Fig. 2: Study design and overview of the computational method.
From: Cost-effective methylome sequencing of cell-free DNA for accurately detecting and locating cancer
a Overview of the sample usage for marker discovery, model training, and validation. All tissue samples are used for marker discovery, and all plasma samples are randomly split into three sets, used for marker discovery, training, and validating the predictive model. The plasma sample split is repeated 10 times and the prediction performance is averaged over the 10 runs. b Details of sample usage for marker discovery. Different types of methylation markers were discovered by using different samples. Note that 30 reference noncancer plasma samples (in blue boxes) correspond to “marker filtration” in a. Abbreviations: TOO tissue of origin.
