一种具有高灵敏度和特异性的全基因组肿瘤知情分子残留病检测方法的分析学评估

Introduction: Molecular residual disease (MRD) testing involves the detection of circulating tumor DNA (ctDNA) in plasma after cancer treatment. Tumor-informed Whole Genome Sequencing (WGS) workflows for the qualitative detection of MRD have been shown to be highly sensitive. We present an automated and scalable solution for patients with bladder, breast, melanoma, non-small cell lung (NSCLC), and colorectal (CRC) cancers with turnaround as fast as 8 to 9 days. This includes automation of DNA extraction, library preparation, sequencing, data analyses, and reporting. Here, we summarize the analytical performance of a research use only workflow intended as a service for research and clinical partners, such as pharmaceutical companies. Methods: The workflow requires 3 different sample types: formalin-fixed, paraffin-embedded (FFPE), Buffy Coat, and 2 to 4ml plasma aliquots. The optimal inputs are: i) 5ng (minimum 2ng) cell-free DNA (cfDNA) from plasma, ii) 100ng (minimum 50ng) tumor tissue DNA, and iii) 50ng DNA from Buffy Coat. DNA is extracted from different sample types with WGS libraries preparation and sequencing on NovaSeq TM 6000. Sequencing analyses, fingerprint generation, and MRD detection are performed using DRAGEN TM . The assay uses a patient's tumor and germline sample to generate a patient-specific somatic variant list, i.e. fingerprint. Subsequently, the cfDNA sequence data is evaluated against the fingerprint to determine the presence or absence of tumor DNA indicative of MRD. Both clinical and contrived samples were used to evaluate accuracy, analytical sensitivity, analytical specificity and precision. Results: Twenty-five clinical samples (5 samples for each cancer type) resulted in 100% overall percent agreement (OPA) in MRD status relative to a reference method, with ctDNA concentration as low as 0.035%. A panel of 20 healthy plasma samples assessed against 25 fingerprints across multiple cancer types resulted in a 100% negative percent agreement. Sensitivity analysis demonstrated 100% detection rate at variant allele frequency of 0.003% for NSCLC, 0.005% for bladder and CRC, and 0.006% for melanoma and breast samples. Precision was evaluated using all specimen types across multiple operators, instruments, and library preparation start days. There was 100% OPA in MRD status across intra-run and inter-run replicates at ctDNA concentrations of 0.009% to 0.018%. Conclusions: We present a whole genome fully automated workflow capable of generating tumor- informed MRD status with high analytical sensitivity and specificity. The assay requires low DNA input from plasma samples and enables a sample-to-report turnaround time as fast as 8 to 9 days. The results demonstrate that the WGS MRD detection platform is a scalable, robust, and highly sensitive assay for multiple cancer types.