Fig. 4. Benchmark Datasets.

For single cells, DNA was multiple displacement amplified (MDA). a Whole genome sequencing (WGS) dataset, generated from a clonal population started with an individual cell²⁸, and expanded further downstream to generate different bulk samples (C1, C2, C3, and IL-1C). b Newly generated whole exome sequencing (WXS) dataset of blood cells. Ground truth genotypes of patient CCS1 were determined from sequencing data of family members (boxes male; circles female). Granulocytes were isolated from blood using magnetic and fluorescence-activated cell sorting (MACS and FACS). c Whole exome sequencing (WXS) dataset from a triple-negative breast cancer (TNBC) patient²¹. For tumor cells (a1 to a8 and h1 to h8), the cell population (pop) sequencing of the normal sample was taken as the ground truth, including known clonal tumor-specific somatic variants previously validated by deep targeted duplex (td, dx) sequencing²¹. For normal cells (n1 to n16), the cell population (pop) sequencing of the tumor sample was taken as the ground truth, after removing known clonal and subclonal tumor-specific somatic variants previously validated by deep targeted duplex (td, dx) sequencing²¹.