Validation and benchmarking of targeted panel sequencing for cancer genomic profiling.

Duo Wang,Sizhen Wang,Yuan Zhang,Xiaolei Cheng,Xin Huang,Yanxi Han,Zhaohui Chen,Cong Liu,Jinming Li,Rui Zhang
DOI: https://doi.org/10.1093/ajcp/aqad078
2023-07-21
American Journal of Clinical Pathology
Abstract:OBJECTIVES To validate a large next-generation sequencing (NGS) panel for comprehensive genomic profiling and improve patient access to more effective precision oncology treatment strategies. METHODS OncoPanScan was designed by targeting 825 cancer-related genes to detect a broad range of genomic alterations. A practical validation strategy was used to evaluate the assay's analytical performance, involving 97 tumor specimens with 25 paired blood specimens, 10 engineered cell lines, and 121 artificial reference DNA samples. RESULTS Overall, 1107 libraries were prepared and the sequencing failure rate was 0.18%. Across alteration classes, sensitivity ranged from 0.938 to more than 0.999, specificity ranged from 0.889 to more than 0.999, positive predictive value ranged from 0.867 to more than 0.999, repeatability ranged from 0.908 to more than 0.999, and reproducibility ranged from 0.832 to more than 0.999. The limit of detection for variants was established based on variant frequency, while for tumor mutation burden and microsatellite instability, it was based on tumor content, resulting in a minimum requirement of 20% tumor content. Benchmarking variant calls against validated NGS assays revealed that variations in the dry-bench processes were the primary cause of discordances. CONCLUSIONS This study presents a detailed validation framework and empirical recommendations for large panel validation and elucidates the sources of discordant alteration calls by comparing with "gold standard measures."
Medicine
What problem does this paper attempt to address?