Table 2.
Selected studies utilizing different fragmentomics approaches
| Study | Fragment feature | Application | Technology | Key findings and limitations |
|---|---|---|---|---|
| Mouliere et al. 201854 | Fragment size difference between cancer and healthy on a global level, 10 bp oscillations on subnucleosomal level | Cancer diagnosis (pan tumor) | Shallow WGS, WES, size selection enrichment | AUC 0.91–0.99 depending on cancer type. Low sequencing depth (0.4x). Sensitive at low MAFs (after size specific enrichment. Only late-stage cancers |
| Cristiano et al. 201961 | Fragment size difference between cancer and healthy on a regional level | Cancer diagnosis (pan-tumor) | Delfi (WGS) | AUC 0.94 (Sens 57–99%, Spec 98%) supervised model, non-age matched controls |
| Ulz et al. 201959 | Coverage at TFBS and TSS | Prostate cancer subtyping and early detection | WGS | High tumor fraction required |
| Snyder et al. 201665 | Fragment endpoints, coverage near TFBS and near TSS | Cell of origin | Windowed protection score – WPS (WGS inc single strand) | Single strand sequencing enriched shorter fragments. Small sample numbers, high sequencing depth required (∼100x). |
| Esfahani et al. 202270 | Fragment length diversity at promoter regions ‘promoter fragment entropy’, coverage at regions near TSS (“nucleosome depleted regions”) | Diagnosis and subtype classification (lung cancer, diffuse large B cell lymphoma) | EPIQ-Seq (Targeted sequencing, also used WGS/WES) | Composite model of PFE/NDR, Lung cancer from healthy AUC 0.91 (training), 0.83 (validation), NSCLC subtype AUC 0.9, DLBCL from healthy AUC 0.92 (training) AUC 0.96 (validation). Requires disease specific panels for EPIC-seq, less sensitive at early stage |
| De Sarkar et al. 202366 | Coverage at TFBS and TSS, nucleosome phasing (periodicity of nucleosome positions) | Prostate cancer phenotyping | Keraon, ctdPheno (WGS) | AUC 0.96 (90.4% sensitivity, 97.5% specificity) for phenotyping. Lower limit of 8% and 3% tumor fraction required. (ctDPheno/Keraon respectively) |
| Sun et al. 201957 | Strand orientation | HCC diagnosis and tissue of origin | Orientation-aware cfDNA fragmentation- OCF (WGS) | 67% sensitivity, 93.8% specificity for HCC. Lower tissue fraction required than some approaches e.g., ∼5%. Based on known open chromatin regions with limited independent validation. |
| Jiang et al. 202058 | End-motif frequency | Cancer diagnosis (mainly HCC) | WGS, WGBS | AUC 0.86 for HCC. Accurate at 4% tumor fraction. Requires deep sequencing for accuracy. Limited independent validation. |
| Doebley et al. 202271 | TFBS coverage by fragment midpoint | Cancer detection (pan-tumor), breast cancer subtyping. | Griffin (WGS) | Ultra-low-pass WGS (0.1x). Cancer vs. non-cancer: AUC 0.89 for 0.1x coverage. AUC 0.92 for breast cancer subtyping. Mainly existing cohorts, limited independent validation. |
WGS, whole-genome sequencing; WES, whole exome sequencing; WGBS, whole genome bisulfite sequencing; TSS, transcription start site; TFBS, transcription factor binding site; HCC, hepatocellular carcinoma; NSCLC, non-small cell lung cancer; VAF, variant allele frequency; PFE, promoter fragment entropy; NDR, nucleosome depleted regions; AUC, area under the curve.