trends = 0.01 and 0.04, respectively)., Conclusions: In our study of heavy smokers, decreased cg05575921 methylation was strongly associated with smoking but not increased lung cancer risk. The observed association between cg05575921 methylation and increased mortality in adenocarcinoma and small cell histotypes requires further examination. Our results do not support using decreased cg05575921 methylation as a biomarker for lung cancer screening risk stratification.

Background: Increasing evidence suggests that prenatal exposure to arsenic, even at common environmental levels, adversely affects child health. These adverse effects include impaired fetal growth, which can carry serious health implications lifelong. However, the mechanisms by which arsenic affects fetal health and development remain unclear., Methods: We addressed this question using a group of 46 pregnant women selected from the New Hampshire Birth Cohort Study (NHBCS), a US cohort exposed to low-to-moderate arsenic levels in drinking water through the use of unregulated private wells. Prenatal arsenic exposure was assessed using maternal urine samples taken at mid-gestation. Samples of the fetal portion of the placenta were taken from the base of the umbilical cord insertion at the time of delivery, stored in RNAlater and frozen. We used RNA sequencing to analyze changes in global gene expression in the fetal placenta associated with in utero arsenic exposure, adjusting for maternal age. Gene set enrichment analysis and enrichment mapping were then used to identify biological processes represented by the differentially expressed genes. Since our previous analyses have identified considerable sex differences in placental gene expression associated with arsenic exposure, we analyzed male and female samples separately., Results: At FDR < 0.05, no genes were differentially expressed in female placenta, while 606 genes were differentially expressed in males. Genes showing the most significant associations with arsenic exposure in females were LEMD1 and UPK3B (fold changes 2.51 and 2.48), and in males, FIBIN and RANBP3L (fold changes 0.14 and 0.15). In gene set enrichment analyses, at FDR < 0.05, a total of 211 gene sets were enriched with differentially expressed genes in female placenta, and 154 in male placenta. In female but not male placenta, 103 of these gene sets were also associated with reduced birth weight., Conclusions: Our results reveal multiple biological functions in the fetal placenta that are potentially affected by increased arsenic exposure, a subset of which is sex-dependent. Further, our data suggest that in female infants, the mechanisms underlying the arsenic-induced reduction of birth weight may involve activation of stress response pathways.

Background: Prenatal exposure to arsenic has been linked to a range of adverse health conditions in later life. Such fetal origins of disease are frequently the result of environmental effects on the epigenome, leading to long-term alterations in gene expression. Several studies have demonstrated effects of prenatal arsenic exposure on DNA methylation; however the impact of arsenic on the generation and decoding of post-translational histone modifications (PTHMs) is less well characterized, and has not been studied in the context of prenatal human exposures., Methods: In the current study, we examined the effect of exposure to low-to-moderate levels of arsenic in a US birth cohort, on the expression of 138 genes encoding key epigenetic regulators in the fetal portion of the placenta. Our candidate genes included readers, writers and erasers of PTHMs, and chromatin remodelers., Results: Arsenic exposure was associated with the expression of 27 of the 138 epigenetic genes analyzed. When the cohort was stratified by fetal sex, arsenic exposure was associated with the expression of 40 genes in male fetal placenta, and only 3 non-overlapping genes in female fetal placenta. In particular, we identified an inverse relationship between arsenic exposure and expression of the gene encoding the histone methyltransferase, PRDM6 (p < 0.001). Mutation of PRDM6 has been linked to the congenital heart defect, patent ductus arteriosus., Conclusions: Our findings suggest that prenatal arsenic exposure may have sex-specific effects on the fetal epigenome, which could plausibly contribute to its subsequent health impacts.

Background: While associated with obesity, the cause of the rapid rise in prevalence of nonalcoholic fatty liver disease (NAFLD) in children, which is highest among Hispanics, is not well understood. Animal experiments have demonstrated that arsenic exposure contributes to liver injury. Our objective was to examine the association between arsenic exposure and NAFLD in humans and to determine if race/ethnicity modifies the association., Methods: Urinary inorganic arsenic concentrations among those ≥12 years in the National Health and Nutrition Examination Survey, 2005-2014 were used to assess the cross-sectional association with serum alanine aminotransferase (ALT) levels, a marker of liver dysfunction. We excluded high alcohol consumers (>4-5 drinks/day; n = 939), positive hepatitis B or C (n = 2330), those missing body mass index (n = 100) and pregnant women (n = 629) for a final sample of 8518. Arsenic was measured using liquid chromatography coupled with mass spectrometry and ALT was measured using standard methods. Sampling weights were used to obtain national estimates. Due to lack of normality, estimates were log transformed and are presented as geometric means. Logistic regression models controlling for age, sex, income, and weight category estimate adjusted odd ratios (aOR) of elevated ALT by quartile of arsenic and tested for effect modification by race/ethnicity and weight. Elevated ALT was defined as >25 IU/L and >22 IU/L for boys and girls ≤17 years, respectively and >30 IU/L and >19 IU/L for men and women, respectively., Results: Among all, aOR of elevated ALT were higher among those in the highest vs. lowest arsenic quartile (referent), 1.4 (95% confidence interval [CI]: 1.1, 1.7) with a borderline significant interaction (p = 0.07) by race/ethnicity but not weight (p = 0.4). In analysis stratified by race/ethnicity, aOR of elevated ALT among those in the 4th quartile were higher among Mexican Americans, 2.0 (CI: 1.3, 3.1) and non-Hispanic whites only, aOR 1.4 (CI: 1.1, 1.8) despite the fact that obesity prevalence was highest among non-Hispanic blacks., Conclusions: Our findings demonstrate a positive association between urinary arsenic exposure and risk of NAFLD among U.S. adolescents and adults that is highest among Mexican Americans and among those obese, regardless of race/ethnicity.

Background: The placenta is the principal organ regulating intrauterine growth and development, performing critical functions on behalf of the developing fetus. The delineation of functional networks and pathways driving placental processes has the potential to provide key insight into intrauterine perturbations that result in adverse birth as well as later life health outcomes., Results: We generated the transcriptome-wide profile of 200 term human placenta using the Illumina HiSeq 2500 platform and characterized the functional placental gene network using weighted gene coexpression network analysis (WGCNA). We identified 17 placental coexpression network modules that were dominated by functional processes including growth, organ development, gas exchange and immune response. Five network modules, enriched for processes including cellular respiration, amino acid transport, hormone signaling, histone modifications and gene expression, were associated with birth weight; hub genes of all five modules (CREB3, DDX3X, DNAJC14, GRHL1 and C21orf91) were significantly associated with fetal growth restriction, and one hub gene (CREB3) was additionally associated with fetal overgrowth., Conclusions: In this largest RNA-Seq based transcriptome-wide profiling study of human term placenta conducted to date, we delineated a placental gene network with functional relevance to fetal growth using a network-based approach with superior scale reduction capacity. Our study findings not only implicate potential molecular mechanisms underlying fetal growth but also provide a reference placenta gene network to inform future studies investigating placental dysfunction as a route to future disease endpoints.

Background: Sex-specific factors play a major role in human health and disease, including responses to environmental stresses such as toxicant exposure. Increasing evidence suggests that such sex differences also exist during fetal development. In a previous report using the resources of the New Hampshire Birth Cohort Study (NHBCS), we found that low-to-moderate in utero exposure to arsenic, a highly toxic and widespread pollutant, was associated with altered expression of several key developmental genes in the fetal portion of the placenta. These associations were sex-dependent, suggesting that in utero arsenic exposure differentially impacts male and female fetuses. In the present study, we investigated the molecular basis for these sex-specific responses to arsenic., Methods: Using NanoString technology, we further analyzed the fetal placenta samples from the NHBCS for the expression of genes encoding arsenic transporters and metabolic enzymes. Multivariable linear regression analysis was used to examine their relationship with arsenic exposure and with key developmental genes, after stratification by fetal sex., Results: We found that maternal arsenic exposure was strongly associated with expression of the AQP9 gene, encoding an aquaglyceroporin transporter, in female but not male fetal placenta. Moreover, AQP9 expression associated with that of a subset of female-specific arsenic-responsive genes., Conclusions: Our results suggest that AQP9 is upregulated in response to arsenic exposure in female, but not male, fetal placenta. Based on these results and prior studies, increased AQP9 expression may lead to increased arsenic transport in the female fetal placenta, which in turn may alter the expression patterns of key developmental genes that we have previously shown to be associated with arsenic exposure. Thus, this study suggests that AQP9 may play a role in the sex-specific effects of in utero arsenic exposure.

Background: In utero arsenic exposure may alter fetal developmental programming by altering DNA methylation, which may result in a higher risk of disease in later life. We evaluated the association between in utero arsenic exposure and DNA methylation (DNAm) in cord blood and its influence in later life., Methods: Genome-wide DNA methylation in cord blood from 64 subjects in the Taiwanese maternal infant and birth cohort was analyzed. Robust regressions were applied to assess the association of DNA methylation with in utero arsenic exposure. Multiple testing was adjusted by controlling false discovery rate (FDR) of 0.05. The DAVID bioinformatics tool was implemented for functional annotation analyses on the detected CpGs. The identified CpGs were further tested in an independent cohort. For the CpGs replicated in the independent cohort, linear mixed models were applied to assess the association of DNA methylation with low-density lipoprotein (LDL) at different ages (2, 5, 8, 11 and 14 years)., Results: In total, 579 out of 385,183 CpGs were identified after adjusting for multiple testing (FDR = 0.05), of which ~60% were positively associated with arsenic exposure. Functional annotation analysis on these CpGs detected 17 KEGG pathways (FDR = 0.05) including pathways for cardiovascular diseases (CVD) and diabetes mellitus. In the independent cohort, about 46% (252 out of 553 CpGs) of the identified CpGs showed associations consistent with those in the study cohort. In total, 11 CpGs replicated in the independent cohort were in the pathways related to CVD and diabetes mellitus. Via longitudinal analyses, we found at 5 out of the 11 CpGs methylation was associated with LDL over time and interactions between DNA methylation and time were observed at 4 of the 5 CpGs, cg25189764 (coeff = 0.157, p-value = 0.047), cg04986899 (coeff. For interaction [coeff.int] = 0.030, p-value = 0.024), cg04903360 (coeff.int = 0.026, p-value = 0.032), cg08198265 (coeff.int = -0.063, p-value = 0.0021), cg10473311 (coeff.int = -0.021, p-value = 0.027)., Conclusion: In utero arsenic exposure was associated with cord blood DNA methylation at various CpGs. The identified CpGs may help determine pathological epigenetic mechanisms linked to in utero arsenic exposure. Five CpGs (cg25189764, cg04986899, cg04903360, cg08198265 and cg10473311) may serve as epigenetic markers for changes in LDL later in life.

Background: Recent interest in reference-free deconvolution of DNA methylation data has led to several supervised methods, but these methods do not easily permit the interpretation of underlying cell types., Results: We propose a simple method for reference-free deconvolution that provides both proportions of putative cell types defined by their underlying methylomes, the number of these constituent cell types, as well as a method for evaluating the extent to which the underlying methylomes reflect specific types of cells. We demonstrate these methods in an analysis of 23 Infinium data sets from 13 distinct data collection efforts; these empirical evaluations show that our algorithm can reasonably estimate the number of constituent types, return cell proportion estimates that demonstrate anticipated associations with underlying phenotypic data; and methylomes that reflect the underlying biology of constituent cell types., Conclusions: Our methodology permits an explicit quantitation of the mediation of phenotypic associations with DNA methylation by cell composition effects. Although more work is needed to investigate functional information related to estimated methylomes, our proposed method provides a novel and useful foundation for conducting DNA methylation studies on heterogeneous tissues lacking reference data.

Background: MicroRNAs have been identified as potential cancer biomarkers due to their presence and stability in many body fluids including urine and plasma, but the relationship of the pattern of expression of these messengers across various biological media has not been addressed and could provide important information in order to translate these biomarkers for epidemiologic or clinical use., Methods: We analyzed microRNA of matched FFPE-tumor tissue, plasma, urine exosomes (n = 16) and WBCs (n = 11) from patients with bladder cancer, using Nanostring miRNA assays and droplet digital PCR for validation. Pearson correlations were used to compare expression between media., Results: Numerous microRNAs were detected and overlapping from specific bio-specimen sources. MiR-4454 and miR-21 overexpression was found in three sources: tumor, WBCs and urine. Additionally, miR-15b-5p, miR-126-3p, miR-93-5p, and miR-150-5p were common to tumor/WBCs, while miR-720/3007a, miR-205, miR-200c-3p and miR-29b-3p common to tumor/urine. Significant associations were noted between the log-adjusted average miRNA counts in tumor vs. WBCs (r = 0.418 p < 0.001), and tumor vs. urine (r = 0.38 p < 0.001). No association was seen tumor vs. plasma exosome miRs (r = 0.07 p = 0.06)., Conclusions: MicroRNA profiling from matched samples in patients shows a significant number of microRNAs up regulated in bladder tumors are identifiable in urine exosomes and WBCs of the same patient, but not in blood plasma. This study demonstrated varying relationships between miRNA detected in biological media from the same patient, and serves to inform the potential of urine-based microRNAs as biomarkers for bladder cancer and potentially other malignancies.

Background: We sought to identify and characterize predictors of poor fetal growth among variables extracted from perinatal medical records to gain insight into potential etiologic mechanisms. In this process we reevaluated a previously observed association between poor fetal growth and maternal psychiatric disease., Methods: We evaluated 449 deliveries of >36 weeks gestation that occurred between 9/2008 and 9/2010 at the Women and Infants Hospital in Providence Rhode Island. This study group was oversampled for Small-for-Gestational-Age (SGA) infants and excluded Large-for-Gestational-Age (LGA) infants. We assessed the associations between recorded clinical variables and impaired fetal growth: SGA or Intrauterine Growth Restriction (IUGR) diagnosis. After validating the previously observed association between maternal psychiatric disease and impaired fetal growth we addressed weaknesses in the prior studies by explicitly considering antidepressant use and the timing of symptoms with respect to pregnancy. We then evaluated DNA methylation levels at 27 candidate loci in placenta from a subset of these deliveries (n = 197) to examine if epigenetic variation could provide insight into the mechanisms that cause this co-morbidity., Results: Infants of mothers with prenatal psychiatric disease (Depression, Anxiety, OCD/Panic) had increased odds of poor fetal growth (ORadjusted = 3.36, 95%CI: 1.38-8.14). This relationship was similar among those who were treated with antidepressants (ORadjusted = 3.69, 95%CI: 1.31-10.45) and among those who were not (ORadjusted = 3.19, 95%CI: 1.30-7.83). Among those with a history of psychiatric disease but no active disease in pregnancy the ORadjusted was 0.45 (95%CI: 0.09-2.35). A locus near the transcription start site of the leptin receptor (cg21655790) had methylation levels that were decreased in the presence of: 1) SGA/IUGR, and 2) active but not resolved psychiatric disease (among mothers not on antidepressants)., Conclusions: These results validate and further characterize the association between maternal psychiatric disease and poor fetal growth. Because the association appears to depend on active psychiatric disease, this suggests a transient and potentially modifiable pathophysiology. The molecular findings in this study suggest that altered leptin signaling may be involved in the biological mechanisms that link prenatal maternal psychiatric symptoms and poor fetal growth.

Background: The impact of cell-composition effects in analysis of DNA methylation data is now widely appreciated. With the availability of a reference data set consisting of DNA methylation measurements on isolated cell types, it is possible to impute cell proportions and adjust for them, but there is increasing interest in methods that adjust for cell composition effects when reference sets are incomplete or unavailable., Results: In this article we present a theoretical basis for one such method, showing that the total effect of a phenotype on DNA methylation can be decomposed into orthogonal components, one representing the effect of phenotype on proportions of major cell types, the other representing either subtle effects in composition or global effects at focused loci, and that it is possible to separate these two types of effects in a finite data set. We demonstrate this principle empirically on nine DNA methylation data sets, showing that the first few principal components generally contain a majority of the information on cell-type present in the data, but that later principal components nevertheless contain information about a small number of loci that may represent more focused associations. We also present a new method for determining the number of linear terms to interpret as cell-mixture effects and demonstrate robustness to the choice of this parameter., Conclusions: Taken together, our work demonstrates that reference-free algorithms for cell-mixture adjustment can produce biologically valid results, separating cell-mediated epigenetic effects (i.e. apparent effects arising from differences in cell composition) from those that are not cell mediated, and that in general the interpretation of associations evident from DNA methylation should be carefully considered.

Background: Epidemiologic studies and animal models suggest that in utero arsenic exposure affects fetal health, with a negative association between maternal arsenic ingestion and infant birth weight often observed. However, the molecular mechanisms for this association remain elusive. In the present study, we aimed to increase our understanding of the impact of low-dose arsenic exposure on fetal health by identifying possible arsenic-associated fetal tissue biomarkers in a cohort of pregnant women exposed to arsenic at low levels., Methods: Arsenic concentrations were determined from the urine samples of a cohort of 133 pregnant women from New Hampshire. Placental tissue samples collected from enrollees were homogenized and profiled for gene expression across a panel of candidate genes, including known arsenic regulated targets and genes involved in arsenic transport, metabolism, or disease susceptibility. Multivariable adjusted linear regression models were used to examine the relationship of candidate gene expression with arsenic exposure or with birth weight of the baby., Results: Placental expression of the arsenic transporter AQP9 was positively associated with maternal urinary arsenic levels during pregnancy (coefficient estimate: 0.25; 95% confidence interval: 0.05 - 0.45). Placental expression of AQP9 related to expression of the phospholipase ENPP2 which was positively associated with infant birth weight (coefficient estimate: 0.28; 95% CI: 0.09 - 0.47). A structural equation model indicated that these genes may mediate arsenic's effect on infant birth weight (coefficient estimate: -0.009; 95% confidence interval: -0.032 - -0.001; 10,000 replications for bootstrapping)., Conclusions: We identified the expression of AQP9 as a potential fetal biomarker for arsenic exposure. Further, we identified a positive association between the placental expression of phospholipase ENPP2 and infant birth weight. These findings suggest a path by which arsenic may affect birth outcomes.

Background: There has been a long-standing need in biomedical research for a method that quantifies the normally mixed composition of leukocytes beyond what is possible by simple histological or flow cytometric assessments. The latter is restricted by the labile nature of protein epitopes, requirements for cell processing, and timely cell analysis. In a diverse array of diseases and following numerous immune-toxic exposures, leukocyte composition will critically inform the underlying immuno-biology to most chronic medical conditions. Emerging research demonstrates that DNA methylation is responsible for cellular differentiation, and when measured in whole peripheral blood, serves to distinguish cancer cases from controls., Results: Here we present a method, similar to regression calibration, for inferring changes in the distribution of white blood cells between different subpopulations (e.g. cases and controls) using DNA methylation signatures, in combination with a previously obtained external validation set consisting of signatures from purified leukocyte samples. We validate the fundamental idea in a cell mixture reconstruction experiment, then demonstrate our method on DNA methylation data sets from several studies, including data from a Head and Neck Squamous Cell Carcinoma (HNSCC) study and an ovarian cancer study. Our method produces results consistent with prior biological findings, thereby validating the approach., Conclusions: Our method, in combination with an appropriate external validation set, promises new opportunities for large-scale immunological studies of both disease states and noxious exposures.

Background: Epigenetics is the study of heritable changes in gene function that cannot be explained by changes in DNA sequence. One of the most commonly studied epigenetic alterations is cytosine methylation, which is a well recognized mechanism of epigenetic gene silencing and often occurs at tumor suppressor gene loci in human cancer. Arrays are now being used to study DNA methylation at a large number of loci; for example, the Illumina GoldenGate platform assesses DNA methylation at 1505 loci associated with over 800 cancer-related genes. Model-based cluster analysis is often used to identify DNA methylation subgroups in data, but it is unclear how to cluster DNA methylation data from arrays in a scalable and reliable manner., Results: We propose a novel model-based recursive-partitioning algorithm to navigate clusters in a beta mixture model. We present simulations that show that the method is more reliable than competing nonparametric clustering approaches, and is at least as reliable as conventional mixture model methods. We also show that our proposed method is more computationally efficient than conventional mixture model approaches. We demonstrate our method on the normal tissue samples and show that the clusters are associated with tissue type as well as age., Conclusion: Our proposed recursively-partitioned mixture model is an effective and computationally efficient method for clustering DNA methylation data.