Cancer-Like Fragmentomic Characteristics of Somatic Variants in Cell-Free DNA

Area of Science:

• Molecular Biology
• Genomics
• Cancer Research

Background:

• Cell-free DNA (cfDNA) in plasma exhibits unique fragmentomic signatures linked to cellular origins.
• Somatic variants within cfDNA retain cancer-associated signals, differentiating cancer patients from controls.

Purpose of the Study:

• To investigate the fragmentomic characteristics of cfDNA harboring somatic variants.
• To develop AI-driven diagnostic models for cancer detection using cfDNA features.

Main Methods:

• Analysis of cfDNA fragmentomics, including size, DNA methylation, and nucleosome structure.
• Development and validation of AI models (FreeSV and FreeSV+) integrating genomic, fragmentomic, and epigenomic data.
• Utilizing a large pan-cancer cohort for model training and evaluation.

Main Results:

• cfDNA with somatic variants shows cancer-like fragmentomic traits even in controls, associated with clonal hematopoiesis.
• These fragmentomic aberrations are amplified in cancer patients, aiding diagnosis.
• The FreeSV+ model, integrating genomewide features, achieved high AUCs (0.93-0.99) across cancer types.

Conclusions:

• Somatic variant-associated fragmentomic aberrations in cfDNA are crucial for cancer detection.
• Integrative analysis of genomic and fragmentomic cfDNA data enhances AI-based cancer liquid biopsy accuracy.
• AI models like FreeSV+ demonstrate significant potential for non-invasive cancer diagnosis.