Evaluation of three read-depth based CNV detection tools using whole-exome sequencing data

Abstract

Background: Whole exome sequencing (WES) has been widely accepted as a robust and cost-effective approach for clinical genetic testing of small sequence variants. Detection of copy number variants (CNV) within WES data have become possible through the development of various algorithms and software programs that utilize read-depth as the main information. The aim of this study was to evaluate three commonly used, WES read-depth based CNV detection programs using high-resolution chromosomal microarray analysis (CMA) as a standard. Methods: Paired CMA and WES data were acquired for 45 samples. A total of 219 CNVs (size ranged from 2.3 kb - 35 mb) identified on three CMA platforms (Affymetrix, Agilent and Illumina) were used as standards. CNVs were called from WES data using XHMM, CoNIFER, and CNVnator with modified settings. Results: All three software packages detected an elevated proportion of small variants (< 20 kb) compared to CMA. XHMM and CoNIFER had poor detection sensitivity (22.2 and 14.6%), which correlated with the number of capturing probes involved. CNVnator detected most variants and had better sensitivity (87.7%); however, suffered from an overwhelming detection of small CNVs below 20 kb, which required further confirmation. Size estimation of variants was exaggerated by CNVnator and understated by XHMM and CoNIFER. Conclusion: Low concordances of CNV, detected by three different read-depth based programs, indicate the immature status of WES-based CNV detection. Low sensitivity and uncertain specificity of WES-based CNV detection in comparison with CMA based CNV detection suggests that CMA will continue to play an important role in detecting clinical grade CNV in the NGS era, which is largely based on WES.