Fig. 1: Overview of THEMIS approach for cancer detection based on plasma whole-methylome sequencing.

Schematic illustration of the experimental and bioinformatics procedures. Blood samples were collected from cancer patients or noncancer control donors. Plasma cfDNA was extracted from the participant’s blood sample and subject to low-pass WMS using TET2 and APOBEC enzymes for cytosine conversion. Four modalities were extracted from uniquely mapped WMS sequencing reads, including Methylated Fragment Ratio (MFR), Fragment Size Index (FSI), Chromosomal Aneuploidy of Featured Fragments (CAFF), and Fragment End Motif (FEM). An ensemble model integrating prediction scores from all four modalities was constructed to yield the final probability of having cancer, termed THEMIS score, for a sample.