Systematically Evaluating Cell-Free DNA Fragmentation Patterns for Cancer Diagnosis and Enhanced Cancer Detection via...

Area of Science:

• Genomics
• Bioinformatics
• Oncology

Background:

• Cell-free DNA (cfDNA) fragmentation patterns are crucial for early cancer detection.
• Existing definitions of cfDNA fragmentation lack standardization, hindering research and clinical application.
• Systematic comparison of fragmentation patterns, particularly within open chromatin regions, is needed.

Purpose of the Study:

• To systematically compare 10 cfDNA fragmentation patterns within open chromatin regions for cancer detection.
• To evaluate the diagnostic value of these patterns using machine learning.
• To develop an improved classifier for enhanced cancer detection and tissue-of-origin determination.

Main Methods:

• Collected 1382 plasma cfDNA sequencing samples from 8 cancer types.
• Examined 10 cfDNA fragmentation patterns within open chromatin regions.
• Employed machine learning techniques, including an ensemble classifier, for performance evaluation.

Main Results:

• All 10 fragmentation patterns showed classification capabilities, with the end motif being most effective in cross-validation.
• Patterns combining fragment length and coverage demonstrated robust predictive capacity.
• An ensemble classifier integrating all patterns significantly improved cancer detection and tissue-of-origin prediction.
• Bioinformatics analysis identified critical regulatory regions involved in cancer pathogenesis.

Conclusions:

• cfDNA fragmentation patterns within open chromatin regions hold significant potential for early cancer detection.
• An ensemble machine learning approach integrating multiple fragmentation features enhances diagnostic accuracy.
• This integrated approach aids in both cancer detection and determining the tissue of origin, with implications for understanding cancer pathogenesis.