Evaluating variants classified as pathogenic in ClinVar in the DDD Study

Abstract

Purpose: Automated variant filtering is an essential part of diagnostic genome-wide sequencing but may generate false negative results. We sought to investigate whether some previously identified pathogenic variants may be being routinely excluded by standard variant filtering pipelines.

Methods: We evaluated variants that were previously classified as pathogenic or likely pathogenic in ClinVar in known developmental disorder genes using exome sequence data from the Deciphering Developmental Disorders (DDD) study.

Results: Of these ClinVar pathogenic variants, 3.6% were identified among 13,462 DDD probands, and 1134/1352 (83.9%) had already been independently communicated to clinicians using DDD variant filtering pipelines as plausibly pathogenic. The remaining 218 variants failed consequence, inheritance, or other automated variant filters. Following clinical review of these additional variants, we were able to identify 112 variants in 107 (0.8%) DDD probands as potential diagnoses.

Conclusion: Lower minor allele frequency (<0.0005%) and higher gold star review status in ClinVar (>1 star) are good predictors of a previously identified variant being plausibly diagnostic for developmental disorders. However, around half of previously identified pathogenic variants excluded by automated variant filtering did not appear to be disease-causing, underlining the continued need for clinical evaluation of candidate variants as part of the diagnostic process.

Keywords: developmental disorders; exome sequencing; genomic medicine; reanalysis; variant interpretation.

Fig. 1. Flowchart of variant selection method and results.

DDD Deciphering Developmental Disorders study, DDG2P developmental disorder gene-to-phenotype database, ES exome sequencing, LP likely pathogenic.