This web page was produced as an assignment for an undergraduate course at Davidson College.

Helen Webster's Genomics Home Page

DNA damage is a pervasive cause of sequencing errors, directly confounding variant identification

Main Idea
To test the possibility that mutagenic damage causes sequencing inaccuracies and variation, Chen et al. created a standard score (GIV score) for determining read imbalances and sequencing variants. Public sequencing datasets contain this mutant variation mixed in with natural variation, and it is unclear to those who use the data that some variation is caused by mutagens. Mutagenic damage of DNA causes an imbalance in base transversions between the reads of the two DNA strands, and is a major confounding effect of damage. This imbalance is used to the researchers' advantage, as it can ideally be used as a basis to track mutagenic damage and therefore sequencing errors in these databases, as opposed to natural variance present in the population. This GIV score can ideally be used to accurately estimate the damage present in public data sets. Chen et al. conclude that the GIV score accurately quantifies mutagenic damage in somatic variant cells, which occurs in very low frequencies. This affects a substantial portion of

My Opinion
I found this paper especially intriguing given the integral role DNA sequencing has played in the biological research I have conducted at Davidson, as well as vital to nearly every part of the work we read about in our genomics course. It is incredible to me that mutagens could play such a detrimental role not just physically for the DNA itself, but for further findings and usaage of DNA reads and sequences. I think this paper did an impressive job of establishing a necessary a score system for damage, that would have obvious positive benefits for improving public data sets. In addition, the display of proving the scoring system and proceeding to apply it with tumor cells thoroughly convinced the reader of the benefit of the GIV score in addition to the credibility of the method. Finally, this GIV score is a novel idea with a necessary application, that I think can quickly be useful to genomicists and bioinformaticists.

Figure 1
Figure 1A outlines the principle behind the GIV score. Mutagenic variation causes an imblanace of transversion between the two reads because the base switch between the two strands does not happen equally. The left side of Panel A shows the base transversion imbalance that results from sequencing mutagenically damaged sequences. When the variant is just a natural SNP (right side of panel), reads from both strands will transverse equally. The degree of imbalance due to damage is the basis of the GIV score. Figure 1B visually shows the fraction of G to T transversions and the inequity between reads 1 and 2.

Figure 2
To estimate the amount of damage present in public DNA sets, Chen et al. calculated the GIV scores for the 1000 Genomes Project dataset (Figure 2A) and a subset of the TCGA data set (Figure 2B). The gray line in both A and B demarcates a GIV score of 1.5, above which the score implicates damage, below which is non-damage. Both sets show widespread erroneous sequencing calls, 30% of which were G to T variant reads. T to A and C to T were 0.5% and 3% of erroneous calls, respectively. Overall, Figure 2 shows there is indeed DNA sequence damage in public data sets that leads to erroneous sequencing calls in at least 1/3 of G to T variant reads.

Figure 3

Supplementary data found G to T transversions to be randomly generated, implying they occur at low allelic fractions (known as somatic variants, as opposed to the high frequency germline variants). Figure 3 looks at how damage affects somatic variant identification. DNA repair eliminates 82% of G-to-T and C-to-A variant positions in the low frequency groups (less than 1% and 1% to 5%), proving those positions are erroneous and due to damage in somatic variants. This leads to false positives and direct confounding in the identification of variance in sequences reads.

Figure 4
Figure 4 sorts approximately 1800 tumor sequencing runs by G-to-T variant GIV score. There are more G-to-T somatic variants than C-to-A, and the fraction increases with increasing GIV score. In addition, panel 4D shows germline variants remaining consistent in GIV score. Estimated false positives in somatic variants is strongly correlated to estimated damage in these tumor samples, ultimately supporting the application of the GIV score to accurately detect high somatic damage and false positives in sequence reads.

Citation: Chen L, Liu P, Evans T, Ettwiller L. (2017). DNA damage is a pervasive cause of sequencing errors, directly confounding variant identification. Science 355, 752-756.

  Helen's Home Page

Genomics Page

Biology Home Page

Email Questions or Comments:

Copyright 2016 Department of Biology, Davidson College, Davidson, NC 28035