(A) In bulk sequencing, a somatic mutation present in k out of n cells pooled together for sequencing (i.e. mosaicism of k/n), with read coverage D at the mutation locus, will be detected on average in k/n·D/2 reads with a variance depending on sampling error; i.e. the number of reads detecting the mutation correlates linearly with the percent mosaicism. In contrast, germline heterozygous and homozygous variants are present in D/2 and D reads, respectively. Due to sequencing artifacts and sequencing errors, a mutation must be detected above a threshold number of reads, T, which also depends on the sequencing depth, D, since errors occur at rates, e, that are a constant fraction on average of the total number of reads (T=z·e·D; z is a constant chosen based on desired detection sensitivity and specificity). The fraction of error reads, e, is a constant on average that is independent of total sequencing depth, D, because library artifacts and sequencing errors occur at rates that are independent of total sequencing depth. The threshold, T, can be reduced with methods reducing sequencing error, but errors are still present in any current sequencing technology. Combining equations simplifies to k/n ≥ 2·z·e. This means that the mosaicism of a somatic mutation must be at least twice the sequencing error rate (or more, depending on the confidence factor) for detection to be possible in bulk DNA sequencing, regardless of sequencing depth. Below a certain level of mosaicism that depends on the sequencing error rate, detection is unlikely. Note: for simplicity, the height of the histograms (# of mutations) is scaled to the same height, and the equations do not include variance terms. (B) In single-cell sequencing, somatic mutations are present at the same signal level on average as germline heterozygous variants (i.e. D/2, since k/n = 1), enabling detection of low mosaicism mutations that would otherwise be below detection thresholds of bulk sequencing due to sequencing error. Due to whole genome amplification, single-cell sequencing also leads to greater variance in mutation and error signal level distributions (non-uniform amplification and dropout) and entails additional artifacts not present in bulk sequencing, which increases the noise level, e', but still a lower level on average than true heterozygous mutations. However, the signal distribution of artifacts may still overlap that of true mutations, necessitating careful bioinformatics and modeling of error and true mutation signals along with rigorous validation. Note, for simplicity, the equations here do not include variance terms and bioinformatic modeling usually includes additional parameters other than read count illustrated here. Single-cell sequencing does not achieve increased sensitivity for somatic mutations without cost, because to detect a given mutation with k/n mosaicism, more than n/k single cells may need to be sequenced. The benefit of single-cell sequencing is not to reduce sequencing costs, but rather its ability to overcome limitations due to sequencing error rates on the minimum mosaicism detectable and maintaining information as to which somatic mutations are found within the same cell, which enables lineage tracing.