Fragmentomics

The density profile of cfDNA molecule sizes, typically peaking at ~166 bp (mono-nucleosome) with a 10-bp periodicity. Tumor-derived ctDNA is often more fragmented, showing enrichment in shorter fragments (<150 bp) compared to non-cancerous cfDNA. This shift is a foundational feature for cancer detection algorithms.

• Peak analysis: Distance between local maxima in the density curve
• Short-to-long ratio: Proportion of fragments below vs. above 150 bp
• Modal shift: Deviation of the primary peak from the canonical 166 bp position

The nucleotide sequence preferences at the 5' and 3' termini of cfDNA fragments. Cleavage is not uniform; specific 4-mer end motifs (e.g., CCCA, CCTG) are preferentially generated by the nucleases active in the tissue of origin. These motifs serve as a highly specific epigenetic fingerprint for tissue-of-origin deconvolution.

• Motif frequency vectors: Counts of all 256 possible 4-base end sequences
• Diversity score: Shannon entropy of the end motif distribution
• Coordinate-specific bias: Local sequence preference around the cleavage site

The mapping of cfDNA fragment endpoints to reference genome coordinates reveals the precise positions of nucleosome-protected regions. Active gene promoters and enhancers exhibit nucleosome-depleted regions (NDRs) with high fragment endpoint density. The spacing between nucleosomes and the depth of protection valleys provide a direct readout of chromatin accessibility in the cell of origin.

• Windowed protection score (WPS): A metric quantifying local nucleosome occupancy
• Transcription start site (TSS) coverage: Fragment density profile around gene promoters
• Phasogram: Autocorrelation of fragment endpoints revealing nucleosome repeat length

Integrating fragment length with copy number alteration (CNA) status corrects for the confounding effect of tumor aneuploidy on the fragmentome. A region with a copy number gain will produce more fragments, but the fragmentation profile of those fragments (e.g., their size distribution) can differ from normal regions, revealing the altered chromatin compaction of the amplified locus.

• CNA-stratified size ratios: Fragment length distributions computed separately for gained, lost, and neutral regions
• Allele-specific fragmentation: Examining fragment patterns on the maternal vs. paternal haplotype

The relative orientation of overlapping paired-end reads. A fragment where the two reads overlap with a 'jagged' (staggered) end, rather than a blunt end, indicates a specific cleavage pattern. The ratio of jagged to blunt ends is a stable, non-random feature that correlates with the nuclease environment (e.g., DNASE1L3 vs. DFFB activity) and is perturbed in cancer and autoimmune disease.

• Jaggedness index: Proportion of overlapping read pairs with non-blunt ends
• Overhang length distribution: The size of the single-stranded overhang in jagged fragments
• Strand-specific orientation bias: The 5' to 3' directionality of fragment endpoints relative to the reference

A measure of fragmentation randomness computed across sliding windows of the genome. Healthy tissues produce a highly structured, reproducible fragmentation pattern. In cancer, the disruption of normal chromatin organization and the presence of aberrant nuclease activity increases the local entropy of fragment endpoints, creating a genome-wide signal detectable by machine learning classifiers.

• Shannon entropy: Calculated on fragment endpoint positions in fixed-size bins
• KL divergence: Comparing a sample's entropy profile to a healthy reference
• Multiscale entropy: Entropy computed at multiple genomic resolutions (1 kb, 10 kb, 100 kb)

The characteristic fragmentation pattern of cfDNA reflecting the protection of DNA wrapped around histone octamers. In healthy cells, nucleosomes are regularly spaced, creating a ~167 bp ladder pattern in cfDNA fragment size distributions. In cancer, this footprint is disrupted due to altered chromatin accessibility and nuclease activity, providing information about gene regulatory elements and cell-type origin. The depth of coverage at transcription start sites directly correlates with gene expression levels.

The specific nucleotide sequences at the termini of cfDNA fragments, reflecting the cleavage preferences of the nucleases responsible for DNA degradation. Key characteristics include:

• CCCA motif: Enriched in fetal-derived cfDNA
• CCN trinucleotides: Associated with DNASE1L3 activity
• Diversity Score: Measures the entropy of end motifs; reduced diversity is a hallmark of cancer
• 4-mer end profiles: 256 possible motifs used to infer tissue of origin These motifs serve as orthogonal biomarkers independent of genetic mutations.

The probability density function of cfDNA fragment lengths, typically ranging from 50 to 350 bp. Key features include:

• Mononucleosomal peak: Dominant peak at ~167 bp from apoptotic chromatin digestion
• Dinucleosomal peak: Secondary peak at ~334 bp
• Short fragment enrichment: Tumor-derived ctDNA is often 20-50 bp shorter than non-tumor cfDNA
• Size selection: In vitro enrichment for sub-150 bp fragments improves variant detection sensitivity This distribution is a fundamental biomarker for distinguishing malignant from benign cfDNA.

The genomic coordinates where histone octamers are bound, inferred from the oscillating read depth of cfDNA sequencing data. Well-positioned nucleosomes create phased coverage peaks across gene bodies and regulatory elements. Key applications:

• Transcription factor footprinting: Nucleosome-depleted regions indicate active promoters
• Chromatin accessibility inference: Correlates with ATAC-seq and DNase-seq profiles
• Tissue-of-origin deconvolution: Different cell types exhibit distinct nucleosome architectures
• Cancer detection: Disorganized positioning indicates aberrant chromatin remodeling

Genomic coordinates where cfDNA fragments terminate with high frequency, representing hotspots of nuclease cleavage. These positions are non-random and correlate with chromatin accessibility and transcription factor binding. Key concepts:

• Window Protection Score (WPS) : Measures nucleosome occupancy by quantifying fragment endpoints
• Occurrence peaks: Clusters of fragment ends at open chromatin regions
• Cancer-specific ends: Aberrant preferred end profiles distinguish tumor-derived cfDNA
• Multi-feature integration: Combined with end motifs and size for maximum classification power

Single-stranded overhangs at cfDNA fragment termini, reflecting the specific cleavage mechanism of the nuclease that generated the fragment. Key characteristics:

• Blunt ends: Produced by caspase-activated DNase (CAD) during apoptosis
• 5' overhangs: Associated with DNASE1L3 cleavage
• 3' overhangs: Less common, linked to alternative nucleases
• Jaggedness Index: Quantifies the proportion of non-blunt ends; elevated in cancer
• Sequencing detection: Requires specialized library preparation that preserves end structure This feature provides mechanistic insight into the enzymatic processes driving cfDNA release.