Performance evaluation of DREAM. (A) Schematic outline of the principles of DNA methylation analysis. In DREAM, DNA is sequentially cut using 2 restriction enzymes, SmaI and XmaI. Estimated durations are indicated alongside each step. Individual sample sequencing could be completed in 4 to 5 days, including SmaI digestion (3 hours), XmaI digestion (16 hours), end repair (2 hours), adaptor ligation (1 hour), PCR amplification (2 hours), final library quality control (1 day), sequencing (2-3 days), and data analysis time. The hands-on time was <6 hours per sample for the entire process. In reduced representation bisulfite sequencing (RRBS), DNA is digested by MspI to generate short fragments that contain CpG dinucleotides at the 3' ends. (B) In DREAM, the ends of the sequencing reads generated by SmaI or XmaI locate at 5'-CCCGGG-3' sites, with 4 types of read ends. Left-1, left-3, right-2, and right-4 can be caused by these restriction enzymes, whereas others can be caused by random fragmentation. Unmethylated 5'-CCCGGG-3' sites (presumed to be digested by SmaI) generate left-3 and right-2 ends (unmethylated signature [u]), and methylated 5'-CCC(me)GGG-3' sites (presumed to be digested by XmaI) generate left-1 and right-4 (methylated signature [m]). The methylation ratio was simply calculated as the fraction of total 5'-CCCGGG-3' site sequencing reads (m + u) that were mapped to m. (C) The human genome has 28.0 million CpG sites, and the GRCh37/hg19 annotation provides coordinates for 374 921 5'-CCCGGG-3' sites targeted by DREAM, including the 6455 sites covered by 450K. (D) Results of DREAM of 4 pg of methylation calibrators spiked in 100 ng of sample DNA from 19 patients are shown. EC247, EC293, and EC466 were used as unmethylated or 100% methylated calibration standards. (E) The median number of CpG sites and the median Pearson’s correlation coefficient (r²) between DREAM methylation ratios and 450K β values at 6455 overlapping CpG sites classified according to DREAM sequence depth for 19 cases are shown. Error bars represent standard errors. (F) Correlation of the discovery cohort (n = 99) between ratios obtained using DREAM and 450K β values at 1703 overlapping CpG sites with sufficient coverage (≥20 reads in 95% of 99 samples). The best (G) and worst (H) correlations are shown. TSS, transcription start site.