Evaluation of computational genotyping of Structural Variations for clinical diagnoses.

BackgroundIn recent years, Structural Variation (SV) has been identified as having a pivotal role in causing genetic disease. Nevertheless, the discovery of SVs based on short DNA sequence reads from next-generation DNA sequence methods is still error-prone, suffering from low sensitivity and high false discovery. These shortcomings can be partially overcome with the use of long reads, but the current expense precludes their application for routine clinical diagnostics. Structural Variation genotyping, on the other hand, offers cost-effective application as diagnostic tool in the clinic, with potentially no false positives and low occurrence of false negatives. ResultsWe assess five state- of-the- art SV genotyping software methods that test short read sequence data. These methods are characterized based on their ability to genotype certain SV types and size ranges. Furthermore, we analyze their applicability to parse different VCF file sub-formats, or to rely on specific meta information that is not always at hand. We compare the SV genotyping methods across a range of simulated and real data including SVs that were not found with Illumina data alone. We assess their sensitivity and ability to filter out initial false discovery calls to assess their reliability. ConclusionOur results indicate that, although SV genotypers have superior performance to SV callers, there are performance limitations that suggest the need for further innovation.