Pingfang Liu

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

When is a mutation a true genetic variant? Large-scale sequencing studies have set out to determine the low-frequency pathogenic genetic variants in individuals and populations. However, Chen et al. demonstrate that many so-called low-frequency genetic variants in large public databases may be due to DNA damage. They scored libraries sequenced with and without a DNA damage–repairing enzymatic mix to assess the proportion of true rare variants. It remains to be seen how best to repair DNA before sequencing to provide more accurate assessments of mutation. Science, this issue p. 752 Damaged DNA may artificially increase the number of rare variants identified in large-scale sequencing projects. Mutations in somatic cells generate a heterogeneous genomic population and may result in serious medical conditions. Although cancer is typically associated with somatic variations, advances in DNA sequencing indicate that cell-specific variants affect a number of phenotypes and pathologies. Here, we show that mutagenic damage accounts for the majority of the erroneous identification of variants with low to moderate (1 to 5%) frequency. More important, we found signatures of damage in most sequencing data sets in widely used resources, including the 1000 Genomes Project and The Cancer Genome Atlas, establishing damage as a pervasive cause of sequencing errors. The extent of this damage directly confounds the determination of somatic variants in these data sets.

Response to Comment on “DNA damage is a pervasive cause of sequencing errors, directly confounding variant identification”

Following the Comment of Stewart et al., we repeated our analysis on sequencing runs from The Cancer Genome Atlas (TCGA) using their suggested parameters. We found signs of oxidative damage in all sequence contexts and irrespective of the sequencing date, reaffirming that DNA damage affects mutation-calling pipelines in their ability to accurately identify somatic variations.

Abstract 1425: A multi-enzyme DNA repair mix improves library quality and sequencing accuracy in FFPE tumor samples

Next-generation sequencing (NGS) methods are used extensively to profile mutations present in diseased human tissues. These genomic approaches hold great promise for personalized medicine but sequencing accuracy is essential for proper patient diagnosis and determining a treatment plan. A common source of DNA for genomic profiling is formalin-fixed, paraffin-embedded (FFPE) tissue samples obtained from patient biopsy. FFPE DNA poses important challenges for preparing NGS libraries including low input amounts and poor DNA quality, resulting from extensive fixation- and storage-induced DNA damage. Additionally, these damage-induced sequencing artifacts raise the background level of mutations, making it difficult to discern true, low frequency, disease-causing variants from noise. We previously showed that a major fraction of somatic mutations described in publicly available datasets are due to such sequencing artifacts (Chen et al., Science 2017). Furthermore, we showed that enzymatic repair of DNA before library preparation improves the library quality and reduces background noise. We developed a second-generation DNA repair enzyme mix (V2) that efficiently repairs the most prevalent damage types found in FFPE DNA and further improves the quality and yield of NGS libraries. Additionally, we tested the efficacy of the V2 repair mix in improving sequencing accuracy for FFPE DNA samples obtained from different cancer tissues. We performed target enrichment on a panel of 151 cancer genes, deep sequenced, and performed variant analysis. For a subset of variants, we further validated our results using a droplet digital PCR (ddPCR) assay. Both methods showed that the V2 repair mix did not alter the overall frequency of variants identified, thus it did not introduce bias, but significantly improved the sequencing accuracy by reducing the number of false variant calls. Therefore, enzymatic repair is a critical first step in preparing FFPE DNA sequencing libraries, allowing more sensitive and robust detection of low frequency, disease variants. Citation Format: Pingfang Liu, Margaret Heider, Chen Song, Lixin Chen, Laurence Ettwiller, Lauren Higgins, Eileen Dimalanta, Theodore Davis, Thomas Evans. A multi-enzyme DNA repair mix improves library quality and sequencing accuracy in FFPE tumor samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1425.

Abstract 5365: Combining enzymatic DNA fragmentation with NGS library construction results in high quality, high yield libraries

The use of Next Generation Sequencing (NGS) data has been instrumental in advancing our understanding of human genetics, identifying the molecular events that contribute to human disease, and supporting drug development targeted towards precision medicine. Continued advancement relies on overcoming the limitations and bottlenecks associated with NGS. In this work, we have focused on NGS library preparation, where the requirement for expensive equipment and numerous steps can lead to sample loss, errors, and limited throughput. Specifically, we have developed a library construction method that integrates enzymatic DNA fragmentation into the workflow and combines fragmentation with end repair and dA-tailing in a single step. Integrating these reactions eliminates the need for costly equipment to shear DNA and reduces the number of sample transfers and losses. Adaptor ligation is also carried out in the same tube, after which a single cleanup step is performed. For low input samples, PCR amplification is performed prior to sequencing. This method is compatible with a broad range of DNA inputs and insert sizes. Libraries generated using this streamlined method with inputs ranging from 500 pg to 500 ng of intact DNA show no significant difference in coverage uniformity or sequence quality metrics, compared to libraries generated with mechanically sheared DNA. Similarly, libraries generated to contain insert sizes that range from 150bp to 1kb display no significant difference in sequence quality from each other or from those generated with mechanically sheared DNA. Finally, this streamlined method generates libraries of substantially higher yields than those generated using mechanically fragmented DNA, allowing the use of lower DNA inputs and fewer PCR cycles. The ability to generate high quality NGS libraries from intact DNA without the need for costly equipment and numerous cleanup or liquid transfer steps substantially reduces the time, cost and errors associated with library construction. In addition, these advances will enable greater use and adoption of NGS technologies in clinical and diagnostic settings. Citation Format: Fiona Stewart, Lynne Apone, Vaishnavi Panchapakesa, Karen Duggan, Timur Shtatland, Bradley Langhorst, John Murdoch, Christine Sumner, Christine Rozzi, Pingfang Liu, Keerthana Krishnan, Deyra Rodriguez, Joanna Bybee, Danielle Rivizzigno, Laurie Mazzola, Eileen Dimalanta, Theodore Davis. Combining enzymatic DNA fragmentation with NGS library construction results in high quality, high yield libraries [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 5365. doi:10.1158/1538-7445.AM2017-5365

Abstract 5360: DNA repair increases sequencing accuracy without altering actual mutation frequency in clinical samples

Targeted cancer therapy based on genomic alterations can be remarkably effective. Currently, cancer genome profiling using next generation sequencing (NGS) is routinely applied in cancer care to guide personalized treatment. The accuracy of this profiling directly impacts therapeutic choices and the outcomes of patient care. We previously showed that false positive variants are abundant and can account for a major fraction of identified somatic variations in publicly available datasets (doi: http://dx.doi.org/10.1101/070334). These false positive variants show signs of mutagenic DNA damage. We further demonstrated that enzymatic DNA repair increases sequencing quality by lowering damage-induced background noise. Therefore, enzymatic DNA repair has the potential to improve sequencing accuracy, avoiding incorrect somatic variant calls and consequently reducing incorrect diagnostic conclusions. In this study, we investigated whether enzymatic DNA repair introduces any bias to NGS libraries using analysis by droplet digital PCR (ddPCR) and deep sequencing. DNA Reference Standards containing multiple common cancer mutations (Horizon Discovery, Inc.) were spiked into formalin-fixed paraffin-embedded (FFPE) DNA isolated from tumor samples from different tissue types at defined frequencies (0.5-10% quantified by ddPCR). Genotyping of the FFPE DNA ensured that they were free of any of the spiked-in mutations. After DNA repair and library preparation, mutation frequencies were quantified by ddPCR, and compared to the mutation levels in input DNA and control libraries without repair. Deep sequencing of 151 cancer genes including these spike-ins showed no difference in mutation frequency for the spiked-in mutations between the control and repair groups. However, the number of false positive variant calls was reduced in the repair group. Our data demonstrates that DNA repair significantly increases sequencing accuracy without altering the frequency of actual mutations in tumor samples. Citation Format: Pingfang Liu, Lixin Chen, Laurence Ettwiller, Eileen Dimalanta, Theodore B. Davis, Thomas C. Evans. DNA repair increases sequencing accuracy without altering actual mutation frequency in clinical samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 5360. doi:10.1158/1538-7445.AM2017-5360

Abstract 3628: Improving sequencing quality of libraries prepared from FFPE DNA

Targeted cancer therapy based on genomic alterations can be remarkably effective, and has made significant strides with the recent advances in next-generation sequencing (NGS) technology. Although samples of blood and other bodily fluids are being actively explored for early disease diagnosis and treatment monitoring, DNA isolated from FFPE samples is currently the main source for NGS-based cancer profiling in clinical settings. Unfortunately, sequencing DNA from FFPE samples is challenging due to limited quantities and poor quality, a result of DNA damage incurred during fixation and storage. Artifacts associated with FFPE DNA have limited the mutation detection sensitivity to ≥ 5% mutant allele frequency (Frampton et al, Nature Biotechnology 2013), which would unfortunately leave many low-abundance genetic variants of clinical significance undetected. For example, clinical resistance-causing KIT and EGFR mutations can be present in tumors at levels In this study, we investigated the effects of DNA repair and different sample handling workflows on sequencing quality of libraries prepared from FFPE samples. Careful analysis of sequencing data showed that base calling qualities for all 4 bases are improved, and aberrant G:C to A:T mutations were significantly reduced upon DNA repair. Because the large majority of mutations encountered in human tumors are G:C to A:T mutations (Greenman, C. Nature 2007), we expect that lowering the damage induced background noise of FFPE DNA would allow more reliable detection of clinically important, actionable mutations at lower abundance. In addition, we observed specific sequencing artifacts associated with the method of handling FFPE samples and have since identified effective measurements to avoid such artifacts. We expect that these improvements in sequencing quality of FFPE samples would ultimately enable more sensitive and robust detection of many low level genetic variations in clinically and biologically relevant cancer genes. Citation Format: pingfang liu, Lixin Chen, Laurence Ettwiller, Christine Sumner, Fiona J. Stewart, Eileen T. Dimalanta, Theodore B. Davis, Evans C. Thomas. Improving sequencing quality of libraries prepared from FFPE DNA. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3628.

Abstract 3620: Enhancing clinical utility of NGS with reduced bias, low DNA input, library construction

Early detection and diagnosis of cancer substantially increases the likelihood for successful treatment. Tools that aid in detecting and diagnosing cancer early, therefore, have the potential to greatly impact the clinical outcome for cancer patients. Next Generation Sequencing (NGS) has emerged as an important tool in this area. The technology is sensitive, fast and high throughput to allow sequencing of many samples at once. Unfortunately, many clinical samples go unanalyzed because they do not yield sufficient quantities of DNA to generate NGS libraries or the libraries generated require so many rounds of PCR amplification that they display extreme sequence bias. Bias not only hampers data analysis, but also increases costs by requiring excess sequencing to obtain sufficient coverage over all relevant genomic regions. To enable the increased use of NGS in the clinic and reduce the amount of sequence bias generated during library preparation, we have developed a PCR free library construction method that uses low quantities of DNA as input. As an initial test of the method, we generated PCR free libraries from 100ng, 50ng and 25ng of human genomic DNA. The libraries where pooled and sequenced on the Illumina NextSeq 500 instrument to approximately 10X coverage. All libraries, irrespective of input amount, showed minimal AT/GC bias and excellent coverage distributions, with most bases covered within 5X of the expected coverage depth. In addition, regions identified as difficult to sequence (Aird, D., et.al., 2011 and Ross, M. G., et.al., 2013) showed coverage at near expected levels for all libraries. This method can easily be adapted for use with extremely low DNA inputs by the introduction of a minimal number of PCR cycles. In fact, we have used this method to construct high quality NGS libraries with picogram quantities of DNA input. Standard library construction methods require DNA inputs of 2ug to 500ng when PCR amplification is omitted. This new method utilizes inputs as low as 25ng to generate high-quality PCR free libraries and picogram quantities when amplification is performed. We are currently investigating the possibility of reducing input levels further and exploring the limits of the method with low quality DNA samples. Interestingly, we have observed substantial sample loss during DNA shearing and reaction cleanup. Samples that do not require fragmentation, such as DNA isolated from plasma (cfDNA) and low quality FFPE DNA, may reduce the input requirements even further. Finally, this new method utilizes low sample and reagent volumes, possibly paving the way for its use in microfluidic devices. Citation Format: Lynne Apone, Pingfang Liu, Vaish Panchapakesa, Deyra Rodriguez, Karen Duggan, Krishnan Keerthana, Nicole Nichols, Yanxia Bei, Julie Menin, Brad Langhorst, Christine Sumner, Christine Chater, Joanna Bybee, Laurie Mazzola, Danielle Rivizzigno, Fiona Stewart, Eileen Dimalanta, Theodore Davis. Enhancing clinical utility of NGS with reduced bias, low DNA input, library construction. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3620.

A fast solution to NGS library preparation with low nanogram DNA input.

Next-generation sequencing (NGS) has significantly impacted human genetics, enabling a comprehensive characterization of human genome as well as better understanding of many genomic abnormalities. By delivering massive DNA sequences at unprecedented speed and cost, NGS promises to make personalized medicine a reality in the foreseeable future. To date, library construction with clinical samples has been a challenge, primarily due to the limited quantities of sample DNA available. To overcome this challenge, we have developed a fast library preparation method using novel NEBNext reagents and adaptors, including a new DNA polymerase that has been optimized to minimize GC bias. This method enables library construction from an amount of DNA as low as 5 ng, and can be used for both intact and fragmented DNA. Moreover, the workflow is compatible with multiple NGS platforms.