Comparison of different copy number variations (CNVs) detection tools using whole genome sequencing (WGS) data

Retinitis Pigmentosa (RP) encompasses a group of inherited diseases of the retina (IRD) associated with progressive dysfunction of the rods and/or cones, leading to loss of vision. There is a great genetic heterogeneity within IRD with more than 250 genes identified so far, which has made its genetic-molecular characterization a challenge. A significant percentage of genome variability is caused by copy number variations (CNVs), which encompass unbalanced rearrangements that increase or decrease DNA content. Whole genome sequencing (WGS) is becoming a viable and sensitive method for the detection of all types of genetic variability. There are four main CNV detection methodologies from WGS data: read-depth (RD), split-read (SR), read-pair (RP) and de novo assembly (AS). In recent years there has been a bloom of CNV detection tools based on WGS data; with more than 70 available methods currently available. The aim of this study is to evaluate the performance of seven bioinformatic tools to detect CNVs from WGS data, in terms of sensitivity and accuracy. This will allow us to determine the most suitable tool for its use in the genetic diagnosis of RP, based on WGS data. We evaluated BreakDancer, CNVpytor, Delly, Lumpy, Manta, Pindel and SvAba with twenty samples from the Human Genome Structural Variation Consortium, Phase 2 (HGSVC2). We found Lumpy to be the best performing tool for the detection of genomic losses. However, its performance at detecting genomic gains was not as good, with the best performing tool being Manta.