AbstractCopy number variation (CNV) is a common type of mutation that often drives cancer progression. With advances in next-generation sequencing (NGS), CNVs can be detected in a detailed manner via newly developed computational tools but quality of such CNV calls has not been carefully evaluated. We analyzed CNV calls reported by 6 cutting-edge callers for 91 samples which were derived from the same cancer cell line, prepared and sequenced by varying the following factors: type of tissue sample (Fresh vs. Formalin Fixed Paraffin Embedded (FFPE)), library DNA amount, tumor purity, sequencing platform (Whole-Genome Sequencing (WGS) versus Whole-Exome Sequencing (WES)), and sequencing coverage. We found that callers greatly determined the pattern of CNV calls. Calling quality was drastically impaired by low purity (<50%) and became variable when WES, FFPE, and medium purity (50%-75%) were applied. Effects of low DNA amount and low coverage were relatively minor. Our analysis demonstrates the limitation of benchmarking somatic CNV callers when the real ground truth is not available. Our comprehensive analysis has further characterized each caller with respect to confounding factors and examined the consistency of CNV calls, thereby providing guidelines for conducting somatic CNV analysis.
Computational tools for copy number variation (CNV) detection using next-generation sequencing data: features and perspectives
