For years, circulating cell-free DNA (cfDNA) isolated from blood or urine was treated primarily as a genetic ledger—a soup of degraded fragments mined exclusively for specific tumor mutations or fetal chromosomal abnormalities. However, looking only at the genetic sequence ignores a fundamental physical reality: cfDNA is not chopped up at random.
When a cell undergoes apoptosis or necrosis, its genome is exposed to intracellular and extracellular nucleases. This degradation process is governed by the physical architecture of the chromatin. DNA that is tightly wrapped around nucleosomes or bound by transcription factors is physically shielded from enzymatic cleavage, while open, accessible, and unmethylated regions are highly vulnerable. As a result, the lengths, endpoints, and structural features of cfDNA fragments preserve a high-fidelity footprint of the epigenetic state and chromatin architecture of their cells of origin. The study of these patterns is known as fragmentomics.
The Physical Link Between Fragmentation and Epigenetics
To understand why fragmentomics works, consider the structural constraints of a nucleosome. Roughly 147 base pairs of DNA wrap around a histone core, with an additional 20 base pairs spanning the linker region to the next nucleosome, totaling approximately 167 base pairs.
When nucleases cleave healthy cell-free DNA, they preferentially cut the exposed linker regions, yielding a highly characteristic peak in fragment length distributions around 167 base pairs. However, this baseline changes dramatically depending on the local epigenetic landscape:
- Methylation and Open Chromatin: Unmethylated DNA typically resides in open, transcriptionally active chromatin (euchromatin). Because this DNA lacks the dense, protective compaction of hypermethylated heterochromatin, it is far more susceptible to rapid enzymatic breakdown. As demonstrated by Jensen et al., this creates a direct correlation between methylation status and fragment length: shorter cfDNA fragments are significantly more likely to be unmethylated. While some microbial metagenomic applications treat an organism's methylome as a stable fingerprint independent of fragment length (Heidelbach et al.), in human physiology, the two are deeply intertwined.
- The Tumor Signature: Malignant cells are characterized by global hypomethylation and disrupted chromatin folding—a nanoscale physical decompaction that can be observed directly in early carcinogenesis using super-resolution microscopy (Xu et al.). Because of this open structure, tumor-derived cfDNA (ctDNA) is consistently shorter than healthy background cfDNA, frequently dropping well below the 167-base-pair nucleosomal threshold. As reviewed by Stejskal et al., this shortened size profile acts as a physical indicator of malignancy. The author of MSK-ACCESS leveraged this exact property, using fragment size distributions as an independent biological signature to validate whether low-frequency genetic variants originated from a tumor or from healthy background noise.
By reversing this logic, researchers can infer the original epigenetic state of a molecule without performing destructive chemical treatments like bisulfite sequencing. For instance, Liu et al. developed a computational framework, FinaleMe, that predicts DNA methylation directly from standard whole-genome sequencing. The model succeeds by analyzing the spatial positioning of CpG sites relative to the center of each cfDNA fragment; because nucleosome wrapping protects specific regions, a CpG site's distance from the fragment center serves as a reliable proxy for its underlying methylation and protection status.
Mapping Chromatin Accessibility Across the Genome
Because fragmentation patterns reflect local chromatin states, analyzing these profiles across the entire genome allows researchers to map tissue-specific gene regulation non-invasively.
Instead of looking at a single locus, approaches like DELFI, developed by Mathios et al., evaluate genome-wide fragmentation profiles by measuring the size and distribution of millions of fragments across large genomic windows. When a tumor is present, the influx of short, aberrantly cleaved fragments alters the local size distribution within specific genomic bins. By tracking these shifts at known regulatory regions—such as the binding sites for the transcription factor ASCL1—the authors successfully differentiated between small cell and non-small cell lung cancer subtypes based entirely on the chromatin accessibility signatures left behind in the blood.
Healthy Cell (Closed Chromatin)
[---Nucleosome Core---] ======= [---Nucleosome Core---]
Protected Exposed Linker (Cut Site)
Result: Standard ~167 bp fragments
Tumor Cell (Open/Hypomethylated Chromatin)
[---Compacted Core Disrupted---] =====================
Highly Exposed to Nucleases (Multiple Cut Sites)
Result: Shorter, highly variable fragments (<150 bp)
To decode these complex mixtures without prior knowledge of the tumor's specific mutations, Renaud et al. introduced an unsupervised approach using Non-Negative Matrix Factorization (NMF). By applying NMF to fragment length histograms across genomic bins, their method deconvolves the overlapping fragment signatures of various tissues, allowing for the direct inference of regional chromatin accessibility and tumor burden from shallow, low-coverage sequencing data.
The Anatomy of a Breakpoint: End Motifs and Jagged Ends
The structural information embedded in cfDNA extends down to the exact nucleotides where a strand is cut. The orientation of the nuclease, the twist of the DNA helix, and the local sequence context combine to leave highly specific signatures at the fragment ends.
End Motifs and Nuclease Landscapes
Nucleases do not slice DNA blindly; they have distinct sequence preferences. By analyzing the short nucleotide sequences immediately adjacent to a cleavage site (known as end motifs or breakpoint motifs), fragmentomic assays can identify which enzymes were active. Zhang et al. demonstrated that extending the analysis of these end motifs to six base pairs captures highly complex biological signatures unique to specific tumor microenvironments, offering a mutation-independent signal for early-stage cancer detection.
Furthermore, these end motifs reveal that the enzymatic landscape varies dramatically across biological fluids. As shown by Chen et al. using nuclease knockout models, DNASE1L3 is the dominant nuclease shaping cfDNA in blood plasma, whereas DNASE1 drives fragmentation in urine. By mapping the density of these urinary fragment ends to known DNase I hypersensitive sites, the authors demonstrated that urinary fragmentomics directly reflects the open chromatin states of the urological tract.
Single-Stranded Jagged Ends
When a double-stranded break occurs, nucleases rarely leave a perfectly blunt cut. Instead, one strand often extends past the other, creating a single-stranded overhang termed a "jagged end."
Using a high-resolution sequencing method called Jag-seq, Xie et al. discovered that urinary cfDNA possesses exceptionally long jagged ends that exhibit a strict 10-nucleotide periodicity. This 10-bp interval is not an accident; it corresponds precisely to the physical helical pitch of the DNA molecule as it wraps around the histone core. Because the outer surface of the DNA helix faces away from the histone every 10 base pairs, it is exposed to enzymatic cleavage at exact periodic intervals. When a tumor alters the local chromatin structure, this strict structural periodicity is disrupted, providing a highly sensitive diagnostic signal for malignancies like renal cell carcinoma.
Clinical Applications and Confounding Factors
Because fragmentomics relies on the physical properties of millions of circulating molecules rather than the presence of a single rare mutation, it remains highly effective even when tumor concentrations are exceptionally low. This makes it a powerful tool for early cancer screening, distinguishing malignant tissue from benign nodules (Liu et al.), and expanding liquid biopsies into non-cancerous conditions. For example, Peddu et al. showed that viral cfDNA from cytomegalovirus is significantly more fragmented than host DNA due to its distinct non-nucleosomal viral structure, while Liao et al. leveraged cell-free mitochondrial DNA breakage patterns to diagnose and predict outcomes in sepsis.
However, interpreting these physical signatures requires careful accounting for systemic physiological noise. Because cfDNA captures any major cellular turnover event, acute physical trauma can distort the circulating pool. Henriksen et al. discovered that surgical trauma floods the bloodstream with normal, size-distinct cfDNA fragments for up to four weeks post-operation. Without size fractionation or delayed sampling to filter out this trauma-induced background, the massive influx of healthy, long DNA fragments can physically mask the shorter, more subtle fragmentomic signatures shed by a residual tumor.
Ultimately, fragmentomics succeeds because it treats cell-free DNA not merely as digital sequence data, but as a physical artifact. Every cut, overhang, and length distribution carries a structural memory of the nuclear architecture that once protected it.
Key Foundational Work
High-resolution profiling of human cytomegalovirus cell-free DNA in human plasma highlights its exceptionally fragmented nature (2020) — contributed 2 concept elements to this direction.
Whole genome bisulfite sequencing of cell-free DNA and its cellular contributors uncovers placenta hypomethylated domains (2015) — contributed 2 concept elements to this direction.
The effect of surgical trauma on circulating free DNA levels in cancer patients—implications for studies of circulating tumor DNA (2020) — contributed 1 concept elements to this direction.