Abstract 4590: Larger family sizes improve stochastic error correction during untargeted searches of very-low frequency variant alleles in circulating cell-free DNA (ccfDNA)

Abstract

Abstract Detection of ccfDNA derived from non-metastatic solid tumors requires discrimination of very-low frequency true variants from NGS library preparation and sequencing noise that commonly occurs at a similar frequency (<1%). Correction of NGS-related errors is particularly important during searches of unknown tumor-associated variants with multi-gene panels. Unique molecular identifiers (UMIs) have been used during in silico error correction by molecular barcoding and performing a consensus base calling at each position from DNA labelled with the same UMI (i.e., PCR duplicates). In this study, we sought to determine the utility of UMIs for error correction of stochastic noise in ccfDNA. ccfDNA was isolated from the plasma of 11 healthy participants. An eight base-pair random barcode (UMI) was ligated prior to library amplification. Libraries were enriched using a custom capture probe set (128 genes; 128 kb) and then sequenced (HiSeq 2500). Reads were aligned, grouped based on UMI similarity, and a consensus read sequence was determined. The number of PCR duplicates that yielded a single consensus sequence was defined as family size (FS). For non-reference alleles detected at a variant allele frequency (VAF) ≥ 0.1% and ≤ 1.0%, there was a significant reduction in observed counts with increasingly larger FSs (Table 1). This effect was observed for both unique and common non-reference allele counts (Table 1). Of note, counts of unique non-reference alleles (stochastic noise) were consistently greater than the common non-reference alleles regardless of family size (Table 1). The use of larger family sizes reduced the stochastic noise that may confound detection of very-low frequency variants present in ccfDNA during searches for unknown tumor-associated variants. Strategies that generate larger family sizes or suppress early PCR errors may further improve NGS specificity to broaden liquid biopsy clinical applications. Table 1. Mean (SD) Non-Reference Allele Counts with a VAF between 0.1% and 1.0% in Cell-Free DNAFS >= 1FS >= 5FS >= 10FS >= 15FS >= 20ANOVA Results F(4,50)P-valueAll Variants5536 (1230)3988 (881)2621 (805)1028 (552)1028 (552)49.8<0.001Unique Variants3597 (1107)2635 (766)1112 (526)709 (434)709 (434)27.6<0.001Common Variants1939 (161)1353 (188)521 (141)318 (121)318 (121)186.1<0.001 Citation Format: Preetida J. Bhetariya, David Nix, Sabine Hellwig, Gabor Marth, Mary Bronner, Hunter Underhill. Larger family sizes improve stochastic error correction during untargeted searches of very-low frequency variant alleles in circulating cell-free DNA (ccfDNA) [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 4590.