panelcn.MOPS reaches clinical standards as a copy number variation detection tool for targeted panel sequencing

• Verena Haunschmid

Targeted panel sequencing is becoming increasingly important as a cost-effective strategy to identify disease-causing variants in clinical and research applications. While various copy number variation (CNV) detection methods exist for wholegenome and whole-exome sequencing data, highly accurate methods for panel sequencing data that are suitable for clinical purposes are still missing. The challenges with this kind of data are the small size and number of target regions as well as their uneven coverage. For clinical applications a method should furthermore be able to detect both short CNVs, affecting only single exons or even just parts thereof, as well as longer CNVs that affect multiple exons or even an entire gene.

We present panelcn.MOPS for copy number detection which extends our previously developed method cn.MOPS to targeted panel sequencing data. The method is well suited for this type of data since it can estimate technical and biological characteristics inuencing the read counts of each targeted region by a mixture of Poissons model. The design of the count windows, the read counting procedure, the parameters of the model and the segmentation algorithm have been optimized for targeted panel sequencing. panelcn.MOPS supplies integer copy numbers. Additionally we developed a user-friendly and feature rich graphical user interface to make it easy to run the analysis.

We have tested panelcn.MOPS on simulated and real sequencing data. On 150 simulated data sets, that resembled the characteristics of targeted panel sequencing data, panelcn.MOPS has reached an average accuracy of 99.96%. The real sequencing data was enriched with the TruSight cancer panel that targets 94 cancer predisposition genes including NF1/2, BRCA1/2 and APC. panelcn.MOPS detected 87.5% of CNVs known from previous MLPA analyses without any false positives and an accuracy of 93.1%. The size of the CNVs ranged from an 80bp deletion starting in the intron and a_ecting only part of one exon over duplications of several exons to deletions of 350kb a_ecting the entire gene.

These results show that CNVs in targeted panel sequencing data can accurately be predicted with panelcn.MOPS. Consequently additional biotechnologies to detect CNVs, such as MLPA, can be omitted in order to reduce time and costs.