{ "accessed": "2026-08-01", "interpretation_rule": "Performance values are reported in each study's own population and operating point. They are not directly comparable and are not estimates of this prototype's clinical performance.", "studies": [ { "id": "snyder-2016", "year": 2016, "title": "Cell-free DNA comprises an in vivo nucleosome footprint that informs its tissues-of-origin", "method": "Windowed protection score and nucleosome footprints", "evidence": "Mechanistic plasma-WGS study", "cohort_depth": "Public plasma WGS with heterogeneous approximately 1x to deep coverage", "key_result": "Established that cfDNA coverage around nucleosomes reflects chromatin and cell-of-origin biology.", "critical_read": "Foundational feature biology, not an MCED validation study; very few independent healthy donors.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC4715266/" }, { "id": "ulz-tss-2016", "year": 2016, "title": "Inferring expressed genes by whole-genome sequencing of plasma DNA", "method": "TSS nucleosome occupancy", "evidence": "Mechanistic method", "cohort_depth": "Plasma WGS", "key_result": "Linked promoter coverage patterns to gene expression and tissue contributions.", "critical_read": "Supports a chromatin expert; sparse low-pass data require aggregation over predefined programs.", "url": "https://www.nature.com/articles/ng.3648" }, { "id": "ichorcna-2017", "year": 2017, "title": "Scalable whole-exome sequencing of cell-free DNA reveals high concordance with metastatic tumors", "method": "ichorCNA ultra-low-pass copy number and tumor fraction", "evidence": "Large metastatic analytical cohort", "cohort_depth": "1,439 samples from 520 metastatic breast/prostate patients; approximately 0.1x ULP-WGS branch", "key_result": "Favorable arm-level CNA profiles were detectable around a modeled 3% tumor fraction.", "critical_read": "The 3% figure is not a clinical early-cancer detection limit and CNA-neutral disease will be missed.", "url": "https://www.nature.com/articles/s41467-017-00965-y" }, { "id": "mouliere-2018", "year": 2018, "title": "Enhanced detection of circulating tumor DNA by fragment size analysis", "method": "Global fragment size, size selection and CNA", "evidence": "Retrospective multi-cancer analytical study", "cohort_depth": "344 plasma samples from 200 cancer patients; low-pass analysis around 0.4x", "key_result": "Short-fragment selection often enriched ctDNA more than twofold and size complemented CNA.", "critical_read": "Mostly advanced disease; physical size selection discards molecules and can introduce bias.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6483061/" }, { "id": "sun-ocf-2019", "year": 2019, "title": "Orientation-aware plasma cell-free DNA fragmentation analysis in open chromatin regions informs tissue of origin", "method": "Orientation-aware fragmentation", "evidence": "Mechanistic method", "cohort_depth": "Plasma WGS", "key_result": "Strand-specific endpoints around open chromatin carry tissue information.", "critical_read": "Requires pinned genome build, strand policy and cell/tissue atlas; tissue-origin must be evaluated conditionally after detection.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6396422/" }, { "id": "delfi-2019", "year": 2019, "title": "Genome-wide cell-free DNA fragmentation in patients with cancer", "method": "DELFI regional short/long fragmentation and coverage", "evidence": "Foundational retrospective case-control MCED study", "cohort_depth": "Classifier: 208 cancers, 215 healthy controls, seven cancer types, 1-2x WGS and 504 5-Mb windows", "key_result": "AUC 0.94; 73% sensitivity at 98% specificity; reported stage I/II/III sensitivities 68%/72%/79%.", "critical_read": "Internal case-control cross-validation with healthy controls and batch/site risk; not prospective screening performance.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6774252/" }, { "id": "jiang-endmotif-2020", "year": 2020, "title": "Plasma DNA end-motif profiling as a fragmentomic marker in cancer, pregnancy, and transplantation", "method": "Four-base endpoint motif diversity", "evidence": "Mechanistic and liver-cancer case-control study", "cohort_depth": "Plasma WGS", "key_result": "Showed disease-associated changes in endpoint motifs and motif diversity.", "critical_read": "End repair, library chemistry, sequencer and reference context can dominate; assay lock is mandatory.", "url": "https://aacrjournals.org/cancerdiscovery/article/10/5/664/2457/Plasma-DNA-End-Motif-Profiling-as-a-Fragmentomic" }, { "id": "liquorice-2021", "year": 2021, "title": "The LIQUORICE algorithm identifies tissue-specific cell-free DNA signatures in cancer patients", "method": "Regulatory-region coverage and fragment length", "evidence": "Retrospective method study", "cohort_depth": "263 samples, including sarcomas and 22 healthy participants; median 12x with computational 1x/0.1x downsampling", "key_result": "Cancer-versus-healthy AUC reached 0.97 in selected analyses.", "critical_read": "Deep discovery, pediatric/high-shedding biology and in-silico downsampling limit ordinary lpWGS MCED interpretation.", "url": "https://www.nature.com/articles/s41467-021-23445-w" }, { "id": "lucas-2021", "year": 2021, "title": "Detection and characterization of lung cancer using cell-free DNA fragmentomes", "method": "DELFI lung classifier with clinical features", "evidence": "Development plus external retrospective cohort", "cohort_depth": "129 lung cancers and 158 controls; external 46 cancers and 385 controls; 1-2x", "key_result": "Cross-validation AUC 0.90 and stage-I AUC 0.76; combined model reached 94% sensitivity at 80% specificity.", "critical_read": "Symptomatic/imaging-positive context and source differences between cases and controls do not represent population MCED.", "url": "https://www.nature.com/articles/s41467-021-24994-w" }, { "id": "griffin-2022", "year": 2022, "title": "A framework for clinical cancer subtyping from nucleosome profiling of cell-free DNA", "method": "Griffin length-specific GC correction and TFBS profiles", "evidence": "Cross-cohort reanalysis method", "cohort_depth": "DELFI/LUCAS and other WGS cohorts; tested down to 0.1x", "key_result": "DELFI AUC 0.94 and fixed external LUCAS AUC 0.86 in reported analyses.", "critical_read": "Explicit DELFI-LUCAS batch effects prevented direct transfer and performance worsened at 0.1x.", "url": "https://www.nature.com/articles/s41467-022-35076-w" }, { "id": "crag-2022", "year": 2022, "title": "CRAG: de novo characterization of cell-free DNA fragmentation hotspots in plasma whole-genome sequencing", "method": "Endpoint fragmentation hotspots", "evidence": "Small early-stage multi-cancer proof of concept", "cohort_depth": "Open processed coordinates for 66 matched HCC/breast cancer and control subjects", "key_result": "Identified open-chromatin/transposon-associated hotspots with cancer and tissue-origin signal.", "critical_read": "Small matched cohorts can validate software directionality, not population specificity.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9733064/" }, { "id": "pointy-2022", "year": 2022, "title": "Genome-wide mutational signatures in low-coverage whole genome sequencing of cell-free DNA", "method": "Singleton mutational spectra (Pointy) with optional ichorCNA", "evidence": "Small internal repeated nested-CV classifier plus larger signature analyses", "cohort_depth": "Classifier cohort only 16 stage-IV CRC and 19 controls; larger DELFI reanalysis was used mainly for signature detection", "key_result": "Repeated nested CV in the 35-subject classifier cohort gave median AUC 0.96 across models, random forest 0.98 and random forest plus ichorCNA 1.00.", "critical_read": "These are not external results; the tiny advanced-disease cohort and sequencing-error, germline, age, smoking and platform sensitivity make this a high-control exploratory branch.", "url": "https://www.nature.com/articles/s41467-022-32598-1" }, { "id": "walker-foresee-2022", "year": 2022, "title": "Hydroxymethylation profile of cell free DNA is a biomarker for early colorectal cancer", "method": "Regional fragmentation from ordinary input-control cfDNA WGS", "evidence": "Same-study held-out CRC validation with open processed matrices", "cohort_depth": "Released fragment matrices: 405 CRC plus 296 noncancers for training; 220 CRC plus 164 noncancers for validation", "key_result": "The paper reported input-WGS fragmentation-only AUC about 0.83 and 62% sensitivity at 95% specificity in its held-out internal validation.", "critical_read": "CRC only, post hoc same-study validation rather than independent-site MCED; technical/vendor structure and the released vector's exact channel identity require further audit.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9532421/" }, { "id": "liver-delfi-2023", "year": 2023, "title": "Detecting liver cancer using cell-free DNA fragmentomes", "method": "DELFI liver classifier", "evidence": "Multicohort retrospective validation", "cohort_depth": "724 total participants in US/EU and Hong Kong cohorts, about 2.6x", "key_result": "The study reported 88% sensitivity at 98% specificity versus average-risk controls and 85% at 80% specificity among high-risk controls; an independent Hong Kong cohort was also included.", "critical_read": "Those two operating points summarize study/development contexts rather than both being independent external estimates; the change against high-risk controls shows why disease-context controls are essential.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9975663/" }, { "id": "wang-lung-2023", "year": 2023, "title": "Multidimensional analysis of cell-free DNA fragmentomics for lung cancer detection", "method": "Multimodal score fusion of CNA, size, end and breakpoint-motif features (not gene-fusion detection)", "evidence": "Multicohort retrospective study", "cohort_depth": "432 lung cancers and 350 noncancers; original mean 11.08x, development around 5x, computationally downsampled to 0.5x", "key_result": "Reported cohort AUCs 0.974-0.987 and stage-I sensitivity 83.2%.", "critical_read": "The 0.5x claim relies on in-silico downsampling; independently prepared depth, LoD and repeatability were not established.", "url": "https://academic.oup.com/ajrccm/article/207/9/1203/8428147" }, { "id": "acid-2024", "year": 2024, "title": "Language model enables end-to-end accurate detection of cancer from cell-free DNA", "method": "ACID causal sequence model over concatenated reads", "evidence": "Assay-mixed retrospective research", "cohort_depth": "2,094 samples across WGS and bisulfite assays; small WGS evaluation", "key_result": "Reported WGS AUROC 0.924 and high accuracy using 10,000 reads.", "critical_read": "Alphabetical read concatenation is arbitrary and not permutation-invariant; reproduce only as a research comparator.", "url": "https://pubmed.ncbi.nlm.nih.gov/38385880/" }, { "id": "artemis-2024", "year": 2024, "title": "Genome-wide repeat landscapes in cancer and cell-free DNA", "method": "ARTEMIS repeat-element fragmentation", "evidence": "Large multi-assay method study", "cohort_depth": "2,837 tissue/plasma samples from 1,975 people; cfDNA generally 1-2x", "key_result": "LUCAS ARTEMIS AUC 0.82 and ARTEMIS plus DELFI 0.91.", "critical_read": "Repeat features can be orthogonal but are sensitive to ancestry, reference build and reused cohorts.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11323656/" }, { "id": "finaleme-2024", "year": 2024, "title": "FinaleMe: inferring regional methylation from cell-free DNA fragmentation", "method": "HMM-derived methylation/tissue probabilities without bisulfite sequencing", "evidence": "Exploratory method", "cohort_depth": "80 paired WGS/WGBS samples plus ULP breast/prostate data; evaluated to 0.1x", "key_result": "Inferred regional methylation probabilities from fragment length, coverage and CpG context.", "critical_read": "Sparse and prior-dependent at low depth; use only as an exploratory tissue-origin expert.", "url": "https://www.nature.com/articles/s41467-024-47196-6" }, { "id": "fragment-comparison-2024", "year": 2024, "title": "Systematic comparison of cell-free DNA fragmentomic features for cancer detection", "method": "Ten-pattern, cross-source comparison", "evidence": "Cross-dataset benchmarking study", "cohort_depth": "Four data sources and eight cancers", "key_result": "End motifs were strong internally while length and coverage transferred more consistently.", "critical_read": "Dataset origin substantially changed performance, arguing for site-held-out evaluation and late fusion.", "url": "https://advanced.onlinelibrary.wiley.com/doi/10.1002/advs.202308243" }, { "id": "illmac-2024", "year": 2024, "title": "Interpretable large language model for cancer detection from cell-free DNA fragmentomics", "method": "LLaMA-7B over serialized end-motif ranks", "evidence": "Assay-mixed retrospective research", "cohort_depth": "2,451 samples overall; WGS subset 129; external WGS 44 cancers and 30 controls", "key_result": "Reported external AUC 0.912 with specificity 0.667 at its high-sensitivity point.", "critical_read": "Overcapacity and weak simple-model comparisons provide no persuasive reason to use a natural-language LLM for a compact numeric vector.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11367762/" }, { "id": "emit-2024", "year": 2024, "title": "End-motif transformer for cancer detection from cell-free DNA", "method": "Compact motif Transformer", "evidence": "Assay-mixed retrospective research", "cohort_depth": "4,606 assay-mixed samples; WGS HCC subset about 129", "key_result": "Reported motif AUC improvement from approximately 0.826 to 0.895.", "critical_read": "Assay mixing and pretraining/evaluation overlap create transductive leakage concerns.", "url": "https://www.nature.com/articles/s41698-024-00635-5" }, { "id": "decidia-2024", "year": 2024, "title": "Transformer-based representation learning and multiple-instance learning for cancer diagnosis from bisulfite-treated plasma cfDNA", "method": "Read-level Transformer plus MIL", "evidence": "Large targeted-bisulfite research cohort", "cohort_depth": "5,389 samples across colorectal, liver, lung and noncancer classes", "key_result": "CRC cross-validation AUROC 0.980 and external HCC AUROC 0.910 were reported.", "critical_read": "Architecture informs a future methylation branch but ordinary lpWGS does not contain base methylation states.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11547222/" }, { "id": "delfi-l101-2024", "year": 2024, "title": "A blood-based test for lung cancer screening using cell-free DNA fragmentomes", "method": "DELFI-L101 lung-screen triage", "evidence": "Prospective multicenter screening-eligible validation", "cohort_depth": "958 participants from 47 centers; held-out 248 cancers and 134 controls; about 3x", "key_result": "Validation sensitivity 84%, specificity 53%, and stage-I sensitivity 71%.", "critical_read": "Useful LDCT-triage evidence, not an MCED-grade high-specificity screen.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11528203/" }, { "id": "dnabert2-2024", "year": 2024, "title": "DNABERT-2: Efficient foundation model and benchmark for multi-species genome", "method": "117M-parameter DNA foundation model", "evidence": "General genome representation model", "cohort_depth": "Pretrained reference sequences, not cfDNA samples", "key_result": "Provides an Apache-2.0 frozen encoder suitable for a controlled fragment-embedding benchmark.", "critical_read": "Reference-sequence grammar is not plasma-fragmentation biology; it must beat k-mer/CNN baselines under identical bags and budgets.", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/b633e7052970b8f5aa1a69164d99e9e8-Abstract-Conference.html" }, { "id": "caduceus-2024", "year": 2024, "title": "Caduceus: Bi-directional equivariant long-range DNA sequence modeling", "method": "Reverse-complement-aware DNA state-space model", "evidence": "General genome representation model", "cohort_depth": "Pretrained genomic sequence", "key_result": "Biologically attractive reverse-complement symmetry and Apache-2.0 code make it a second frozen-encoder comparator.", "critical_read": "It still lacks cfDNA cleavage-boundary pretraining and sample-level validation.", "url": "https://arxiv.org/abs/2403.03234" }, { "id": "nucleotide-transformer-2024", "year": 2024, "title": "Nucleotide Transformer: building and evaluating robust foundation models for human genomics", "method": "Large DNA Transformer", "evidence": "General genome foundation model", "cohort_depth": "Reference genomes, not plasma WGS", "key_result": "Provides strong general sequence embeddings across genomics tasks.", "critical_read": "Large compute and noncommercial/share-alike weight terms make it a research comparator, not the MVP.", "url": "https://www.nature.com/articles/s41592-024-02523-z" }, { "id": "fragle-2025", "year": 2025, "title": "A deep-learning model for quantifying circulating tumour DNA from the density distribution of DNA-fragment lengths", "method": "Fragle fragment-length density model", "evidence": "Cross-cohort method validation", "cohort_depth": "Multiple low-pass WGS cancer and control cohorts", "key_result": "Reported improved tumor-fraction quantification and longitudinal/MRD utility.", "critical_read": "Best treated as an auxiliary tumor-fraction/QC expert rather than a standalone MCED classifier.", "url": "https://www.nature.com/articles/s41551-025-01370-3" }, { "id": "lionheart-2025", "year": 2025, "title": "Cross-dataset pan-cancer detection by correlating cell-free DNA fragment coverage with open chromatin sites across cell types", "method": "LIONHEART cross-dataset chromatin correlations", "evidence": "Leave-one-dataset-out multi-cancer benchmark", "cohort_depth": "Nine datasets, 14 cancers, 1,449 cancers and 1,106 controls; 63.3% at 1-3x", "key_result": "Mean cross-dataset AUC 0.826; mean sensitivity 49.5% at 95% specificity.", "critical_read": "More credible generalization blueprint and materially lower than typical within-study performance.", "url": "https://www.nature.com/articles/s41467-025-66503-3" }, { "id": "mercury-2025", "year": 2025, "title": "Early detection of multiple cancer types using multidimensional cell-free DNA fragmentomics", "method": "Mercury integrated WGS genetic/fragmentomic model", "evidence": "Large case-control plus preliminary prospective asymptomatic study", "cohort_depth": "Internal 3,021 cancers/3,370 controls; independent 677/687; prospective asymptomatic 3,724", "key_result": "Independent sensitivity 87.4%, specificity 97.8%, tissue-origin accuracy 82.4%; prospective sensitivity 53.5%, specificity 98.1%.", "critical_read": "The large case-control-to-screening sensitivity drop is the clearest warning against quoting discovery performance as clinical performance.", "url": "https://www.nature.com/articles/s41591-025-03735-2" }, { "id": "methylbert-2025", "year": 2025, "title": "MethylBERT learns read-level DNA methylation patterns for cancer detection", "method": "Sequence-plus-methylation Transformer", "evidence": "Assay-specific foundation model", "cohort_depth": "Bisulfite methylation reads", "key_result": "Demonstrated read-level representation and tumor-fraction deconvolution for methylation assays.", "critical_read": "Not applicable to ordinary WGS without methylation-preserving chemistry.", "url": "https://doi.org/10.1038/s41467-025-55920-z" }, { "id": "dna-fm-benchmark-2025", "year": 2025, "title": "Benchmarking DNA foundation models for genomic sequence prediction", "method": "Independent DNA foundation-model comparison", "evidence": "Cross-model benchmark", "cohort_depth": "General genomic tasks", "key_result": "Found no uniformly superior DNA foundation model across tasks.", "critical_read": "Requires task-specific scratch baselines and fair compute/data budgets before adopting pretraining.", "url": "https://www.nature.com/articles/s41467-025-65823-8" }, { "id": "tufest-2026", "year": 2026, "title": "Fragmentomic liquid biopsy enables early breast cancer detection, molecular subtyping and lymph node assessment", "method": "TuFEst repeats, motifs, size and TF coverage", "evidence": "Multicenter breast-specific case-control study", "cohort_depth": "503 breast cancers, 289 benign controls; 77.1% stage 0-IIa; 2x from about 1 mL plasma", "key_result": "In the labeled external cohort (95 malignant, 61 benign), AUC was 0.968 (95% CI 0.941-0.995), with 91/95 sensitivity and 53/61 specificity at the reported confusion-matrix point.", "critical_read": "Breast-specific, not MCED; libraries were 2x and the 1x/0.5x/0.1x experiments were computational downsamplings. The abstract's 95%/78.3% is not labeled as that external operating point.", "url": "https://www.nature.com/articles/s41467-026-70204-w" }, { "id": "fragmentia-ai-2026", "year": 2026, "title": "Toward generalizable prediction of cancer signal using a cell-free DNA language model", "method": "Fragmentia-AI fragment encoder and sample-level learning", "evidence": "Peer-reviewed multi-cohort research", "cohort_depth": "Panels and ultra-low-pass WGS at approximately 0.1-1% of conventional sequencing input", "key_result": "Reported cancer-associated fragment-level sequence patterns across assay contexts.", "critical_read": "Directly relevant architecture but proprietary/restricted resources and lack of independent replication limit reproducibility.", "url": "https://www.sciencedirect.com/science/article/pii/S2666379126002831" }, { "id": "leaf1-2026", "year": 2026, "title": "Semantic fragment representations for coordinate-free analysis of genomics data", "method": "LEAF-1 boundary-aware 90.4M fragment foundation model plus MIL", "evidence": "Preprint", "cohort_depth": "Pretrained on about 58 billion ATAC/cfDNA fragments; pan-cancer and held-cancer experiments", "key_result": "Reported pan-cancer AUC around 0.95 and frozen clear-cell renal cancer AUC 0.83.", "critical_read": "Best research benchmark, but July 2026 preprint, possible cohort overlap, small held-cancer test, noncommercial terms and approximately 2 GB embeddings per sample by default.", "url": "https://www.biorxiv.org/content/10.64898/2026.07.09.737627v1" }, { "id": "coverage-bias-2026", "year": 2026, "title": "Fragmentation-induced coverage biases in cell-free DNA sequencing affect the clinical sensitivity of liquid biopsies", "method": "Coverage-bias analysis", "evidence": "Mechanistic methods paper", "cohort_depth": "cfDNA WGS", "key_result": "Shows that epigenetic-context-dependent fragmentation changes locus coverage and effective sensitivity.", "critical_read": "Coverage cannot be treated as uniform Poisson sampling; fragment context and assay preparation alter callability.", "url": "https://www.nature.com/articles/s42003-026-10414-6" }, { "id": "multimodal-wgsm-2026", "year": 2026, "title": "Enhanced multicancer screening assay through whole-genome methylation sequencing-based multimodal cell-free DNA analysis", "method": "Methylation, CNV, fragment-size ratio and distribution ensemble", "evidence": "Multimodal retrospective study", "cohort_depth": "Whole-genome methylation sequencing", "key_result": "Reported sensitivity 93.2% at 95% specificity with strong early-stage results.", "critical_read": "Assay chemistry differs from ordinary lpWGS and 95% specificity is insufficient for many low-prevalence screening settings.", "url": "https://www.nature.com/articles/s12276-026-01674-7" }, { "id": "methylome-compendium-2026", "year": 2026, "title": "A pan-cancer compendium of 1,294 plasma cell-free DNA methylomes and fragmentomes enabling multicancer detection", "method": "cfMeDIP-seq methylome plus fragmentomics", "evidence": "Pan-cancer compendium", "cohort_depth": "1,294 plasma methylomes/fragmentomes", "key_result": "Demonstrated cancer-type heterogeneity across length, motif, nucleosome and methylation features.", "critical_read": "Useful multimodal evidence but immunoprecipitation enriches a different molecular view than ordinary lpWGS.", "url": "https://www.nature.com/articles/s43018-026-01116-3" }, { "id": "nci-status-2026", "year": 2026, "title": "NCI Cancer Screening Overview: Multi-Cancer Detection", "method": "Regulatory and clinical-evidence context", "evidence": "US National Cancer Institute", "cohort_depth": "Not applicable", "key_result": "As of access, no MCD assay is FDA-authorized and no published randomized trial has shown mortality reduction.", "critical_read": "A research classifier must not be marketed or interpreted as a screening test; clinical utility and downstream harms remain separate endpoints.", "url": "https://www.cancer.gov/about-cancer/screening/hp-screening-overview-pdq" } ] }