Shearing DNA in 10 mM Tris alone or in combination with EDTA resulted in fewer G-to-T mutations than when the specimen was sheared in water alone or 1 mM Tris, but the difference was eliminated if DNA was repaired prior to sequencing. DNA repair eliminated 77-82% of G-to-T and C-to-A of very and low frequency variants (<1% and 1-5%, respectively). When DNA enriched for the cancer panel was sequenced without repair, 195 G-to-T and C-to-A variants of moderate frequency (<5%) were detected. In contrast only 12 such variants were detected if DNA was repaired prior to sequencing, indicating that if DNA is not repaired prior to sequencing approximately one false mutation will be detected for every cancer gene investigated.
An excess of G-to-T variants was also observed in the 1000 Genomes and The Cancer Genome Atlas (TCGA) datasets. The percentage of TCGA sequencing runs with a Global Imbalance Value (GIV) G to T of > 2 was 73%, indicating that the majority of these variants were a result of DNA damage. When a GIV threshold of 1.5 was applied, an A-to-T imbalance was found in 0.5% of TCGA sequencing runs and a C-to-T imbalance was found in 3% of TCGA sequencing runs. The authors conclude that stochastic damage is responsible for the G-to-T transversions, which suggests that such artifacts would be more prevalent among low allelic fractions. The occurrence of such artifacts could confound detection of mutations with a similar low allelic frequency. Importantly, 78% of TCGA sequencing reads from tumor specimens yielded false positives for more that 50% of G-to-T variants; and, the percentage of false positives was strongly correlated with DNA damage (r=0.79). Notably, the fraction of germline variants (G-to-T), which occur at higher frequency was unaffected by DNA damage. In a subset of TCGA specimens (~10%), there was a higher number of high confidence somatic C-to-T type mutations which were false positives. TCGA reference files also displayed a 9% increase in G-to-T and C-to-A variants identified with high confidence in specimens with heavily damaged DNA in comparison to those with no damage or weak damage.