Abstract 3521: Critical Assessment of CNV Calling Using Next Generation Sequencing

Abstract

Abstract Tumor-derived Copy Number Variations (CNVs) are among the most important genomic aberrations in cancers. Since a CNV amplification is often associated with oncogene activation, and a CNV deletion is frequently attributed to the gene inactivation, an accurate detection of CNVs plays a crucial role in cancer diagnosis and treatment. In this study, using the next generation sequencing, we systematically interrogated somatic CNVs in paired tumor-normal cell lines to identify factors affecting detection reproducibility and accuracy. Whole-genome and whole-exome sequencing were carried out at six sequencing centers followed by processing with six CNV callers to evaluate their reproducibility. Different types of samples with varying input amount and tumor purity were processed using multiple library construction protocols and read depths. We compared the consistency of reference replicates, the effect of experimental, technical, and biological confounding factors, and benchmarked our callers using cytogenetic array. Our evaluation was based on two approaches; bin-based and gene-based approach. The bin-based comparison used a sliding window of 10 Kb, while gene-based comparison used segments at the gene level. Our evaluations indicate variations among CNV calls are mainly driven by callers and less by the sequencing centers. In addition to caller effect, PCA analyses also show site-specific effect, largely pairwise distances on the PC plot for pairs between different centers than for those within the same centers. Among confounding factors, namely FFPE, library concentration, and tumor purity with several sequencing depths, we report that tumor purity has a dominant and major effect in caller’s performance. PCA analysis shows low tumor purity (<= 50%) has a large effect on CNV calls which are evidenced by low purity samples being away from reference samples on PC plots for all callers. Tumor purity <= 50% leads to severe missing calls and lower copy numbers in amplified regions while sequencing depth seems to affect the outcome. We illustrated that most of the other confounding factors such as WES, FFPE, and LBP are caller-specific. Comparing to other factors, the effect of low sequence depth seems to be minimal. FFPE also affects CNV calls by increased variability in several callers. Consensus call and slightly stringent cutoffs for calling amplification/deletion can produce high confidence calls. Comparing to cytogenetic array, we found high correlations between the array and WGS segments in some callers while other callers show large variability. Citation Format: Yun-Ching Chen, Fayaz Seifuddin, Cu Nguyen, Chunhua Yan, Qingrong Chen, Wenming Xiao, Mehdi Pirooznia, Daoud Meerzaman. Critical Assessment of CNV Calling Using Next Generation Sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3521.