Sage Journals: Discover world-class research

Abstract

Epigenetic age is a biological age estimate based on nuclear DNA methylation patterns. Epigenetic clocks measure biological age by analyzing predictable changes in DNA methylation sites associated with aging. This study introduces EpigeneticAgePipeline, an R package that streamlines the estimation of epigenetic age metrics including Horvath, Horvath skin and blood, Hannum, PhenoAge/Levine, GrimAge (V1 and V2), and DunedinPACE plus additional acceleration metrics based on all other clocks. Quality control includes detection P-value filtering (sample- and probe-level), bead-count thresholds, and Illumina quality control intensity checks. EpigeneticAgePipeline supports Illumina Infinium methylation microarrays (HumanMethylation27, HumanMethylation450, HumanMethylationEPIC/EPICv2, and Human Methylation Screening Array). It offers functionalities including data preprocessing, normalization, cell count imputation, residual generation accounting for principal components and batch effects, and extensive visualizations for improved interpretability. Validation was performed using GEO data set GSE237561, confirming the accuracy of the pipeline. EpigeneticAgePipeline provides an integrated workflow from raw data to advanced statistical analyses and visualizations, improving usability over existing tools. In addition to the traditional clocks mentioned, the package also integrates a set of additional epigenetic age clocks (PedBE, Wu, TL, BLUP, and EN). Future updates will include emerging epigenetic age measures to maintain relevance in this evolving field.

Keywords

Epigenetic age DNA methylation epigenetic clocks biological aging bioinformatics pipeline R package methylation microarrays

Introduction

The field of epigenomics has led to the understanding that environmental factors can significantly influence our genome and how particular regions of our genome are regulated and expressed. As the field has expanded, research into the association of epigenetic features, human health, and disease risk, as well as aging has emerged, leading to the concept of epigenetic age and the creation of epigenetic age clocks. Epigenetic clocks are computational tools that estimate biological age by analyzing specific DNA methylation patterns that are associated with cellular aging across the genome. These clocks use measures of DNA methylation, chemical modifications that regulate gene expression without altering the underlying DNA sequence, to provide insight into the aging process. This is important because epigenetic age may be a better measure of an individual’s health status than chronological age alone. EpigeneticAgePipeline aims to build off the functionality of other well-established R packages such as dnaMethyAge,¹ methylclock,² and minfi³ to provide additional functionality and significantly improve usability.

The aim of this article is to showcase the complete workflow performed by the package, from data preprocessing to visualization. This comprehensive approach emphasizes the seamless integration of functionalities, highlighting the distinct advantages of EpigeneticAgePipeline compared to other tools that typically require multiple separate steps to achieve similar outcomes.

What makes EpigeneticAgePipeline unique is that it integrates the entire workflow, including data preprocessing, epigenetic age calculations, and downstream visualization tasks/final analysis into 1 seamless pipeline. The package allows for direct input of raw intensity data (intensity data file [IDAT]) files (27K, 450K, EPICv1/v2, and MSA microarray), which are then processed via sample and probe-level detection P-value filtering, bead-count thresholds at both the sample and probe level, Illumina control intensity checks, and the removal of cross-reactive probes and sex probes.^4-8 Alternatively, users can input their own preprocessed methylation beta values. The package also allows for residual generation to account for cell count data and batch effects that arise during methylation data collection. Residual generation also includes the option to regress out principal components (PCs) generated from the methylation data set to account for additional sources of variation in the data. Cell proportion estimation is also included, allowing the user to specify tissue type and methodology.

Furthermore, EpigeneticAgePipeline provides a range of visualizations, including various correlation and distribution plots that enhance the data’s interpretability. These features, combined with the foundational functionalities of other well-established R packages, allow EpigeneticAgePipeline to fill a unique niche in computational epigenetics.

Methods

EpigeneticAgePipeline generates 6 common epigenetic age clocks, which include Horvath clock⁹ and Horvath Bayesian Neural Network (BNN) clock,¹⁰ Horvath skin and blood clock,¹¹ Hannum clock,¹² PhenoAge/Levine clock¹³ and GrimAge (V1 and V2 based on their published surrogate panels).^14,15 We have also included the PC versions for each of these clocks.¹ In addition to these 5 common clocks, the pipeline also generates the PedBE clock,¹⁶ the Wu clock,¹⁷ the Telomere Length (TL) clock,¹⁸ the BLUP clock,¹⁹ and the EN clock.¹⁹ Each 1 analyzes the methylation pattern of a unique set of CpG (Cytosine-phosphate-Guanine) sites, which allows for more specific estimations depending on the tissue type and/or phenotype. Along with epigenetic age estimation, the package allows for the generation of DunedinPACE,²⁰ which reflects the pace of biological aging and age acceleration for each of the generated clocks.

When raw IDATs are provided, the pipeline performs detection P-value filtering at 2 levels: (1) sample-level retention if the mean detection P-value is < .05 and (2) probe-level retention only if the detection P-value is < .01 across passing samples. Bead-count filters exclude samples with >.05 low-bead observations (bead count < 3) and drop probes with >.05 low-bead observations among retained samples. Default parameters concerning detection P-value and bead-count filtering can also be modified by the user. Illumina control intensity metrics are assessed via minfi, where samples with mean log2 intensity of mMed and uMed < 10.5 are removed.³ The pipeline then removes known cross-reactive probes and sex-chromosome probes. For each of the microarray formats accepted (27K, 450K, EPICv1/v2, MSA), a separate list of previously identified cross-reactive probes is removed during this process.^21-24 Currently, there has been no formal study that compiled a list of cross-reactive probes for the MSA platform; however, due to its similarity to EPICv2, we used the EPICv2 list.

To handle replicate probes in newer EPICv2/MSA array formats targeting the same legacy CpG, beta values are collapsed by probe-ID prefix using betasCollapseToPfx from the sesame R package, yielding a single averaged beta per CpG prior to clock computation.²⁵

To calculate cell proportion estimates, the package supports Constrained Projection (CP) and Robust Partial Correlations (RPC) style generation for adult blood, cord blood, saliva, placenta (first and third trimester), and frontal cortex cells.^26-33 Buccal cell proportions are also supported using the HEpiDISH (from EpiDISH R Package) hierarchical approach by first deconvolving epithelial, fibroblast, and immune fractions with a solid-tissue reference, then partitioning the immune fraction into blood cell subtypes.^27,34,35

For all clocks except DunedinPACE, the package provides 3 variations of epigenetic age acceleration: (1) the difference between Epigenetic Age and chronological age, (2) residuals from Epigenetic Age ~ Age + estimated cell proportions, and (3) residuals from Epigenetic Age ~ Age. These are analogous to the age accelerations provided by methylclock.²

In designing EpigeneticAgePipeline, we implement a curated set of epigenetic aging measures intended to cover the major classes of DNAm aging estimators that dominate comparative disease, mortality, and aging research.

Systematic reviews and large comparative studies consistently emphasize a core family of clocks spanning first-generation chronological age predictors (eg, Horvath, Hannum), phenotypic/mortality-trained clocks (eg, PhenoAge, GrimAge), pace-of-aging measures (eg, DunedinPACE/DunedinPoAm38), and DNAm-based telomere length estimators (DNAmTL). We therefore prioritize these families for conceptual coverage and practical relevance.^36-40 We also include pediatric postnatal clocks (PedBE and Wu) to support child/adolescent cohorts and nonlinear (BNN) chronological predictors.^41,42

The original Horvath clock was one of the first epigenetic clocks created and remains a commonly used epigenetic age measure. The clock uses a set of 353 CpG sites trained on 27K/450K arrays and is designed to be used on a wide range of tissue types. We also included a Horvath BNN clock, which generates a BNN using Horvath CpG sites to predict age. It has been shown to be slightly more accurate than default Horvath in predicting age. The second original epigenetic clock is the Hannum clock. This clock relies on a set of 71 CpG sites trained on a 450K array and is primarily designed for use on blood samples. Both the Horvath and Hannum clocks have been shown to be significant predictors of age- and age-related conditions. After the creation of the Horvath clock, the Horvath skin and blood clock was developed using a set of 391 CpG sites trained on 450K/EPICv1 array types for better estimations from skin, blood, saliva, human fibroblasts, keratinocytes, buccal cell, endothelial cell, and lymphoblastoid cell inputs. Another epigenetic age clock included is the Levine/PhenoAge clock, which was created as a potential outperformer of other clocks concerning predictions of age-based phenotypic outcomes; it uses a set of 513 CpG sites trained on 27K/450K/EPICv1. GrimAge is a measure primarily designed for blood samples, with versions 1 and 2 based on a set of 1030 CpGs, difference being that version 2 uses an additional 2 DNAm (DNA methylation) surrogates. The surrogates are DNAm log(C-reactive protein) and DNAm log(Hemoglobin A1c), which allows GrimAge2 to outperform the first in mortality prediction.¹⁵ GrimAge has been shown to predict a high Mortality Hazard Ratio compared to other common epigenetic age measures. The PedBE clock uses 94 CpGs and was trained primarily on buccal epithelial samples (children and adolescents) profiled on 450K/EPIC platforms. It is optimized for pediatric tissues and has been used to track developmental timing and age acceleration in childhood cohorts. The Wu clock uses 111 CpGs and was trained on pediatric blood using 27K and 450K arrays. It provides highly accurate age estimates in children and has been applied to investigate developmental and clinical outcomes in pediatric populations. The TL clock uses 140 CpGs trained on blood using 450K and EPIC arrays to estimate leukocyte telomere length. It has been used to identify associations between shorter telomere length and advancing age and is related to aging-related morbidity. The BLUP clock uses 319 607 CpGs trained on blood and saliva measured on 450K and EPIC arrays. It leverages genome-wide features to yield high-precision age prediction and robust performance across cohorts and tissues. The EN clock uses 514 CpGs trained on the same 450K/EPIC blood and saliva data sets as BLUP. The elastic-net (EN) clock provides a sparse, portable predictor with strong cross-platform accuracy and reliable performance in external data.

The DunedinPACE measure is a measure of the pace of biological aging, which has been known to show strong associations with phenotypic outcomes like morbidity, disability, and mortality. DunedinPACE was developed for use in blood samples and uses a set of 173 CpG sites.

Unless otherwise justified, we recommend applying each epigenetic age measure to the tissue type and age range in which it was originally trained and validated (see Table 1), as performance can degrade or become difficult to interpret outside that domain. Accordingly, EpigeneticAgePipeline documents the intended tissue and age context for every clock and is designed with the expectation that users will supply methylation data derived from those recommended sample types, treating out-of-domain applications as exploratory.

Table 1.

Implemented clocks in EpigeneticAgePipeline, their recommended sample/tissue contexts, and key alternative clocks excluded with rationale.

Clock included in EpigeneticAgePipeline	Recommended sample parameters	Number of CpG sites	Potential alternatives considered but not implemented	Reason(s) for non-inclusion of alternatives
First-generation chronological age predictors
Horvath/Horvath BNN^9,10	Pan-tissue, trained on 27K, 450K	353	Weidner et al,⁴³ Lin and Wagner,⁴⁴ Vidal-Bralo et al,⁴⁵ Garagnani et al,⁴⁶ Yang et al,⁴⁷ Bocklandt et al⁴⁸	Largely redundant within the chronological age category, alternative clocks show substantially weaker age correlation in large cohorts.⁴⁹
Hannum et al¹²	Whole peripheral blood, trained on 450K, 71 CpG sites
Horvath Skin and Blood¹¹	Fibroblasts, keratinocytes, buccal, endothelial, saliva, lymphoblastoid, skin and blood, trained on 450K	391
Zhang EN/BLUP¹⁹	Blood, saliva, trained on 450K and EPICv1	EN uses 514, BLUP uses 319607
Second-generation phenotypic/mortality clocks
PhenoAge (Levine)¹³	Whole peripheral blood, trained on 27K, 450K and EPICv1	513	Zhang et al⁵⁰	Largely redundant as PhenoAge/GrimAgeV1/GrimAgeV2 show larger disease associations.⁵¹
GrimAgeV1/GrimAgeV2^14,15	Whole peripheral blood, trained on 450K and EPICv1	1030	Zhang et al⁵⁰
Pace of aging estimator
DunedinPACE²⁰	Whole peripheral blood, trained on 450K and EPICv1	173	DunedinPoAm⁵²	DunedinPACE is more precise and has higher test-retest reliability.²⁰
Pediatric chronological age predictor
PedBE¹⁶	Buccal, trained on 450K and EPICv1	94	Knight et al⁵³, Bohlin et al⁵⁴	Out of scope for postnatal aging, designed as gestational clocks for pregnancy/neonatal development.
Wu¹⁷	Buccal, trained on 27K and 450K	111	Knight et al⁵³, Bohlin et al⁵⁴
Proxy of leukocyte telomere length
DNAmTL et al¹⁸	Whole peripheral blood, trained on 450K and EPICv1	140	NA	DNAmTL is the standard DNAm telomere estimator.¹⁸

EpigeneticAgePipeline also incorporates residual generation to allow for correction of important confounding variables. Principal component analysis is performed on beta values, and residuals are generated adjusting for the first 5 PCs (default) generated on the methylation data. Samples that are beyond 3 times (default) the median absolute deviation in any of the first PCs are marked as outliers. Residual generation also incorporates adjustments for technical effects that commonly arise from methylation array measurements. The user may provide data for array variables such as Row, Column, Slide, and Batch. Depending on which ones are given, the model will dynamically construct a suitable formula for the regression (see Figure 1). There is also a built-in option to compute and adjust for cell count proportion estimates. Residuals generated when chronological age is set as a predictor for biological age effectively gives a measure of age acceleration.

Figure 1.

Flowchart illustrating formula construction for the linear model during residual generation (generateResiduals function).

A series of plots and visualizations are also included in the final output to present the correlations between epigenetic age measures and chronological age as well as the distribution of epigenetic age estimates for every sample. Raw data are exported as a .txt file which can be used for further analysis.

Results

The package’s main function is the primary entry point for epigenetic age analysis. Users provide the directory containing methylation and covariate data, if available (eg, age, sex, smoking status, batch effects, etc), along with options for normalizing beta values, selecting the methylation array type, and processing raw or precomputed data. The function performs preprocessing steps, cell count generation, and the creation of a file containing beta values for downstream analysis. This function also supports parallel processing for efficient handling of large data sets and allows users to specify sample tissue assumptions during cell count generation.

The generateResiduals function is designed for advanced analysis following epigenetic age estimation. This function builds a linear model based on user-specified variables or PCs and outputs the corresponding residuals, which can be used for further investigation of epigenetic age-related patterns. Users can input custom formulas for the model, specify the type of methylation array used, and include additional covariates such as cell counts in the analysis. The function also supports the detection and removal of highly correlated variables (if enabled) and identifies outliers using median absolute deviation (MAD). As with the main function, this process can leverage parallel processing for efficiency and accommodates both adult and cord blood sample assumptions during cell count generation. To view a complete outline of the workflow for either the main or generateResiduals function (see Figures 2 and 3).

Figure 2.

Workflow diagram illustrating the primary functionalities of the main() function.

Figure 3.

Workflow diagram illustrating the primary functionalities of the generateResiduals() function.

To ensure the robustness of our pipeline, we first validated its implementation using the publicly available Gene Expression Omnibus (GEO) data set GSE237561.⁵⁵ This data set contains nuclear DNA methylation data for 126 samples: 92 measured using the 450K array and 34 using the EPIC platform. Since both methylation and phenotypic data are publicly available, we were able to compare our results against previously published findings. Specifically, a prior study investigating the relationship between clozapine use and epigenetic changes utilized this data set to calculate PhenoAge/Levine and DunedinPACE epigenetic age measures.⁵⁶ These epigenetic markers were incorporated into linear mixed models to assess whether clozapine use predicts epigenetic age or age acceleration. To validate our pipeline, we used all 126 samples and compared our computed PhenoAge/Levine and DunedinPACE measures against the published results. Our package generates these estimates using the methylclock and dnaMethyAge packages, respectively; this comparison, therefore, seeks to confirm the accuracy of our implementation. The full validation results can be found in the Supplementary Information. Following this validation, we applied our pipeline specifically to the 34 EPIC-measured samples for visualization. The covariates included in this analysis were sex, age, smoking score, and number of days on clozapine. Figure 4 as well as all Supplementary Figures (except validation) were generated using our EpigeneticAgePipeline on the EPIC-measured samples from the GSE237561 data set.

Figure 4.

A correlation matrix generated for Horvath epigenetic age from the GSE237561 data set.

Inputs: The package accepts methylation data either in the form of a .csv file containing methylation beta values for each respective sample or a directory containing raw IDAT files.

Outputs: The package outputs raw epigenetic age/acceleration values (.txt and .md, Supplemental Table S1), raw residual results (.csv), age and epigenetic age correlation plots (Supplemental Figure S2), matrix correlation plots (.png, Figure 3), and an epigenetic age distribution plot (.png, Figure 3). A clock-coverage summary table is also outputted, reporting for each clock/surrogate the total CpGs, missing CpGs, and percent missing (Supplemental Table S2). Examples of all the plots can be found in the Supplementary Information files and our GitHub package page (https://github.com/CastellaniLab/EpigeneticAgePipeline).

Specific information on installation procedures and file specificities, as well as vignettes, is located within the GitHub readme.⁵⁷

Discussion

EpigeneticAgePipeline provides a comprehensive and user-friendly method to generate epigenetic age measures for those studying the effects of biological aging. Our pipeline better accounts for potential batch effects and confounding variables to minimize the effect of potential biases on the results in studies concerning biological aging. Our pipeline provides a complete workflow for epigenetic age estimation, starting with raw methylation data all the way to the generation of epigenetic age measures, incorporating a variety of quality control methods and visualizations.

Limitations

The accuracy and interpretability of epigenetic age estimates depend on applying each clock within the tissue types and age ranges for which it was trained on. Although the pipeline allows for flexible inputs, results obtained from out-of-domain sample specifications should be interpreted as exploratory. EpigeneticAgePipeline implements a comprehensive collection of epigenetic age clocks for the most studied tissues and use cases. More specialized or less commonly applied clocks are not included by default and would need to be generated separately by users.

Conclusion

EpigeneticAgePipeline serves as a valuable tool for investigating the relationships between epigenetic age measures and a wide range of health outcomes. Future updates will include the addition of emerging epigenetic aging measures/methods to ensure the pipeline remains relevant to this expanding field.

Supplemental Material

sj-pdf-1-bbi-10.1177_11779322261433665 – Supplemental material for EpigeneticAgePipeline: An R Package for Comprehensive Assessment of Epigenetic Age Metrics From Methylation Microarrays

Supplemental material, sj-pdf-1-bbi-10.1177_11779322261433665 for EpigeneticAgePipeline: An R Package for Comprehensive Assessment of Epigenetic Age Metrics From Methylation Microarrays by Stanislav Rayevskiy, Quinn Le, Julia Nguyen, Steven Q. Chen and Christina A. Castellani in Bioinformatics and Biology Insights

Footnotes

Acknowledgements

We acknowledge the Digital Research Alliance and the Canadian Foundation for Innovation J. Evans Leaders Fund grant #43481 for the support of the computing infrastructure used throughout our analyses.

Authors’ Note

Project Name: EpigeneticAgePipeline.

Project Homepage: .

Operating Systems: Platform-independent.

Programming Language: R.

Other Requirements: R 4.2+.

License: CC0 1.0.

Any restrictions to use by non-academics: No.

Author Contributions

Stanislav Rayevskiy: led the software development portion of the project, contributed to testing and manuscript preparation, and read and approved the final manuscript.

Quinn Le: involved in extensive software testing, co-wrote the manuscript, and read and approved the final manuscript.

Julia Nguyen: contributed to code development, provided critical feedback during the early stages of the project, and read and approved the final manuscript.

Steven Q Chen: assisted with testing, provided additional validation support, and read and approved the final manuscript.

Christina A. Castellani: conceptualized and supervised the project, provided funding, oversaw all aspects of development and manuscript preparation, and read and approved the final manuscript.

ORCID iDs

Stanislav Rayevskiy

Quinn Le

Steven Q. Chen

Christina A. Castellani

Ethics Approval and Consent to Participate

Not applicable, this study used publicly available data sets and did not involve any human participants, human tissue, or human data collected by the authors.

Consent for Publication

Not applicable.

Data Availability Statement

The data sets generated and/or analyzed during the current study are available in the Gene Expression Omnibus repository, accession code GSE237561 ().

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by start-up funds to CC from the Department of Pathology and Laboratory Medicine at Western University. It was further supported by institutional summer support (USRI) from Western University to SR.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental Material for this article is available online.

References

Wang

Grant

Zhai

Mcdonald-Maier

Schalkwyk

LC.

Insights into ageing rates comparison across tissues from recalibrating cerebellum DNA methylation clock. GeroScience. 2024;46:39-56. doi:10.1007/s11357-023-00871-w

Pelegí-Sisó

de Prado

Ronkainen

Bustamante

González

JR.

Methylclock: a Bioconductor package to estimate DNA methylation age. Bioinformatics. 2021;37:1759-1760. doi:10.1093/bioinformatics/btaa825

Aryee

Jaffe

Corrada-Bravo

, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;30:1363-1369. doi:10.1093/bioinformatics/btu049

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/IlluminaHumanMethylationEPICanno.ilm10b4.hg19/

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/IlluminaHumanMethylation27kanno.ilmn12.hg19/

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/IlluminaHumanMethylationEPICv2anno.20a1.hg38/

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/IlluminaHumanMethylation450kanno.ilmn12.hg19/

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/IlluminaHumanMethylationMSAanno.ilm10a1.hg38/

Horvath

DNA methylation age of human tissues and cell types. Genome Biol. 2013;14:3156. doi:10.1186/gb-2013-14-10-r115

10.

Alfonso

Gonzalez

JR.

Bayesian neural networks for the optimisation of biological clocks in humans. bioRxiv. 2020. doi:10.1101/2020.04.21.052605

11.

Horvath

Oshima

Martin

, et al. Epigenetic clock for skin and blood cells applied to Hutchinson Gilford Progeria Syndrome and ex vivo studies. Aging. 2018;10:1758-1775. doi:10.18632/aging.101508

12.

Hannum

Guinney

Zhao

, et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell. 2013;49:359-367. doi:10.1016/j.molcel.2012.10.016

13.

Levine

Quach

, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging. 2018;10:573-591. doi:10.18632/aging.101414

14.

Quach

Wilson

, et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging. 2019;11:303-327. doi:10.18632/aging.101684

15.

Binder

Zhang

, et al. DNA methylation GrimAge version 2. Aging. 2022;14:9484-9549. doi:10.18632/aging.204434

16.

McEwen

O’Donnell

McGill

, et al. The PedBE clock accurately estimates DNA methylation age in pediatric buccal cells. Proc Natl Acad Sci. 2020;117:23329-23335. doi:10.1073/pnas.1820843116

17.

Chen

Lin

, et al. DNA methylation profile is a quantitative measure of biological aging in children. Aging. 2019;11:10031-10051. doi:10.18632/aging.102399

18.

Seeboth

Tsai

P-C

, et al. DNA methylation-based estimator of telomere length. Aging. 2019;11:5895-5923. doi:10.18632/aging.102173

19.

Zhang

Vallerga

Walker

, et al. Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing. Genome Med. 2019;11:54. doi:10.1186/s13073-019-0667-1

20.

Belsky

Caspi

Corcoran

, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife. 2022;11:e73420. doi:10.7554/eLife.73420

21.

Chen

Choufani

Grafodatskaya

Butcher

Ferreira

Weksberg

Cross-reactive DNA microarray probes lead to false discovery of autosomal sex-associated DNA methylation. Am J Hum Genet. 2012;91:762-764. doi:10.1016/j.ajhg.2012.06.020

22.

Chen

Lemire

Choufani

, et al. Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray. Epigenetics. 2013;8:203-209. doi:10.4161/epi.23470

23.

Pidsley

Zotenko

Peters

, et al. Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biol. 2016;17:208. doi:10.1186/s13059-016-1066-1

24.

Peters

Meyer

Ryan

, et al. Characterisation and reproducibility of the HumanMethylationEPIC v2.0 BeadChip for DNA methylation profiling. BMC Genomics. 2024;25:251. doi:10.1186/s12864-024-10027-5

25.

Zhou

Triche

Jr Laird

Shen

SeSAMe: reducing artifactual detection of DNA methylation by Infinium BeadChips in genomic deletions. Nucleic Acids Res. 2018;46:e123. doi:10.1093/nar/gky691

26.

Houseman

Accomando

Koestler

, et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics. 2012;13:86. doi:10.1186/1471-2105-13-86

27.

Teschendorff

Breeze

Zheng

Beck

A comparison of reference-based algorithms for correcting cell-type heterogeneity in Epigenome-Wide Association Studies. BMC Bioinformatics. 2017;18:105. doi:10.1186/s12859-017-1511-5

28.

Bioconductor. n.d. Accessed September 25, 2025. http://bioconductor.org/packages/FlowSorted.Blood.450k/

29.

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/FlowSorted.Blood.EPIC/

30.

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/FlowSorted.CordBloodCombined.450k/

31.

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/BeadSorted.Saliva.EPIC/

32.

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/planet/

33.

Bioconductor. n.d. Accessed September 19, 2025. http://bioconductor.org/packages/FlowSorted.DLPFC.450k/

34.

Zheng

Webster

Dong

, et al. A novel cell-type deconvolution algorithm reveals substantial contamination by immune cells in saliva, buccal and cervix. Epigenomics. 2018;10:925-940. doi:10.2217/epi-2018-0037

35.

Salas

Zhang

Koestler

, et al. Enhanced cell deconvolution of peripheral blood using DNA methylation for high-resolution immune profiling. Nat Commun. 2022;13:761. doi:10.1038/s41467-021-27864-7

36.

Margiotti

Monaco

Fabiani

Mesoraca

Giorlandino

Epigenetic clocks: in aging-related and complex diseases. Cytogenet Genome Res. 2023;163:247-256. doi:10.1159/000534561

37.

Hillary

Stevenson

McCartney

, et al. Epigenetic measures of ageing predict the prevalence and incidence of leading causes of death and disease burden. Clin Epigenetics. 2020;12:115. doi:10.1186/s13148-020-00905-6

38.

Jain

Binder

Chen

, et al. Analysis of epigenetic age acceleration and healthy longevity among older US women. JAMA Netw Open. 2022;5:e2223285. doi:10.1001/jamanetworkopen.2022.23285

39.

Chervova

Panteleeva

Chernysheva

, et al. Breaking new ground on human health and well-being with epigenetic clocks: a systematic review and meta-analysis of epigenetic age acceleration associations. Ageing Res Rev. 2024;102:102552. doi:10.1016/j.arr.2024.102552

40.

Yamada

Epigenetic clocks and EpiScore for preventive medicine: risk stratification and intervention models for age-related diseases. J Clin Med. 2025;14:3604. doi:10.3390/jcm14103604

41.

Fang

Zhou

Perng

, et al. Evaluation of pediatric epigenetic clocks across multiple tissues. Clin Epigenetics. 2023;15:142. doi:10.1186/s13148-023-01552-3

42.

Wang

Zhou

W-H.

Epigenetic clocks in the pediatric population: when and why they tick?

Chin Med J (Engl). 2021;134:2901. doi:10.1097/CM9.0000000000001723

43.

Weidner

Lin

Koch

, et al. Aging of blood can be tracked by DNA methylation changes at just three CpG sites. Genome Biol. 2014;15:R24. doi:10.1186/gb-2014-15-2-r24

44.

Lin

Wagner

Epigenetic aging signatures are coherently modified in cancer. PLoS Genet. 2015;11:e1005334. doi:10.1371/journal.pgen.1005334

45.

Vidal-Bralo

Lopez-Golan

Gonzalez

Simplified assay for epigenetic age estimation in whole blood of adults. Front Genet. 2016;7:126. doi:10.3389/fgene.2016.00126

46.

Garagnani

Bacalini

Pirazzini

, et al. Methylation of ELOVL2 gene as a new epigenetic marker of age. Aging Cell. 2012;11:1132-1134. doi:10.1111/acel.12005

47.

Yang

Wong

Kuh

, et al. Correlation of an epigenetic mitotic clock with cancer risk. Genome Biol. 2016;17:205. doi:10.1186/s13059-016-1064-3

48.

Bocklandt

Lin

Sehl

, et al. Epigenetic predictor of age. PLoS ONE. 2011;6:e14821. doi:10.1371/journal.pone.0014821

49.

Crimmins

Thyagarajan

Levine

Weir

Faul

Associations of age, sex, race/ethnicity, and education with 13 epigenetic clocks in a nationally representative U.S. sample: the health and retirement study. J Gerontol Ser A. 2021;76:1117-1123. doi:10.1093/gerona/glab016

50.

Zhang

Wilson

Heiss

, et al. DNA methylation signatures in peripheral blood strongly predict all-cause mortality. Nat Commun. 2017;8:14617. doi:10.1038/ncomms14617

51.

Mavrommatis

Belsky

Ying

, et al. An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes. Nat Commun. 2025;16:11164. doi:10.1038/s41467-025-66106-y

52.

Belsky

Caspi

Arseneault

, et al. Quantification of the pace of biological aging in humans through a blood test, the DunedinPoAm DNA methylation algorithm. eLife. 2020;9:e54870. doi:10.7554/eLife.54870

53.

Knight

Craig

Theda

, et al. An epigenetic clock for gestational age at birth based on blood methylation data. Genome Biol. 2016;17:206. doi:10.1186/s13059-016-1068-z

54.

Bohlin

Håberg

Magnus

, et al. Prediction of gestational age based on genome-wide differentially methylated regions. Genome Biol. 2016;17:207. doi:10.1186/s13059-016-1063-4

55.

GEO. Accession viewer. n.d. Accessed September 19, 2025. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE237561

56.

Gillespie

Walker

Hannon

, et al. Longitudinal changes in DNA methylation associated with clozapine use in treatment-resistant schizophrenia from two international cohorts. Transl Psychiatry. 2024;14:390. doi:10.1038/s41398-024-03102-8

57.

Castellani

. CastellaniLab/EpigeneticAgePipeline. 2025. https://github.com/CastellaniLab/EpigeneticAgePipeline

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

11.55 MB