Draft:Fragmentomics

Fragment size distribution

Fragment size is highly dependent on the placement of histones on the DNA of cells undergoing apoptosis. During programmed cell death, DNA is degraded by caspase-activated DNAases.^[14] Nucleosomes protect DNA from digestion by DNAases, leading to cuts exclusively occurring at exposed sites between nucleosomes.^[3] This non-random, caspase-dependent cleavage is reflected in the fragment size distribution of cfDNA; the median size of cfDNA fragments is ~167 bp, representing the length of DNA wrapped around a single nucleosome.^[2][4][5] A smaller peak in the distribution is observed at ~320 bp, representing a dinucleosomal unit.^[15] Thus, fragment size distribution is a reflection of chromatin structure.^{[citation needed]}

Given the consistency in distribution, fragment size can be used to classify fragment type. Tumour-derived fragments tend to be shorter overall compared to normal cfDNA, with enrichment for fragment sizes between 90 and 150 bp.^[4] The ratio of short-to-long fragments can serve as a feature for early cancer detection and disease monitoring.^[16][17] Similarly, fetal cfDNA in maternal plasma is shorter than maternal cfDNA.^[18][19] The majority of cffDNA fragments are shorter than 300 bp, while roughly 20% of maternal cfDNA fragments are longer than 300 bp.^[18][19][20] Separating fragments by size and extracting those under 300 bp can help enrich for cffDNA.^[19]

End motifs

cfDNA fragments have been found to terminate at preferred nucleotide contexts, with particular tetranucleotide sequences being highly enriched at the ends of fragments.^[21] These end motifs are shaped by the nucleases responsible for cleavage, such as DFFB, DNASE1L3, and DNASE1.^[22][23][24] Plasma DNA fragment end motif frequencies have been shown to be affected by disease. For example, in a study of patients with hepatocellular carcinoma (HCC), patients with HCC had lower abundance of the CCCA fragment motif compared to healthy patients.^[25] Nuclease dysregulation can also cause changes in end motif frequency. Deletion of Dnase1l3 in mice led to the preferential cleavage of fragments, with the top six motifs all beginning with CC.^[26] The range of unique fragment motifs in an individual or population can be quantified by a motif diversity score.^[25] End motif profiles show promising utility as tissue-of-origin classifiers or as markers of disease.^{[citation needed]}

While some plasma DNA fragments are cleaved with blunt ends, other nucleases can produce jagged ends.^[21] Jagged ends refer to the asymmetry between the 5′ and 3′ termini on opposing strands where the cuts on DNA strands are often slightly offset, leading to overhangs.^[21] The degree and directionality of this jaggedness encodes information about the specific enzymatic mechanism of cleavage, with DNASE1 and DNASE1L3 leading to an increase and decrease in the frequency of jagged edges, respectively.^[27] Jagged ends have also been found to be more frequent in fetal DNA in maternal cfDNA and ctDNA in cancer patients.^[27]

Nucleosome positioning

cfDNA fragmentation patterns can reveal underlying nucleosome occupancy in the cell-of-origin. Nucleosome-protected DNA is resistant to cleavage by nucleases, so cuts primarily occur in nucleosome-depleted regions (NDR).^[5][21] Consequently, the ends of cfDNA fragments are strongly biased to fall within regions of linker DNA. Accessible genomic regions are areas of open chromatin that frequently contain actively bound transcription factor binding sites.^[28] By clustering fragments to specific genomic positions, the location of NDRs can be determined and quantified as a windowed protection score (WPS), which is the number of DNA fragments completely spanning a 120 bp genomic window minus those with endpoints inside it.^[5] Nucleosome organization can also be inferred from sequencing metrics, with a loss of sequencing coverage reflecting DNA degradation at unprotected binding sites and peaks of coverage occurring around nucleosome-protected regions.^[28]Nucleosome maps constructed from fragmentation patterns reflect chromatin state in the cell-of-origin and can reveal sites of active transcription.^[29] In oncology, nucleosome positioning can be leveraged for early screening and non-invasive cancer subtyping with ctDNA.^[30][31] From fragmentation patterns, tissue-of-origin can be determined and expression profiles can be characterized.^{[citation needed]}

Standard workflow

Fragmentomics analyses begin with peripheral blood draw, followed by the isolation of blood plasma and/or serum to obtain circulating cfDNA. Isolated cfDNA fragments then undergo next-generation sequencing (NGS). Sequencing data are then processed through bioinformatic pipelines to extract fragmentomic features and/or DNA methylation signals. This can then be used for tissue-of-origin deconvolution, and additional downstream analyses involving statistical and machine learning approaches.^{[citation needed]}

EPIC-seq

EPIC-seq (Epigenetic Expression Inference from Cell-free DNA Sequencing) is a fragmentomics-based method that infers gene expression from cfDNA by analyzing fragmentation patterns at transcription start sites (TSS).^[32] Since active promoters are less protected by nucleosomes, they exhibit more irregular cleavage patterns, whereas inactive promoters remain more protected. In EPIC-seq, cfDNA is extracted from plasma, libraries are prepared, and custom capture panels are enriched for TSS regions of interest, followed by deep sequencing.^[32] Machine learning models then analyze nucleosome positioning and fragmentation features at each TSS to predict gene expression levels, enabling non-invasive reconstruction of transcriptional activity from plasma cfDNA.^{[citation needed]}

cfMeDIP-seq

cfMeDIP-seq (cell-free methylated DNA immunoprecipitation and high-throughput sequencing) is a fragment-based method used to profile DNA methylation patterns from cfDNA.^[33] DNA methylation, the addition of a methyl group to the 5′ position of cytosine residue (5mC), is a key epigenetic mechanism regulating gene expression, and its patterns are tissue- and cell-type specific. Changes in DNA methylation also reflect development, aging, environmental exposures, and disease states. In cfMeDIP-seq, cfDNA is first extracted from plasma and quantified despite its low input and fragment size (~167 bp due to nucleosome protection). In this assay, a 5mC antibody binds to the methylated fragments, enriching for methylated regions across the genome.^[33] Fragments are quantified using spike-in and filler DNA.^[33] By selectively enriching methylated cfDNA fragments without harsh treatments (bisulfite), cfMeDIP-seq overcomes challenges related to low abundance and degradation, enabling sensitive genome-wide methylation profiling.^{[citation needed]}

EM-seq

EM-seq (Enzymatic methyl sequencing) also profiles cfDNA methylation though enzymatic conversion.^[34] The method preserves cfDNA fragment length from degradation and provides high resolution information from low input amounts. In EM-seq, the 5mC and 5hmC is protected by TET family enzymes, then unmethylated cytosines are deaminated, thereby being read-out as thymines during sequencing.^[34] Enzymatic conversion supports both methylation profiling and downstream fragmentomic analyses that link methylation state with nucleosome positioning and tissue-specific signatures.^{[citation needed]}

Tissue-of-origin detection

cfDNA fragments carry signatures reflecting their tissue-of-origin. The tissue and cell-type contributions of circulating cfDNA fragments can be computationally inferred in a process called cfDNA deconvolution.^[36][37] Fragmentomic features can be quantified by calculating statistical scores at defined genomic loci for predictable characteristics such as fragment size distributions, fragment end motifs, strand orientation, fragment convergence patterns, and nucleosomal positioning around TSSs.^[37] These fragmentation signatures are then used to to reconstruct chromatin accessibility landscapes and computationally correlated with reference epigenomic maps and cell atlases to rank candidate tissues and cells-of-origin.^[37] While DNA methylation markers and single nucleotide polymorphisms (SNPs) are commonly used for cfDNA deconvolution, fragmentation-based markers can also provide an effective framework, particularly when high-depth sequencing is available. ^{[citation needed]}

Cancer detection and tumour profiling

Liquid biopsies are increasingly being used for cancer profiling, where tumor-derived DNA is analyzed to detect somatic variants, monitor disease progression, assess minimal residual disease, and guide targeted therapy decisions. Plasma DNA fragments from tumour cells tend to be shorter and have different end motif profiles compared to cell-free DNA from normal cells, improving the identification of tumour DNA.^[9] Fragmentomics has been applied to several areas of cancer care, including early detection,^[17][13] monitoring,^[5] and subtyping.^[28][30]

Non-invasive prenatal testing

In prenatal medicine, cfDNA analysis is widely used for fetal sex determination and non-invasive prenatal testing (NIPT), including parent-of-origin and haplotype-based analyses using paternal SNPs. However, the availability of fetal cfDNA in maternal cfDNA is low, comprising only 2–19% of the total volume^[19]. Since all cffDNA is less than 300 bp long,^[19] selecting for fragments by size can help enrich for fetal DNA. ^{[citation needed]}

Transplant rejection detection

While organ transplants are often life saving procedures, their efficacy and success are threatened by the possibility of rejection. Identifying rejection responses early is critical to save the graft, prevent irreversible damage, and provide treatment in a timely manner. Donor-derived cell-free DNA (ddcfDNA) has been developed as a biomarker of graft integrity, and fragmentomics can be applied to detect transplant rejection. A previous study found that patients who experience acute rejection had higher ddcfDNA concentrations and showed enrichment for fragments between 100 and 250 bp.^[38] Using these features of graft-derived DNA, patients can be easily monitored, allowing transplant rejection to be caught early. ^{[citation needed]}

Biomarker discovery

cfDNA is increasingly investigated as a biomarker in neurodegenerative diseases such as Alzheimer's disease and Parkinson's disease, as well as in autoimmune, cardiovascular, and inflammatory conditions, where tissue-specific methylation patterns can reveal the cell of origin.^[7] In obstetrics, placental cfDNA dynamics are being studied in pregnancy complications such as preeclampsia.^[39][40] Moreover, differential fragmentation profiles and epigenetic signatures in cfDNA are being leveraged to identify therapeutic targets, biomarkers, and to better comprehend underlying disease mechanisms. ^{[citation needed]}