Abstract 430: Detection of low-frequency variants from highly degraded DNA samples using a novel library preparation method

Abstract

Diagnostic tools based on next generation sequencing are fundamentally transforming clinical oncology. However, there is a lack of adequate library preparation strategies for highly degraded, clinically relevant samples, such as cell-free DNA (cfDNA) and FFPE DNA. Furthermore, clinical samples are often available in limited quantity, making the detection of pathologic variants particularly challenging. Due to the extreme heterogeneity of these sample types, targeted sequencing is often used to achieve deep coverage of genomic loci and enable detection of low-frequency variants. Commercially available protocols for library preparation require stringent size-selection to remove adapter-dimers, which reduces library complexity. Achieving high specificity can be challenging because low-frequency artifacts arise from a variety of sources, including DNA extraction, library construction, PCR, hybrid selection, and sequencing. These artifacts can be identified by “duplex sequencing”, where strand-specific unique molecular identifiers (UMIs) are used to confirm the presence of an alteration on both strands of an input molecule. However, duplex sequencing typically delivers low conversion rates with degraded samples due to poor ligation efficiency and template loss during size-selection. We present a novel library construction chemistry that eliminates adapter dimers and utilizes duplexed UMIs to increase sensitivity. The method employs a two-step ligation procedure that does not require size-selection or adjustment of AMPure bead ratios during cleanup. We achieve maximal library conversion using a unique, mutant DNA ligase and proprietary sequencing adapters that increase ligation efficiency and suppress chimera formation. We demonstrate performance using three sample types: sheared genomic DNA, cfDNA, and FFPE DNA, across a wide input range (1-1000 ng). To mimic performance with degraded DNA, we created libraries from mixtures of NA12878 and NA24385 cell line DNA sheared to 150 bp with inputs of 1-25 ng and mutant allele fractions (MAFs) down to 0.25%. Libraries were enriched using a 75 kb panel of xGen Lockdown Probes, followed by ultra-deep sequencing and variant calling. When compared to commercially available methods, our approach yielded a 50-100% increase in library complexity with significantly improved sensitivity to <1% variants. We also obtained 100% specificity using duplexed UMI correction, which removed all false-positive calls. To highlight clinical utility, we extended our study to cfDNA samples with inputs of 5-25 ng and MAFs down to 1%, and FFPE DNA samples with inputs of 25-100 ng and MAFs down to 1%. Citation Format: Ushati Das Chakravarty, Zac Zwirko, Yu Zheng, Madelyn Light, Kevin Lai, Keith Bryan, Scott Rose, Yun Bao, Mirna Jarosz, Caifu Chen. Detection of low-frequency variants from highly degraded DNA samples using a novel library preparation method [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 430.