Evaluating variants classified as pathogenic in ClinVar in the DDD Study

Caroline F Wright,Ruth Y Eberhardt,Panayiotis Constantinou,Matthew E Hurles,David R Fitzpatrick,Helen V Firth

doi:10.1038/s41436-020-01021-9

Abstract

PurposeAutomated 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.MethodsWe 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.ResultsOf 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.ConclusionLower 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.

Highlights

Over the past decade, the diagnosis of rare pediatric diseases has been transformed by the application of next-generation sequencing.[1]
Using exome sequence data from the Deciphering Developmental Disorders (DDD) study,[8] here we show that around 16% of potentially relevant previously reported pathogenic variants are excluded by the standard DDD clinical variant filtering pipeline.[9]
Variants were evaluated for clinical feedback using a curated developmental disorder gene-to-phenotype database (DDG2P)[12] and a bespoke series of variant filtering rules described previously[7,9]; in brief, variants are excluded based on minor allele frequency (MAF), predicted consequence, and genotype or inheritance inconsistent with either the family history or the allelic requirement of the DDG2P gene–disease pair

Summary

Introduction

The diagnosis of rare pediatric diseases has been transformed by the application of next-generation sequencing.[1]. Hard thresholds are usually applied to exclude all but a small minority of variants, without which diagnostic services would be overwhelmed by false positive results, especially for highly genetically heterogeneous and incompletely penetrant disorders. These hard cutoffs can sometimes inadvertently exclude important diagnoses, leading to false negative results

Methods

Results

Discussion

Conclusion