Improving Reliability of Copy Number Variations Detection

DNA copy number variations (CNVs) and their changes play an important role in biological processes, including well-established roles in cancer and in developmental disorders. All methods detecting CNVs in next-generation... [ view full abstract ]

DNA copy number variations (CNVs) and their changes play an important role in biological processes, including well-established roles in cancer and in developmental disorders.

All methods detecting CNVs in next-generation sequencing data are based on analyzing, either explicitly or implicitly, how coverage varies across analyzed regions, in order to model the coverage overdispersion. However, CNVs are not the only contributor to coverage overdispersion, i.e. the coverage overdispersion caused by the presence of CNVs is mixed with the overdispersion caused by biases introduced at various steps of sequencing library preparation, systematic sequencing errors as well as artefacts of mapping to the reference genome.  Therefore, despite significant progress in the development of tools that detect CNVs in sequencing reads, no single tool detects all types of CNVs.

We developed a data mining method to identify coverage variability for these non-CNV effects by mapping artefactual variability on a reference genome of interest. The corrected coverage distribution will be used for efficient and reliable CNV detection, while testing the reduction of overdispersion provides a very stringent validation criterion.

The method has great potential to advance the analysis of pan-genome CNV presence and to determine whether particular CNVs represent a population or a private variant.