Extended Data Fig. 4: Comparison of UBS-seq and conventional BS-seq in application to cfDNA samples.

a) For partially methylated regions, the proportion of methylated fragments was marked as x, while 1-x was the proportion of unmethylated fragments. BS treatments caused more degradation on reacted C sites and reduced the fraction of unmethylated fragments in sequencing libraries. The measurement of 5mC level would be affected by the degradation. Completely methylated and completely unmethylated sites showed less bias in the quantification results, while the stoichiometry of partially methylated sites could be over-estimated. b) Pairwise comparison of total sequencing coverage of all CpG sites within genomic windows (10 kb). Two replicates of conventional BS-seq samples (BS #1 vs. BS #2) and two replicates of UBS-seq samples (UBS #1 vs. UBS #2) both showed high correlation within methods (r = 0.99), but the correlation is slightly lower between methods (r = 0.96 ~ 0.97). c) Comparison of the methylated levels in two replicates of cfDNA UBS-seq samples within an example region. Highly methylated sites are colored in green and unmethylated/lowly methylated sites are colored in red. Detected 5mC ratios shown in y-axis were consistent between two replicates at single base resolution. d) Another example region showing consistency of methylation profile between two replicates of cfDNA UBS-seq libraries.