Table 2 Exemplary approaches harnessing fragmentomics, potential applications and limitations

Windowed protection score (WPS)

Whole-genome sequencing to generate maps of genome-wide nucleosome occupancy; WPS is calculated by the number of DNA fragments completely spanning a 120 bp window centered at given genomic coordinate, minus number of fragments with an endpoint within that same window

Use of nucleosome footprints can infer cell types contributing to cfDNA; use of short cfDNA fragments to footprint TFs

Nucleosome maps are heterogeneous, comprising signals of all cell types that give rise to cfDNA; profiled only a small number of ubiquitous TFs; small size of reference dataset of cell lines and tissues against which these samples were compared

[153]

Fragment coverage

Whole-genome sequencing of plasma DNA identified two discrete regions at TSS (NDR and 2K region) where nucleosome occupancy results in different read depth coverage patterns for expressed and silent genes

Classification of expressed cancer driver genes; Determination of expressed isoform of genes with several TSSs

High tumor fraction in plasma required; Gene expression prediction limited to amplified regions; binary classification of genes, i.e. expressed vs. non-expressed

[94]

Fragment coverage

Establishment of nucleosome occupancy maps at TFBSs via whole-genome sequencing; Calculation of accessibility scores as a measure of strength of nucleosome phasing at binding sites of a TF, reflecting strength of TF binding

Identification of lineage-specific TFs and profiling of individual TFs from cfDNA; Identification of patient-specific and tumor-specific patterns, including prediction of tumor subtypes in prostate cancer; detection of early-stage colorectal carcinomas

TF nucleosome interaction maps are heterogeneous, comprising signals of all cell types that give rise to cfDNA; use of all 504 TFs in logistic regression model does not make strategy specific for colon cancer; further work required to identify distinct TFs subsets specific for different tumor types

[122]

Incorporation of additional information on DNA fragment lengths and tumor allelic fraction of mutations to enhance the accuracy of ctDNA detection

Analysis of DNA fragment sizes in plasma cfDNA from melanoma patients demonstrated that mutant fragments were shorter than wild-type fragments at the mononucleosome and dinucleosome peaks; Assessed frequency of mutations for any given fragment size and then weighted each mutant read observed with the probability that it came from the cancer distribution as opposed to the wild-type size distribution

Personalized cancer monitoring

Not suitable for early detection or diagnosis of new cancers, as it requires evaluation of signals across a patient-specific list of mutations; only applied to limited number of cases; lack of validation in a larger cohort

[88]

Filtration of CHIP-associated variants according to fragment size

Demonstrated that cfDNA molecules bearing CH-derived variants tend to be longer than those bearing tumor-derived variants, which can be leveraged to improve detection sensitivity of ctDNA

Early-stage lung cancer detection

Larger cohort needed to fully establish performance characteristics of Lung-CLiP; majority of cases were incidentally diagnosed lung cancers and not identified by LDCT screening;

cohort mainly composed of smokers and thus need to assess performance in non-smokers

[27]

Global fragmentation patterns

Evaluation of size distribution and frequency of millions of naturally occurring cfDNA fragments across genome

Non-invasive early detection/prescreening high-risk populations for lung cancer; Use of genome-wide fragmentation profiles across ASCL1 TFBSs to distinguish individuals with SCLC from those with NSCLC

Majority of patients in LUCAS cohort presented symptoms not fully representative of a screening population; lack of large prospective validation in a screening population; Several patients with late-stage disease not detected by DELFI

[29]

Orientation-aware cfDNA fragmentation (OCF)

Assessment of differences in read densities of sequences corresponding to orientation of upstream and downstream ends of cfDNA molecules in relation to reference genome; quantitative analyses of signals to measure relative contributions of various tissues to plasma DNA pool

Noninvasive prenatal testing, organ transplantation monitoring, cancer liquid biopsy

Only small sample size investigated; method based on open chromatin profiles, of which availability was limited at the time of the study

[145]

Preferred end coordinates

Determined that particular genome coordinates had an increased probability of being an ending position for plasma DNA fragment and whether such ends exhibit differences depending on their tissue of origin (i.e., from placenta or mother or from patients with HCC)

Noninvasive fetal whole-genome analysis; diagnosis of early-stage HCC; tissue-of-origin for organ transplant recipients

Determination of preferred-end sites requires high coverage sequencing; lack of large prospective validation cohort

[163, 164]

DNA end motif

Demonstrated that plasma DNA ends show prevalence of certain nucleotide contexts, i.e., preferred fragment end motifs, which represent a distinct type of fragmentation signature. The motifs are defined as a few nucleotides at plasma DNA ends regardless of the site of origin within the genome

End motifs may serve as class of biomarkers for liquid biopsy in oncology, noninvasive prenatal testing, and transplantation monitoring

Small sample size; lack of large-scale validation study

[165]

Jagged ends

Detection of double-stranded plasma molecules carrying single-stranded protruding ends, termed jagged end; Assessment of jaggedness across varying plasma DNA fragment sizes and association with nucleosomal patterns

Fragmentomics-based molecular diagnostics in noninvasive prenatal testing, organ transplantation, oncology, and autoimmune diseases

Lack of large-scale validation study

[166]

Global and regional fragment size distribution, fragment coverage (LIQUORICE)

Combination of several fragmentation-based metrics into an integrated machine learning classifier; Analysis of global fragment size distribution, region fragment size distribution as well as fragment coverage at regions of interest

Use of cfDNA fragmentation patterns as prognostic biomarkers in Ewing sarcoma

Lack of standard reference markers for ctDNA quantification makes calculation of definitive performance metrics for machine learning classifier difficult; investigation of rare sarcomas, which limited size of cohort; retrospective analysis lack of validation in large, prospective cohort

[167]