Our results are also important for the clinical diagnostics field. Typically, the assessment of data in clinical laboratories is focused on larger CNVs (different thresholds are used in different laboratories). We therefore performed a more detailed analysis of CNVs >50 kb. To our surprise, we found that the lack of overlap between platforms, algorithms and replicates that was found in the full data set similarly applied to large CNVs. A closer look at these regions indicates that most can be explained by overlap with complex regions and call fragmentation. In standard clinical assessment of patient data, curation of results and filtering of polymorphic regions would lead to removal of these variants. We therefore do not think that our data contradict previous reports of high accuracy in detecting clinically significant rearrangements in patients across different laboratories and array types. However, our results bring light to the problems of clinical interpretation of variants in complex regions and highlight the risks of incorporating less stringent filtering of data in diagnostics.
