Evaluating the quality of Marfan genotype–phenotype correlations in existing FBN1 databases

Background: Genetic FBN1 testing is pivotal for confirming the clinical diagnosis of Marfan syndrome. In an effort to evaluate variant causality, FBN1 databases are often used. We evaluated the current databases regarding FBN1 variants and validated associated phenotype records with a new Marfan syndrome geno-phenotyping tool called the Marfan score. Methods and results: We evaluated four databases (UMD-FBN1, ClinVar, the Human Gene Mutation Database (HGMD), and Uniprot) containing 2,250 FBN1 variants supported by 4,904 records presented in 307 references. The Marfan score calculated for phenotype data from the records quantified variant associations with Marfan syndrome phenotype. We calculated a Marfan score for 1,283 variants, of which we confirmed the database diagnosis of Marfan syndrome in 77.1%. This represented only 35.8% of the total registered variants; 18.5–33.3% (UMD-FBN1 versus HGMD) of variants associated with Marfan syndrome in the databases could not be confirmed by the recorded phenotype. Conclusion: FBN1 databases can be imprecise and incomplete. Data should be used with caution when evaluating FBN1 variants. At present, the UMD-FBN1 database seems to be the biggest and best curated; therefore, it is the most comprehensive database. However, the need for better genotype–phenotype curated databases is evident, and we hereby present such a database. Genet Med advance online publication 01 December 2016