Fig. 2

a–c The accuracy of variant calling was evaluated for Conbase, Monovar, SCcaller, and LiRA, reflecting the ability of the methods to detect clonal sSNV loci a population of cells. The sensitivity (a) and specificity (b) of Conbase, Monovar, SCcaller, and LiRA to detect clonal mutations in at least two cells in simulated data at increasing dropout probabilities (pDO) at different levels of alignment artifact probabilities (pEAL). c The false discovery rate of Conbase, Monovar, SCcaller, and LiRA when detecting clonal mutations in simulated data at increasing alignment artifact probabilities (pEAL) at different levels of dropout probabilities (pDO). d The accuracy of genotyping was evaluated for Conbase, Monovar, SCcaller, and LiRA, reflecting the ability of the methods to correctly predict genotypes in each sSNV loci and each cell at increasing dropout probabilities (pDO). In each simulation, representing one bar, the true genotype in 50% of the samples were heterozygous, representing samples harboring a sSNV. The remaining samples were homozygous for the reference allele, representing the ancestral unmutated state