Hypothesis

Aging increases stochastic epigenetic variation in circulating cell‑free DNA (cfDNA), but this signal is diluted by heterogeneous fragment sizes that reflect differing nucleosome protection across tissues and chromatin states. By stratifying the Epigenetic Instability Index (EII) according to cfDNA fragment length—specifically isolating short fragments (<150 bp) that originate from nucleosome‑protected regions and long fragments (>200 bp) that arise from more open chromatin—we can obtain a tissue‑agnostic measure of epigenetic noise that tracks biological age more accurately than unstratified EII or current methylation clocks.

Mechanistic Rationale

1. Nucleosome protection and fragment size – cfDNA fragment lengths mirror the positioning of nucleosomes at the tissue of origin. Short fragments are enriched from tightly wrapped DNA (e.g., heterochromatin, lamina‑associated domains), while long fragments derive from nucleosome‑depleted regions such as active promoters or enhancers. Age‑related chromatin remodeling alters the balance between these compartments, shifting the global fragment‑size distribution.

2. Epigenetic instability accumulates with age – Random gain or loss of methylation at CpG islands (CGIs) increases transcriptional noise and contributes to functional decline. The EII captures this variation across 269 CGI regions, proving that stochastic epigenetic patterns are detectable in cfDNA for cancer detection 1.

3. Tissue specificity confounds bulk measures – Existing epigenetic clocks trained on blood or saliva suffer because methylation levels differ by cell type, making it hard to distinguish true age‑related drift from compositional changes 2. Fragment‑size stratification inherently separates signals by chromatin state, which is less variable across tissues than absolute methylation levels.

4. Combining both dimensions improves specificity – By calculating EII separately for short‑fragment and long‑fragment pools, we generate two complementary indices: (a) a short‑fragment EII reflecting instability in heterochromatic regions linked to nuclear lamina stiffness, and (b) a long‑fragment EII capturing changes in euchromatic zones associated with transcriptional activity. The ratio or joint model of these indices should correlate strongly with physiological age and respond to interventions that alter chromatin architecture (e.g., senolytics, HDAC inhibitors).

Testable Predictions

• Prediction 1: In cross‑sectional human plasma cohorts, the short‑fragment EII will increase with chronological age more steeply than the long‑fragment EII, and the combined fs‑EII model will explain >70 % of variance in frailty index after adjusting for cell‑type composition.
• Prediction 2: Longitudinal sampling before and after a known aging intervention (e.g., exercise regimen or metformin) will show a reversible shift in the short‑/long‑fragment EII ratio, whereas global EII will remain unchanged.
• Prediction 3: Sorted cfDNA fractions by size will reveal distinct methylation‑variation signatures: short fragments will show elevated variability at lamina‑associated CGIs, while long fragments will display increased variability at enhancer CGIs.

Experimental Approach

1. Collect plasma from age‑stratified donors (20‑80 y, n = 200). Isolate cfDNA and separate into short (<150 bp) and long (>200 bp) fractions using SPRI bead selection.
2. Perform whole‑genome bisulfite sequencing (or targeted CGI panel) on each fraction; compute EII across the 269 loci used in the cancer study.
3. Model biological age using elastic‑net regression with fs‑EII, traditional clocks, and complete blood count covariates; compare performance via cross‑validated R² and MAE.
4. Validate in an independent interventional trial (n = 50) measuring pre/post changes.

If fs‑EII outperforms existing clocks and tracks intervention‑induced shifts, it will confirm that fragment‑size‑aware epigenetic instability provides a superior, tissue‑agnostic window into biological aging. Failure to observe these patterns would falsify the hypothesis and suggest that fragment size does not meaningfully enrich age‑related epigenetic noise.