A Systematic Re-Analysis Of Copy Number Losses Of Uncertain Clinical Significance

BackgroundRe-analysis of whole exome/genome data improves diagnostic yield. However, the value of re-analysis of clinical array comparative genomic hybridisation (aCGH) data has never been investigated. Case-by-case re-analysis is impractical in busy diagnostic laboratories. Methods and ResultsWe harmonised historical post-natal clinical aCGH results from [~]16,000 patients tested via our diagnostic laboratory over [~]7 years with current clinical guidance. This led to identification of 33,857 benign, 2,173 class 3, and 979 pathogenic copy number losses (CNLs). We found benign CNLs to be significantly less likely to encompass haploinsufficient genes compared to the pathogenic or class 3 CNLs in our database. Using this observation, we developed a re-analysis pipeline (using up-to-date disease association data and haploinsufficiency scores) and shortlisted 207 class 3 CNLs encompassing at least one autosomal dominant disease-gene associated with haploinsufficiency or loss-of-function mechanism. Clinical scientist review led to reclassification of 7.2% shortlisted class 3 CNLs as pathogenic or likely pathogenic. This included first cases of CNV-mediated disease for some genes where all previously described cases involved only point variants. Interestingly, some CNLs could not be re-classified because the phenotypes of patients with CNLs seemed distinct from the known clinical features resulting from point variants, thus raising questions about accepted underlying disease mechanisms. Several potential novel disease-genes were identified that would need further validation. ConclusionsRe-analysis of clinical aCGH data increases diagnostic yield and demonstrates their research value. In future, the aCGH reanalysis program should be expanded to include other copy number variant types.