Abstract 5598: Development and optimization of a comprehensive high-sensitivity NGS cancer assay and bioinformatics pipeline for plasma cfDNA profiling

The accessibility of tumor-derived cell-free DNA (cfDNA) in blood plasma provides a means to non-invasively profile somatic mutations in solid tumor patients. Clinical applications include longitudinal monitoring of disease burden and acquired drug resistance, identification of clinically relevant alterations and mutation signatures, and detection of minimal residual disease. However, the low fraction of tumor-derived cfDNA in plasma in many patients requires assays and bioinformatics methods that are much more sensitive than have been used for traditional tissue-based analysis. The design of our cfDNA NGS panel is based on prospectively-collected clinical sequencing data obtained from more than 20,000 patients at Memorial Sloan Kettering Cancer Center using MSK-IMPACT, a custom 468-gene sequencing test authorized by the FDA for somatic mutation profiling. Exons harboring hotspot mutations, clinically actionable mutations, and elevated somatic mutation rates were selected for inclusion in the cfDNA panel. Additional non-coding content was included to enable optimal detection of selected copy number alterations, regions of loss of heterozygosity, rearrangement breakpoints, and microsatellite instability. Altogether the panel contains 208 kilobases of sequence from 129 cancer genes. Ultra-deep sequencing and unique molecular indexing enable PCR-generated replicate sequences to be collapsed into error-free consensus sequences, thereby facilitating the high-confidence detection of mutations present at low allele fractions. We developed an open source bioinformatics tool, Marianas, for collapsing PCR replicates into consensus sequences and computing associated quality and performance metrics. Marianas incorporates many empirically derived features that lead to significant noise reduction. It efficiently processes a bam file with 20,000X coverage in 20 minutes on a single processor. We benchmarked the performance of Marianas against other available tools for collapsing and consensus base calling. The relative contributions of sources of error such as barcode contamination and sample cross-talk during PCR and sequencing were also quantified. We found that using unique dual sample indexes in multiplexed sequencing runs was essential to suppress these sources of noise. Applying these aggregated methods to analyze plasma cfDNA samples obtained from patients across a range of solid tumor types and disease stages, we were able to reliably detect clinically relevant mutations with variant allele fractions below 0.003, including subclonal mutations associated with acquired drug resistance. This approach, when applied prospectively on clinical specimens, has the potential to facilitate diagnosis, prognosis, and treatment selection in an era of precision oncology. Citation Format: Juber Patel, Maysun Hasan, Fanli Meng, Xiaohong Jing, Dilmi Perera, Jonathan Reichel, Erika Gedvilaite, Julie Yang, Maha Shady, Sandeep Raj, Preethi Srinivasan, Ian Johnson, Jiashi Wang, Mirna Jarosz, Aliaksandra Samoila, Agnes Viale, Bob Li, Pedram Razavi, Dana Tsui, Michael Berger. Development and optimization of a comprehensive high-sensitivity NGS cancer assay and bioinformatics pipeline for plasma cfDNA profiling [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 5598.