Time-series Epigenome-wide Association Study — Longitudinal EWAS
Longitudinal Epigenome-wide Association Study · Also known as: time-series EWAS, longitudinal EWAS, repeated-measures EWAS, dynamic methylation association study
A time-series epigenome-wide association study (time-series EWAS) extends the classic cross-sectional EWAS design to longitudinal settings, measuring DNA methylation across the entire epigenome at multiple time points within the same subjects. The goal is to identify CpG sites whose methylation levels change systematically over time, or to characterise how epigenetic associations with an exposure or phenotype evolve across developmental stages, treatment periods, or disease trajectories.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use time-series EWAS when you have repeated within-subject DNA methylation data and want to detect dynamic epigenetic changes over time, evaluate whether an environmental exposure progressively alters the epigenome, or characterise developmental or treatment-induced methylation trajectories. It is especially valuable in birth-cohort studies, clinical trials with pre- and post-treatment samples, and aging research. Do not apply it to cross-sectional datasets — a standard EWAS is appropriate there. Avoid it when fewer than two matched time points are available per subject, when subject attrition between time points is severe enough to bias within-person comparisons, or when the number of subjects is too small (generally fewer than 50 per group) to support stable mixed-effects estimates across hundreds of thousands of CpG sites.
Strengths & limitations
- Within-subject design controls for stable individual-level confounders, substantially boosting power relative to cross-sectional EWAS.
- Directly models the temporal dynamics of methylation, enabling detection of progressive, reversible, or event-triggered epigenetic changes.
- Can disentangle pre-existing between-person differences in methylation from true longitudinal change.
- Applicable to diverse biological questions — development, aging, disease progression, and environmental exposure effects.
- Repeated measures increase the biological interpretability of findings by confirming that associations persist or evolve predictably.
- Requires matched longitudinal samples, which are expensive and logistically challenging to collect, particularly at multiple time points.
- Mixed-effects models at the genome-wide scale are computationally intensive; large cohorts with many time points demand substantial HPC resources.
- Dropout and missing time points introduce potential bias if missingness is not random (MNAR); complete-case analyses can produce misleading results.
- Cell-type composition changes between time points must be carefully modelled; failure to do so inflates false-positive rates for dynamic methylation.
- Cross-platform comparability is limited when different array generations are used at different time points.
Frequently asked
How is time-series EWAS different from a standard EWAS?
Standard EWAS is cross-sectional: each subject contributes one measurement, and the analysis associates methylation at a single time point with a trait. Time-series EWAS collects methylation at multiple time points from the same subjects and models intra-individual change over time. The key statistical difference is that repeated measures within subjects must be accounted for using mixed-effects models or GEE rather than simple linear regression.
How many time points and subjects are needed?
A minimum of two time points per subject is required; three or more allow non-linear trajectory modelling. For subjects, power depends on effect size and variance, but studies with fewer than 50 subjects per comparison group typically lack the power to detect modest longitudinal methylation effects at genome-wide significance. Simulation-based power calculations tailored to the specific model are recommended at the design stage.
Which software tools are commonly used?
Preprocessing is typically handled with minfi or ChAMP in R. Mixed-effects models genome-wide can be fitted with lme4 (via custom scripting), or specialised packages such as CpGassoc or OSCA. For large cohorts, computationally efficient implementations in OSCA or similar C++-backed tools are preferred. Pathway and annotation analyses use missMethyl, clusterProfiler, or Ingenuity Pathway Analysis.
Can I apply this to bisulfite sequencing data instead of arrays?
Yes. Whole-genome bisulfite sequencing (WGBS) or reduced-representation bisulfite sequencing (RRBS) can provide the methylation input. The same longitudinal modelling framework applies, but preprocessing differs (using tools such as Bismark and BSseq), and the multiple-testing burden increases substantially with full-genome coverage, requiring more stringent FDR control.
How should I handle subjects with missing time points?
Mixed-effects models can accommodate unbalanced data (subjects with different numbers of observations) under a missing-at-random (MAR) assumption. If missingness is potentially informative (e.g., dropout due to disease severity), sensitivity analyses using multiple imputation or mixed-model approaches that model the missingness mechanism are advisable. Complete-case analyses should be avoided unless dropout is clearly negligible.
Sources
- Pidsley, R., Zotenko, E., Peters, T. J., Lawrence, M. G., Risbridger, G. P., Molloy, P., ... & Clark, S. J. (2016). Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biology, 17(1), 208. link ↗
- Waterland, R. A., Kellermayer, R., Laritsky, E., Rayco-Solon, P., Harris, R. A., Travisano, M., ... & Prentice, A. M. (2010). Season of conception in rural Gambia affects DNA methylation at putative human metastable epialleles. PLoS Genetics, 6(12), e1001252. link ↗
How to cite this page
ScholarGate. (2026, June 3). Longitudinal Epigenome-wide Association Study. ScholarGate. https://scholargate.app/en/bioinformatics/time-series-epigenome-wide-association-study
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Epigenome-wide association studyBioinformatics↔ compare
- RNA-seq Differential ExpressionBioinformatics↔ compare
- Time-series single-cell RNA-seq analysisBioinformatics↔ compare