Semi-automation of ClinVar variants
Of 4948 ClinVar variants in 142 deafness-related genes, VIP-HL achieved an overall variant interpretation concordance of 88.0% (4353/4948). The concordant interpretations were 57.1% (376/658) in P/LP variants, 93.6% (3083/3295) in B/LB variants, and 89.8% (894/995) in VUS variants, respectively (Table 2). The lower concordance in P/LP variants could be expained that VIP-HL is a semi-automated tool and only automated 13 out of 24 ACMG/AMP rules.
Of note, VIP-HL generated three P/LP classifications versus B/LB compared to ClinVar.The first one was NM_153676.3(USH1C ):c.2547-1G>T. It was submitted as likely benign in ClinVar, whereas VIP-HL assigned PVS1 and PM2, leading to a likely pathogenic classification. Manual curation revealed that this splicing variant affected an exon in transcript NM_153676.3 with no detectable expression based on the Genotype-Tissue Expression (GTEx) database (GTEx Consortium, 2017). Thus, this variant should be classified as benign/likely benign. This example highlights the need for including information about the most biologically relevant transcripts in VIP-HL for accurate clinical variant interpretation (DiStefano et al., 2018).
The other two discrepant variants were synonymous (NM_206933.3(USH2A ):c.949C>A (p.Arg317=) and NM_022124.6(CDH23 ):c.7362G>A (p.Thr2454=). Both were assigned likely benign classifications by VIP-HL due to the activation of BP4 and BP7. Specifically, NM_206933.3:c.949C>A was submitted as a pathogenic variant in ClinVar. This variant led to abnormal splicing and a premature termination codon (Vaché et al., 2010), strongly supporting its pathogenic classification. As shown below, our VIP-HL user interface will enable curators to manually activate codes for functional studies to avoid such possible misclassifications. The second variant, NM_022124.6(CDH23 ):c.7362G>A, was submitted as likely pathogenic. No supporting RNA/functional, case-level or segregation studies were provided to support this classification. Since the actual effect of this sequence change is still unknown, a likely benign classification may be more appropriate for this case.
Of 4948 ClinVar variants, the most utilized rules were population-related, spanning PM2, BA1, BS1, BS2 and their modified strength levels (supporting through very strong; Figure 2). BP4 and BP7 are also frequently used because 33.5% (1657/4948) deafness-related variants with ClinVar star 2+ were synonymous variants. As expected, the pathogenic criteria were enriched in pathogenic/likely pathogenic variants, whereas benign criteria were enriched in benign/likely benign variants (Figure 2).