基于长读长杂交捕获对95个已知癌症基因进行靶向，通过简单的生物信息学流程和基于AI的临床解读方案检测大型结构变异和复杂变异

Short read sequencing has driven progress in identifying cancer-causing mutations ranging from SNPs to small indels. However, the ability to define complex mutations that span multiple kilobases is limited with traditional NGS approaches due to short read lengths. Targeted long-read sequencing, on the other hand, enables characterization of novel, complex mutations as long reads can span and be phased across large genomic regions. Long-read hybrid capture offers probe design flexibility for capturing targeted regions with unresolved structural variants. Here we describe the development of QIAseq xHYB Long Read Hereditary Cancer Panel and chemistry for detection of large structural variants in 95 known cancer driver genes.QIAseq xHYB Long Read Hereditary Cancer Panel and analysis pipelines were used to identify large structural variants in genes involved in cancer progression that human reference DNA were known to harbor. Libraries were prepared using enzymatic long-read fragmentation, after which targeted regions were captured with probes optimized for long DNA fragments. Captured DNA was amplified with chemistry developed for long, fast PCR and sequenced on both PacBio and Oxford Nanopore platforms. Real-time adaptive sampling was also used on the Oxford Nanopore platform to increase sequencing depth and improve uniformity. The resulting long-read sequencing data was mapped and large structural variants detected with CLC Genomics Workbench Lightspeed Module and Franklin by QIAGEN, a cloud-based, AI-powered platform that integrates the world's first open genomic community to power precision medicine at scale.The QIAseq xHYB Long Read probe design targets entire genes, including introns and UTR's, capturing extended regions of genes in an unbiased manner. Enzymatic fragmentation produced libraries with average read lengths of 4.5 kb. Sequencing eight QIAseq xHYB Hereditary Cancer Panel libraries from one capture pool on a Revio SMRTcell yields 30X average coverage, and uniformity greater than 95% of reads > 0.2X of the mean. Sequencing the same libraries on the Oxford Nanopore Minion platform resulted in 12X average coverage, and uniformity greater than 90% of reads > 0.2X of the mean. To increase Minion sequencing depth, we used real-time adaptive sampling to enrich for our target regions, which resulted in average of 30X coverage, and 93% of reads > 0.2X of the mean for 8 libraries on the Nanopore platform. Downstream analysis with CLC Genomics Workbench Lightspeed Module clearly identified the expected large structural variants in Coriell reference DNA, and subsequent clinical interpretation performed by AI-powered Franklin produced informative actionable conclusions from the structural variant reference DNA investigated.