A fully automated walkaway system programmed with a novel library prep target enrichment protocol for detection of ultra-low frequency mutations in liquid biopsy and tumor samples

Liquid biopsy has transformed oncology through high sensitivity detection of genomic alterations from circulating cell-free DNA (cfDNA). Issues such as sensitivity, PCR bias, contamination, reproducibility, and long and laborious workflows are yet to be resolved. Here, we have fully automated the workflow for a novel library preparation and target enrichment probe technology, Agilent Avida, that tackles some of the existing challenges by being highly optimized to work with circulating tumor DNA (ctDNA). Our automated protocol using pre-aliquoted reagents, was developed on the Agilent Magnis NGS Prep System for a wide range of DNA inputs (1-100 ng) and various sample types (intact, FFPE DNA, and cfDNA). Using this platform, we construct libraries with a high sample recovery that scales linearly with the input amount and is capable of high sensitivity variant detection without a need for pre-enrichment PCR, minimizing the GC bias. Our automated protocol generates up to eight target-enriched Illumina sequencing-ready DNA libraries in about 8 hours without the need of any user intervention, thus minimizing potential contamination. We have generated data using both Avida catalog and custom panels covering a broad range of target sizes to determine assay sensitivity. Using these probes, we were able to reliably detect cancer-associated genomic alterations, including SNVs, indels, CNVs, and translocations across key oncogenes using well characterized cfDNA and formalin compromised reference standards as well as real ctDNA samples. For example, using 15 ng of SeraCare V4 ctDNA sample enriched with Avida DNA Onco LB panel (1.17 Mb covering 164 pan-cancer-associated genes) and sequenced with a budget of 100 M read pairs, we were able to detect SNP frequencies down to 0.25%. Furthermore, in our automated runs, we obtained high reproducibility across 8 technical replicates. For example, for 5 independent Magnis runs (40 samples) of various sample types and inputs per run using the Avida DNA Onco LB panel, the % Coefficient of Variation (CV) ranged from 0.8 to 2.3% for on-target rate, 1.9 to 8.2% for library complexity (unique molecular identifier UMI recovery), 1 to 3.2% for base coverage, and 0.03 to 0.09% for uniformity.