heterogeneous mixtures can result in skewed populations, and
polymerase errors can result in wrong base incorporations and
rearrangements. Furthermore, errors arising during sequencing
process can result in approximately 0.1–1% incorrect bases calling
[6], which are known as sequencing errors. Table1 shows the error
ratios of different major NGS platforms.
Library preparation can also introduce significant errors. For
instance, guanine oxidation is an important source of artificial
mutations because 8-oxoG tends to pair with adenine instead of
cytosine [7]. Long-time heat incubations, which are common in
many DNA extraction and hybrid capture protocols, can signifi-
cantly increase the number of G!Tsubstitutions. Recently a study
showed that a DNA repairing process can eliminate 77% and 82% of
G!T and C!A errors, respectively [8]. This study indicates
DNA lesions can cause a large amount of errors.
Besides errors introduced during sample preparation and
sequencing, software and analysis tools can also introduce errors.
Particularly, false-positive variants can be called in the referenceFig. 2The workflow of ctDNA sequencing. In this workflow, a blood sample is first processed, then DNA is
extracted from the blood plasma, and an initial library is prepared. This initial library will then be amplified by
PCR and be enriched with target capturing methods. The captured library can be amplified again and be
processed to a sequencing library, which can be pooled and then be sequenced to generate the data
70 Shifu Chen et al.