Pre-capture multiplexing provides additional power to detect copy number variation in exome sequencing

• Published in: BMC bioinformatics 2021-07, Vol.22 (1), p.1-374, Article 374
• Main Authors: Filer, Dayne L, Kuo, Fengshen, Brandt, Alicia T, Tilley, Christian R, Mieczkowski, Piotr A, Berg, Jonathan S, Robasky, Kimberly, Li, Yun, Bizon, Chris, Tilson, Jeffery L, Powell, Bradford C, Bost, Darius M, Jeffries, Clark D, Wilhelmsen, Kirk C
• Format: Article
• Language: English
• Subjects: Algorithms; Bayesian analysis; Capture; Comparative analysis; Copy number; Copy number variation; Copy number variations; Data collection; Data mining; DNA sequencing; Efficiency; Exome sequencing; Gene sequencing; Genetic research; Genomes; Information processing; Laboratories; Methods; Multiplexing; Nucleotide sequencing; Probability; Sample variance
• ISSN: 1471-2105; 1471-2105
• DOI: 10.1186/s12859-021-04246-w

As exome sequencing (ES) integrates into clinical practice, we should make every effort to utilize all information generated. Copy-number variation can lead to Mendelian disorders, but small copy-number variants (CNVs) often get overlooked or obscured by under-powered data collection. Many groups have developed methodology for detecting CNVs from ES, but existing methods often perform poorly for small CNVs and rely on large numbers of samples not always available to clinical laboratories. Furthermore, methods often rely on Bayesian approaches requiring user-defined priors in the setting of insufficient prior knowledge. This report first demonstrates the benefit of multiplexed exome capture (pooling samples prior to capture), then presents a novel detection algorithm, mcCNV ("multiplexed capture CNV"), built around multiplexed capture. We demonstrate: (1) multiplexed capture reduces inter-sample variance; (2) our mcCNV method, a novel depth-based algorithm for detecting CNVs from multiplexed capture ES data, improves the detection of small CNVs. We contrast our novel approach, agnostic to prior information, with the the commonly-used ExomeDepth. In a simulation study mcCNV demonstrated a favorable false discovery rate (FDR). When compared to calls made from matched genome sequencing, we find the mcCNV algorithm performs comparably to ExomeDepth. Implementing multiplexed capture increases power to detect single-exon CNVs. The novel mcCNV algorithm may provide a more favorable FDR than ExomeDepth. The greatest benefits of our approach derive from (1) not requiring a database of reference samples and (2) not requiring prior information about the prevalance or size of variants.