CIRCULATING CELL-FREE DNA FOR EARLY CANCER DETECTION: A SYSTEMATIC REVIEW OF MUTATION, METHYLATION, AND FRAGMENTOMIC SIGNATURES

Abstract

Background: Circulating cell-free DNA (cfDNA) has developed from a nonspecific marker of cellular injury into a multidimensional source of molecular information for cancer detection. Tumor-associated cfDNA can differ from background circulating DNA through somatic sequence variants, abnormal methylation, fragment length distributions, nucleosome-associated cleavage patterns, and characteristic fragment ends. These molecular signatures may permit cancer detection before overt clinical presentation. Objective: To systematically evaluate mutation-, methylation-, and fragmentomics-based cfDNA signatures for early cancer detection, emphasizing diagnostic performance, preclinical detection, stage dependence, tissue-of origin prediction, and evidence from prospective asymptomatic populations. Methods: A structured literature search of PubMed/MEDLINE and multidisciplinary academic databases was performed for studies available through July 2026. A total of 1,318 records were identified through database and supplementary searching. After removal of 373 duplicate or otherwise ineligible records, 945 records underwent title and abstract screening. Of these, 767 records were excluded, leaving 178 reports sought for retrieval. Seven reports could not be retrieved, and 171 full-text reports were assessed for eligibility. Following exclusion of 146 reports, 25 primary studies were included in the qualitative synthesis. Search concepts included cell-free DNA, circulating tumor DNA, mutation, methylation, fragmentomics, fragment ends, nucleosome footprints, early cancer detection, and multi-cancer detection. Methodological limitations were evaluated according to QUADAS-3 domains. Owing to substantial heterogeneity, quantitative meta-analysis was not performed. Results: Twenty-five primary studies were included. Mutation analysis demonstrated that ctDNA may precede clinical cancer diagnosis by several years. In a 2025 prospective biobank analysis, cancer-associated mutations were identified 3.1–3.5 years before clinical diagnosis in four of six evaluable participants, although mutant allele fractions were 8.6–79-fold lower than nearer diagnosis. Methylation provided a broader signal space: a 2024 gastrointestinal cancer study evaluating 407 plasma samples achieved 81.3% sensitivity in independent validation, while tumor-origin prediction reached 95.8% for gastric cancer and 93.3% for colorectal cancer. A 2024 fragment-end classifier achieved an AUC of 0.95 with 85.1% sensitivity at 95% specificity. In 2025, multidimensional fragmentomics demonstrated 87.4% sensitivity, 97.8% specificity, and 82.4% tissue-of-origin accuracy in independent validation; however, sensitivity declined to 53.5% in a prospective cohort of 3,724 asymptomatic participants. A prospective multimodal cfDNA study involving 9,024 eligible asymptomatic individuals reported 70.83% sensitivity, 99.71% specificity, 39.53% positive predictive value, and 99.92% negative predictive value. Conclusion: Mutation, methylation, and fragmentomic signatures capture complementary dimensions of tumor biology. Mutation analysis offers high sequence specificity but is constrained by extremely low ctDNA abundance in early disease. Methylation provides a large, tissue-informative target space, whereas fragmentomics exploits genome-wide structural information. Prospective population-based validation remains essential before widespread screening implementation.