Combined Bioinformatic and Splicing Analysis of Likely Benign Intronic and Synonymous Variants Reveals Evidence for Pathogenicity

Purpose Clinical variant analysis pipelines likely have poor sensitivity to the effects on splicing from variants beyond 10-20 bases of exon-intron boundaries. Here, we demonstrate the value of SpliceAI to inform curation of rare variants previously classified as Benign/Likely Benign (B/LB) under current guidelines. Methods Exome sequencing data from 576 pediatric cancer patients enrolled in the Texas KidsCanSeq study were filtered for intronic or synonymous variants absent from population databases, predicted to alter splicing via SpliceAI (>0.20), and scored >10 by Combined Annotation Dependent Depletion. Rare synonymous or intronic B/LB variants in 61 genes submitted to ClinVar were also evaluated and RNA further assessed in monocyte-derived mRNA and/or an in vitro splice reporter assay in HEK-293T cells. Results SpliceAI-supplemented analysis of the KidsCanSeq cohort revealed a DICER1 intronic variant that resulted in missplicing in RNA from a proband with a personal and family history of pleuropulmonary blastoma but negative clinical exome and panel reports. Analysis of 34,188 B/LB ClinVar variants yielded 18 variants predicted to cause disrupted reading frames. Assessment of 8 variants (DICER1 n=4, CDH1 n=2, PALB2 n=2) by in vitro splicing assay demonstrated abnormal splice products (mean 66%; range 6% to 100%). When available, phenotypic information from submitting laboratories demonstrated DICER1-associated tumors in 2 families (1 variant) and breast cancer in 3 families (2 PALB2 variants). Conclusions Incorporation of SpliceAI in variant curation pipelines may improve classification of B/LB intronic and synonymous variants and highlight putative pathogenic variants for functional assays and RNA analysis, thereby increasing diagnostic yield for rare diseases.