Open chromatin-guided interpretable machine learning reveals cancer-specific chromatin features in cell-free DNA

Area of Science:

• Genomics
• Epigenetics
• Cancer Biology

Background:

• Cell-free DNAs (cfDNAs) are DNA fragments in blood, originating from immune cells in healthy individuals and from both immune and cancer cells in patients.
• cfDNAs carry cancer-specific mutations and retain epigenetic features like DNA methylation and nucleosome positioning.

Purpose of the Study:

• To investigate nucleosome enrichment patterns in cfDNAs from breast and pancreatic cancer patients.
• To develop an interpretable machine learning framework for cfDNA-based cancer detection.

Main Methods:

• Analysis of nucleosome enrichment patterns in cfDNAs from cancer patients.
• Utilizing ATAC-seq data to identify open chromatin regions.
• Application of an interpretable machine learning model (XGBoost) trained on cell type specific open chromatin regions.

Main Results:

• Significant nucleosome enrichment was observed at open chromatin regions in cfDNAs.
• Differential enrichment patterns were identified at cancer cell-specific and CD4+ T cell-specific ATAC-seq peaks.
• The machine learning approach improved cancer detection accuracy and highlighted key genomic loci.

Conclusions:

• Nucleosome enrichment patterns in cfDNAs reflect both tumor- and immune-derived signals.
• The developed pipeline offers a robust and interpretable framework for cfDNA-based cancer detection.
• Epigenetic features of cfDNAs hold significant potential for non-invasive cancer diagnostics.