Search

Additional file 4: Figure S1. Carrier frequency (left panel) and NC P/LPs count (right panel) of autosomal recessive (AR) and autosomal dominant (AD) genes across ancestries. Figure S2. Validation of identified NC P/LPs in TCGA by respective ancestral population in gnomAD. Figure S3. Validation of genes impacted by identified NC P/LPs in TCGA for respective ancestral population in gnomAD. Figure S4. Carrier density for the distribution of percentile expression of impacted genes. Figure S5. Lolliplots showing the positions of NC P/LPs in genes suggestively enriched with significant ASE NC P/LPs.

Predicting the effects of genetic variants on splicing is highly relevant for human genetics. We describe the framework MMSplice (modular modeling of splicing) with which we built the winning model of the CAGI5 exon skipping prediction challenge. The MMSplice modules are neural networks scoring exon, intron, and splice sites, trained on distinct large-scale genomics datasets. These modules are combined to predict effects of variants on exon skipping, splice site choice, splicing efficiency, and pathogenicity, with matched or higher performance than state-of-the-art. Our models, available in the repository Kipoi, apply to variants including indels directly from VCF files. Electronic supplementary material The online version of this article (10.1186/s13059-019-1653-z) contains supplementary material, which is available to authorized users.

Supplementary methods and figures. (PDF 674 kb)