Graphical AbstractHighlights d SpliceAI, a 32-layer deep neural network, predicts splicing from a pre-mRNA sequence d 75% of predicted cryptic splice variants validate on RNA-seq d Cryptic splicing may yield 10% of pathogenic variants in neurodevelopmental disorders d Cryptic splice variants frequently give rise to alternative splicing A deep neural network precisely models mRNA splicing from a genomic sequence and accurately predicts noncoding cryptic splice mutations in patients with rare genetic diseases. SUMMARYThe splicing of pre-mRNAs into mature transcripts is remarkable for its precision, but the mechanisms by which the cellular machinery achieves such specificity are incompletely understood. Here, we describe a deep neural network that accurately predicts splice junctions from an arbitrary pre-mRNA transcript sequence, enabling precise prediction of noncoding genetic variants that cause cryptic splicing. Synonymous and intronic mutations with predicted splice-altering consequence validate at a high rate on RNA-seq and are strongly deleterious in the human population. De novo mutations with predicted splice-altering consequence are significantly enriched in patients with autism and intellectual disability compared to healthy controls and validate against RNA-seq in 21 out of 28 of these patients. We estimate that 9%-11% of pathogenic mutations in patients with rare genetic disorders are caused by this previously underappreciated class of disease variation.(legend continued on next page) (F) Relationship between exon-intron length and the strength of the adjoining splice sites, as predicted by SpliceAI-80 nt (local motif score) and SpliceAI-10k. The genome-wide distributions of exon length (yellow) and intron length (pink) are shown in the background. The x axis is in log-scale. (G) A pair of splice acceptor and donor motifs, placed 150 nt apart, are walked along the HMGCR gene. Shown are, at each position, K562 nucleosome signal and the likelihood of the pair forming an exon at that position, as predicted by SpliceAI-10k. The genome-wide Spearman correlation between the two tracks is shown. (H) Average K562 and GM12878 nucleosome signal near private mutations that are predicted by the SpliceAI-10k model to create novel exons in the GTEx cohort.

SummaryBackgroundHuman genome sequencing has transformed our understanding of genomic variation and its relevance to health and disease, and is now starting to enter clinical practice for the diagnosis of rare diseases. The question of whether and how some categories of genomic findings should be shared with individual research participants is currently a topic of international debate, and development of robust analytical workflows to identify and communicate clinically relevant variants is paramount.MethodsThe Deciphering Developmental Disorders (DDD) study has developed a UK-wide patient recruitment network involving over 180 clinicians across all 24 regional genetics services, and has performed genome-wide microarray and whole exome sequencing on children with undiagnosed developmental disorders and their parents. After data analysis, pertinent genomic variants were returned to individual research participants via their local clinical genetics team.FindingsAround 80 000 genomic variants were identified from exome sequencing and microarray analysis in each individual, of which on average 400 were rare and predicted to be protein altering. By focusing only on de novo and segregating variants in known developmental disorder genes, we achieved a diagnostic yield of 27% among 1133 previously investigated yet undiagnosed children with developmental disorders, whilst minimising incidental findings. In families with developmentally normal parents, whole exome sequencing of the child and both parents resulted in a 10-fold reduction in the number of potential causal variants that needed clinical evaluation compared to sequencing only the child. Most diagnostic variants identified in known genes were novel and not present in current databases of known disease variation.InterpretationImplementation of a robust translational genomics workflow is achievable within a large-scale rare disease research study to allow feedback of potentially diagnostic findings to clinicians and research participants. Systematic recording of relevant clinical data, curation of a gene–phenotype knowledge base, and development of clinical decision support software are needed in addition to automated exclusion of almost all variants, which is crucial for scalable prioritisation and review of possible diagnostic variants. However, the resource requirements of development and maintenance of a clinical reporting system within a research setting are substantial.FundingHealth Innovation Challenge Fund, a parallel funding partnership between the Wellcome Trust and the UK Department of Health.

De novo mutations (DNMs) in protein-coding genes are a well-established cause of developmental disorders (DD). However, known DD-associated genes only account for a minority of the observed excess of such DNMs. To identify novel DD-associated genes, we integrated healthcare and research exome sequences on 31,058 DD parent-offspring trios, and developed a simulation-based statistical test to identify gene-specific enrichments of DNMs. We identified 285 significantly DD-associated genes, including 28 not previously robustly associated with DDs. Despite detecting more DD-associated genes than in any previous study, much of the excess of DNMs of protein-coding genes remains unaccounted for. Modelling suggests that over 1,000 novel DD-associated genes await discovery, many of which are likely to be less penetrant than the currently known genes. Research access to clinical diagnostic datasets will be critical for completing the map of dominant DDs.

By analyzing the whole-exome sequences of 4,264 schizophrenia cases, 9,343 controls and 1,077 trios, we identified a genome-wide significant association between rare loss-of-function (LoF) variants in SETD1A and risk for schizophrenia (P = 3.3 × 10(-9)). We found only two heterozygous LoF variants in 45,376 exomes from individuals without a neuropsychiatric diagnosis, indicating that SETD1A is substantially depleted of LoF variants in the general population. Seven of the ten individuals with schizophrenia carrying SETD1A LoF variants also had learning difficulties. We further identified four SETD1A LoF carriers among 4,281 children with severe developmental disorders and two more carriers in an independent sample of 5,720 Finnish exomes, both with notable neuropsychiatric phenotypes. Together, our observations indicate that LoF variants in SETD1A cause a range of neurodevelopmental disorders, including schizophrenia. Combining these data with previous common variant evidence, we suggest that epigenetic dysregulation, specifically in the histone H3K4 methylation pathway, is an important mechanism in the pathogenesis of schizophrenia.

Millions of human genomes and exomes have been sequenced, but their clinical applications remain limited due to the difficulty of distinguishing disease-causing mutations from benign genetic variation. Here we demonstrate that common missense variants in other primate species are largely clinically benign in human, enabling pathogenic mutations to be systematically identified by the process of elimination. Using hundreds of thousands of common variants from population sequencing of six non-human primate species, we train a deep neural network that identifies pathogenic mutations in rare disease patients with 88% accuracy and enables the discovery of 14 new candidate genes in intellectual disability at genome-wide significance. Cataloging common variation from additional primate species would improve interpretation for millions of variants of uncertain significance, further advancing the clinical utility of human genome sequencing.