Fig. 1

cfDNA-FEP overview and study design. A Schematic representation of cfDNA-FEP. Targeted amplification is done with a single specific primer for each target region and preserves the original fragment ends. Subsequent high-coverage NGS reveals the distribution of cfDNA fragment ends and sequence motifs. B Experiment design. The study cohort (n = 175) was split into two batches. Exploratory data analysis and feature preparation were done on the full dataset after quality control. Model tuning was done on the training subset, followed by a performance evaluation on the separate testing subset. C Fragment end profiles for stage IV cancer samples and healthy controls in 3 COAD-specific regions are visualized as empirical cumulative distribution functions (standard deviation range, top panel) and densities (bottom panel). D. Frequencies of cfDNA dinucleotide end motifs in healthy (n = 59), RCC (n = 56) and COAD (n = 56) samples. Significance determined with Wilcoxon test is indicated by asterisks (**** - p < 0.0001; *** - p < 0.001; ** - p < 0.01, * - p < 0.05, ns - not significant)