Integrating Plasma Cell-Free DNA Fragment End Motif and Size with Genomic Features Enables Lung Cancer Detection

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Early lung cancer detection is critical for improving patient survival rates.
• Current lung cancer screening methods have limitations.
• Liquid biopsy using cell-free DNA (cfDNA) presents a promising alternative or complementary screening approach.

Purpose of the Study:

• To develop and validate an accurate lung cancer detection method by integrating cfDNA fragmentomic and genomic features.
• To enhance lung cancer detection accuracy across diverse populations using deep learning classifiers.
• To assess the performance of novel cfDNA-based features for lung cancer screening.

Main Methods:

• Trained deep learning-based classifiers using cfDNA fragmentomic features (fragment end motif by size, genomic coverage) from multi-institutional studies.
• Utilized discovery and validation datasets including Korean and Caucasian cohorts.
• Evaluated classifier performance using area under the curve (AUC).

Main Results:

• Classifiers integrating fragment end motif by size achieved an AUC of 0.917 in the discovery dataset.
• An ensemble classifier combining fragment end motif by size and genomic coverage reached an AUC of 0.937.
• The developed classifier demonstrated consistent performance in Korean and Caucasian validation cohorts, showing ethnic generalizability.

Conclusions:

• Integrating cfDNA fragmentomic and genomic features with deep learning offers a highly accurate method for lung cancer detection.
• This approach shows potential for effective lung cancer screening across diverse ethnic populations.
• The study highlights a promising advancement in early cancer detection, addressing challenges of current screening modalities.