Presentation Information
[PE2-9]In silico strategy for the identification of deep-intronic variants causing aberrant splicing
○Ryo Kurosawa1, Masahiko Ajiro1, Kei Iida1,2, Tomonari Awaya1, Mamiko Yamada3, Kenjiro Kosaki3, Masatoshi Hagiwara1 (1.Kyoto University Graduate School of Medicine, 2.Kindai University Faculty of Science and Engineering, 3.Center for Medical Genetics, Keio University School of Medicine)
Despite recent efforts to apply next-generation sequencing for genetic disease diagnosis, detection rates of causal variants are still limited to 50-75%. This is assumingly attributed to the current lack of a practical schema to predict pathogenic non-coding variants. Here, we propose a diagnostic schema for deep-intronic splicing-associated variants (SAVs), using a splicing predictor, SpliceAI. We comprehensively collected pathogenic deep-intronic SAVs (n=370) to measure the SpliceAI sensitivity. Subsequently, we evaluate its clinical utility by analyzing 2,504 whole-genome sequencing (WGS) samples from 1000 Genomes Project, and validating a subset of SAV candidates (n=10,801) in immortalized B cells. These approaches revealed that a 0.20 score threshold of SpliceAI detected 90% of pathogenic deep-intronic SAVs. This threshold is clinically applicable because it predicted about 37 SAV candidates per individual, whose number is small enough for clinicians to perform manual reviews for causal variants. A 0.76 threshold (40% sensitivity) is also recommended because of its high precision of 58 %. Lastly, we applied our approach to a patient previously diagnosed as beta-propeller protein-associated neurodegeneration and successfully detected the causal deep-intronic SAV in WDR45 gene. In conclusion, this diagnostic schema will shed a light on previously overlooked deep-intronic SAVs in genetic diseases. The study was approved by the institutional review boards. After written consent was obtained, peripheral blood samples were collected from each subject and the parents.