PO.MCB05.02 · 分子与细胞生物学

Comprehensive & precise structural variant detection with a single assay

海报缩略图:Comprehensive & precise structural variant detection with a single assay
编号 517 展板 8 时间 4/19 02:00–05:00 区域 Section 21 主讲 John Sanborn
分会场 Mechanisms and Targets in DNA Damage Repair
查看完整资料 下载 PDF 登录后可访问当前开放资料 AACR 官方页面 ↗

作者与单位

J. Zachary Sanborn, James Durbin, Mital Bhakta, Lisa Munding

Cantata Bio, LLC, Scotts Valley, CA

摘要 Abstract

Structural variants (SVs) are increasingly recognized as key drivers of tumorigenesis, yet current technologies struggle to detect and accurately characterize them due to resolution limitations and sequencing constraints. As a result, attempts to produce complete somatic SV truth sets often suffer due to the gaps in detection performance of the sequencing technology utilized, impacting the truth sets' usefulness for benchmarking alternative technologies and/or SV callers. To this end, the National Institute of Standards & Technology (NIST) performed a deep benchmarking experiment for somatic SV detection, fully characterizing the structural changes of a pancreatic ductal carcinoma (PDAC, labeled HG008T) across 16 whole genome sequencing technologies and analyzing with multiple datatype-specific SV callers. NIST has published its Benchmark SV callset that is the result of integrating and manually curating calls from all analytes. With this benchmarking data, we tested the capability of Dovetail's LinkPrep chemistry alone to reproduce the NIST Benchmark callset using both open source and proprietary SV callers. Dovetail LinkPrep libraries were generated for the HG008T cell line sample and sequenced to approximately 55x genomic coverage. Sequencing data was aligned to hg38 with bwa and SV calls generated by running a standard, HiC-based SV caller, HiC-Breakfinder, and Dovetail Precise (DP), a new SV caller developed by Dovetail Genomics. These callsets were separately compared to the NIST Benchmark callset to assess their performance in categories based on SV type and size. Of the 109 total SVs in the NIST Benchmark callset, HiC-Breakfinder and DP detected 19 (18%) and 97 (92%) SVs, respectively. HiC-Breakfinder's ability to detect SVs was limited to translocations (8 of 12, 67%) and large intra-chromosomal SVs greater than 2 Mb in size (10 of 11, 91%), while unable to find any intra-chromosomal SV smaller than 2 Mb. In contrast, DP detected all translocations (12 of 12, 100%), the same number of large intra-chromosomal SVs (10 of 11, 91%), and the majority of smaller intra-chromosomal SVs (75 of 85, 88%) down to 47 bp in size. Furthermore, the difference in breakend locations between HiC-Breakfinder and the NIST Benchmark SVs ranged from 1,636 bp to 189 kb, while over 96% of DP calls (94 of 97) achieved base-level precision. Dovetail's LinkPrep chemistry with the Dovetail Precise (DP) caller detected 92% of somatic structural variants (SVs) in the NIST Benchmark callset, including small variants down to 47 bp, strongly outperforming HiC-Breakfinder across all SV types and sizes. This high-resolution SV detection enables more comprehensive cancer genome characterization from a single assay and supports more accurate diagnostics and targeted therapy development.
利益披露 Disclosure
J. Z. Sanborn, Cantata Bio, LLC Employment, Stock. J. Durbin, Cantata Bio, LLC Employment, Stock Option. M. Bhakta, Cantata Bio, LLC Employment, Stock Option. L. Munding, Cantata Bio, LLC Employment, Stock Option.

在会议检索中打开