would be incorrectly called a genetic variant. A similar process can occur in CirSeq and singlestranded barcoding, whereby a misinsertion event occurring during the first round of synthesis can be propagated to subsequently synthesized daughter molecules. When sequenced, these subpopulations of related molecules will all contain a copy of the original misincorporation event, and they will be erroneously scored as a mutation. However, because the same mutation is unlikely to occur in unrelated molecules, these artifactual variants would give the appearance of a subclonal mutation. DNA polymerases typically used in library construction make misinsertions at a frequency between 10−4 and 10−6, which can lead to thousands of false positives on a typical sequencing run. Damaged or degraded DNA is particularly sensitive to this form of error because of the prevalence of DNA adducts that cause erroneous base pairings during polymerization. Removal of DNA damage with the addition of glycosylases or in vitro repair kits has been shown to reduce the number of false mutations in these samples31,32. However, not all mutagenic lesions are recognized by these enzymes, nor is the fidelity of repair perfect, thus limiting their utility in error correction.
