Highly parameterized polygenic scores tend to overfit to population stratification via random effects.

Polygenic scores (PGSs), increasingly used in clinical settings, frequently include many genetic variants, with performance typically peaking at thousands of variants. Such highly parameterized PGSs often include variants that do not pass a genome-wide significance threshold. We propose a mathematical perspective that renders the effects of many of these non-significant variants random rather than causal, with the randomness capturing population structure. We devise methods to assess variant effect randomness and population stratification bias. Applying these methods to 141 traits from the UK Biobank, we find that, for many PGSs, the effects of non-significant variants are considerably random, with the extent of randomness associated with the degree of overfitting to population structure of the discovery cohort. Our findings explain why highly parameterized PGSs simultaneously have superior cohort-specific performance and limited generalizability, suggesting the critical need for variant randomness tests in PGS evaluation. Supporting code and a dashboard are available at https://github.com/songlab-cal/StratPGS.

Rare penetrant mutations confer severe risk of common diseases.

We examined 454,712 exomes for genes associated with a wide spectrum of complex traits and common diseases and observed that rare, penetrant mutations in genes implicated by genome-wide association studies confer ~10-fold larger effects than common variants in the same genes. Consequently, an individual at the phenotypic extreme and at the greatest risk for severe, early-onset disease is better identified by a few rare penetrant variants than by the collective action of many common variants with weak effects. By combining rare variants across phenotype-associated genes into a unified genetic risk model, we demonstrate superior portability across diverse global populations compared with common-variant polygenic risk scores, greatly improving the clinical utility of genetic-based risk prediction.

The landscape of tolerated genetic variation in humans and primates.

Personalized genome sequencing has revealed millions of genetic differences between individuals, but our understanding of their clinical relevance remains largely incomplete. To systematically decipher the effects of human genetic variants, we obtained whole-genome sequencing data for 809 individuals from 233 primate species and identified 4.3 million common protein-altering variants with orthologs in humans. We show that these variants can be inferred to have nondeleterious effects in humans based on their presence at high allele frequencies in other primate populations. We use this resource to classify 6% of all possible human protein-altering variants as likely benign and impute the pathogenicity of the remaining 94% of variants with deep learning, achieving state-of-the-art accuracy for diagnosing pathogenic variants in patients with genetic diseases.

Rare penetrant mutations confer severe risk of common diseases.

We examined 454,712 exomes for genes associated with a wide spectrum of complex traits and common diseases and observed that rare, penetrant mutations in genes implicated by genome-wide association studies confer ∼10-fold larger effects than common variants in the same genes. Consequently, an individual at the phenotypic extreme and at the greatest risk for severe, early-onset disease is better identified by a few rare penetrant variants than by the collective action of many common variants with weak effects. By combining rare variants across phenotype-associated genes into a unified genetic risk model, we demonstrate superior portability across diverse global populations compared to common variant polygenic risk scores, greatly improving the clinical utility of genetic-based risk prediction. One sentence summary: Rare variant polygenic risk scores identify individuals with outlier phenotypes in common human diseases and complex traits.

The landscape of tolerated genetic variation in humans and primates.

Personalized genome sequencing has revealed millions of genetic differences between individuals, but our understanding of their clinical relevance remains largely incomplete. To systematically decipher the effects of human genetic variants, we obtained whole genome sequencing data for 809 individuals from 233 primate species, and identified 4.3 million common protein-altering variants with orthologs in human. We show that these variants can be inferred to have non-deleterious effects in human based on their presence at high allele frequencies in other primate populations. We use this resource to classify 6% of all possible human protein-altering variants as likely benign and impute the pathogenicity of the remaining 94% of variants with deep learning, achieving state-of-the-art accuracy for diagnosing pathogenic variants in patients with genetic diseases. One Sentence Summary: Deep learning classifier trained on 4.3 million common primate missense variants predicts variant pathogenicity in humans.

Genetic and chemotherapeutic influences on germline hypermutation.

Mutations in the germline generates all evolutionary genetic variation and is a cause of genetic disease. Parental age is the primary determinant of the number of new germline mutations in an individual's genome1,2. Here we analysed the genome-wide sequences of 21,879 families with rare genetic diseases and identified 12 individuals with a hypermutated genome with between two and seven times more de novo single-nucleotide variants than expected. In most families (9 out of 12), the excess mutations came from the father. Two families had genetic drivers of germline hypermutation, with fathers carrying damaging genetic variation in DNA-repair genes. For five of the families, paternal exposure to chemotherapeutic agents before conception was probably a key driver of hypermutation. Our results suggest that the germline is well protected from mutagenic effects, hypermutation is rare, the number of excess mutations is relatively modest and most individuals with a hypermutated genome will not have a genetic disease.

The contribution of X-linked coding variation to severe developmental disorders.

Over 130 X-linked genes have been robustly associated with developmental disorders, and X-linked causes have been hypothesised to underlie the higher developmental disorder rates in males. Here, we evaluate the burden of X-linked coding variation in 11,044 developmental disorder patients, and find a similar rate of X-linked causes in males and females (6.0% and 6.9%, respectively), indicating that such variants do not account for the 1.4-fold male bias. We develop an improved strategy to detect X-linked developmental disorders and identify 23 significant genes, all of which were previously known, consistent with our inference that the vast majority of the X-linked burden is in known developmental disorder-associated genes. Importantly, we estimate that, in male probands, only 13% of inherited rare missense variants in known developmental disorder-associated genes are likely to be pathogenic. Our results demonstrate that statistical analysis of large datasets can refine our understanding of modes of inheritance for individual X-linked disorders.

Evidence for 28 genetic disorders discovered by combining healthcare and research data.

De novo mutations in protein-coding genes are a well-established cause of developmental disorders1. However, genes known to be associated with developmental disorders account for only a minority of the observed excess of such de novo mutations1,2. Here, to identify previously undescribed genes associated with developmental disorders, we integrate healthcare and research exome-sequence data from 31,058 parent-offspring trios of individuals with developmental disorders, and develop a simulation-based statistical test to identify gene-specific enrichment of de novo mutations. We identified 285 genes that were significantly associated with developmental disorders, including 28 that had not previously been robustly associated with developmental disorders. Although we detected more genes associated with developmental disorders, much of the excess of de novo mutations in protein-coding genes remains unaccounted for. Modelling suggests that more than 1,000 genes associated with developmental disorders have not yet been described, 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 genes associated with developmental disorders.

Finding Diagnostically Useful Patterns in Quantitative Phenotypic Data.

Trio-based whole-exome sequence (WES) data have established confident genetic diagnoses in ∼40% of previously undiagnosed individuals recruited to the Deciphering Developmental Disorders (DDD) study. Here we aim to use the breadth of phenotypic information recorded in DDD to augment diagnosis and disease variant discovery in probands. Median Euclidean distances (mEuD) were employed as a simple measure of similarity of quantitative phenotypic data within sets of ≥10 individuals with plausibly causative de novo mutations (DNM) in 28 different developmental disorder genes. 13/28 (46.4%) showed significant similarity for growth or developmental milestone metrics, 10/28 (35.7%) showed similarity in HPO term usage, and 12/28 (43%) showed no phenotypic similarity. Pairwise comparisons of individuals with high-impact inherited variants to the 32 individuals with causative DNM in ANKRD11 using only growth z-scores highlighted 5 likely causative inherited variants and two unrecognized DNM resulting in an 18% diagnostic uplift for this gene. Using an independent approach, naive Bayes classification of growth and developmental data produced reasonably discriminative models for the 24 DNM genes with sufficiently complete data. An unsupervised naive Bayes classification of 6,993 probands with WES data and sufficient phenotypic information defined 23 in silico syndromes (ISSs) and was used to test a "phenotype first" approach to the discovery of causative genotypes using WES variants strictly filtered on allele frequency, mutation consequence, and evidence of constraint in humans. This highlighted heterozygous de novo nonsynonymous variants in SPTBN2 as causative in three DDD probands.

Exome-wide assessment of the functional impact and pathogenicity of multinucleotide mutations.

Approximately 2% of de novo single-nucleotide variants (SNVs) appear as part of clustered mutations that create multinucleotide variants (MNVs). MNVs are an important source of genomic variability as they are more likely to alter an encoded protein than a SNV, which has important implications in disease as well as evolution. Previous studies of MNVs have focused on their mutational origins and have not systematically evaluated their functional impact and contribution to disease. We identified 69,940 MNVs and 91 de novo MNVs in 6688 exome-sequenced parent-offspring trios from the Deciphering Developmental Disorders Study comprising families with severe developmental disorders. We replicated the previously described MNV mutational signatures associated with DNA polymerase zeta, an error-prone translesion polymerase, and the APOBEC family of DNA deaminases. We estimate the simultaneous MNV germline mutation rate to be 1.78 × 10-10 mutations per base pair per generation. We found that most MNVs within a single codon create a missense change that could not have been created by a SNV. MNV-induced missense changes were, on average, more physicochemically divergent, were more depleted in highly constrained genes (pLI ≥ 0.9), and were under stronger purifying selection compared with SNV-induced missense changes. We found that de novo MNVs were significantly enriched in genes previously associated with developmental disorders in affected children. This shows that MNVs can be more damaging than SNVs even when both induce missense changes, and are an important variant type to consider in relation to human disease.

Predicting Splicing from Primary Sequence with Deep Learning.

The 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.

Author Correction: Predicting the clinical impact of human mutation with deep neural networks.

In the version of this article originally published, the name of author Serafim Batzoglou was misspelled. The error has been corrected in the HTML and PDF versions of the article.

Pathogenicity and selective constraint on variation near splice sites.

Mutations that perturb normal pre-mRNA splicing are significant contributors to human disease. We used exome sequencing data from 7833 probands with developmental disorders (DDs) and their unaffected parents, as well as more than 60,000 aggregated exomes from the Exome Aggregation Consortium, to investigate selection around the splice sites and quantify the contribution of splicing mutations to DDs. Patterns of purifying selection, a deficit of variants in highly constrained genes in healthy subjects, and excess de novo mutations in patients highlighted particular positions within and around the consensus splice site of greater functional relevance. By using mutational burden analyses in this large cohort of proband-parent trios, we could estimate in an unbiased manner the relative contributions of mutations at canonical dinucleotides (73%) and flanking noncanonical positions (27%), and calculate the positive predictive value of pathogenicity for different classes of mutations. We identified 18 patients with likely diagnostic de novo mutations in dominant DD-associated genes at noncanonical positions in splice sites. We estimate 35%-40% of pathogenic variants in noncanonical splice site positions are missing from public databases.

Quantifying the contribution of recessive coding variation to developmental disorders.

We estimated the genome-wide contribution of recessive coding variation in 6040 families from the Deciphering Developmental Disorders study. The proportion of cases attributable to recessive coding variants was 3.6% in patients of European ancestry, compared with 50% explained by de novo coding mutations. It was higher (31%) in patients with Pakistani ancestry, owing to elevated autozygosity. Half of this recessive burden is attributable to known genes. We identified two genes not previously associated with recessive developmental disorders, KDM5B and EIF3F, and functionally validated them with mouse and cellular models. Our results suggest that recessive coding variants account for a small fraction of currently undiagnosed nonconsanguineous individuals, and that the role of noncoding variants, incomplete penetrance, and polygenic mechanisms need further exploration.

Common genetic variants contribute to risk of rare severe neurodevelopmental disorders.

There are thousands of rare human disorders that are caused by single deleterious, protein-coding genetic variants1. However, patients with the same genetic defect can have different clinical presentations2-4, and some individuals who carry known disease-causing variants can appear unaffected5. Here, to understand what explains these differences, we study a cohort of 6,987 children assessed by clinical geneticists to have severe neurodevelopmental disorders such as global developmental delay and autism, often in combination with abnormalities of other organ systems. Although the genetic causes of these neurodevelopmental disorders are expected to be almost entirely monogenic, we show that 7.7% of variance in risk is attributable to inherited common genetic variation. We replicated this genome-wide common variant burden by showing, in an independent sample of 728 trios (comprising a child plus both parents) from the same cohort, that this burden is over-transmitted from parents to children with neurodevelopmental disorders. Our common-variant signal is significantly positively correlated with genetic predisposition to lower educational attainment, decreased intelligence and risk of schizophrenia. We found that common-variant risk was not significantly different between individuals with and without a known protein-coding diagnostic variant, which suggests that common-variant risk affects patients both with and without a monogenic diagnosis. In addition, previously published common-variant scores for autism, height, birth weight and intracranial volume were all correlated with these traits within our cohort, which suggests that phenotypic expression in individuals with monogenic disorders is affected by the same variants as in the general population. Our results demonstrate that common genetic variation affects both overall risk and clinical presentation in neurodevelopmental disorders that are typically considered to be monogenic.

Predicting the clinical impact of human mutation with deep neural networks.

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.

De novo mutations in regulatory elements in neurodevelopmental disorders.

We previously estimated that 42% of patients with severe developmental disorders carry pathogenic de novo mutations in coding sequences. The role of de novo mutations in regulatory elements affecting genes associated with developmental disorders, or other genes, has been essentially unexplored. We identified de novo mutations in three classes of putative regulatory elements in almost 8,000 patients with developmental disorders. Here we show that de novo mutations in highly evolutionarily conserved fetal brain-active elements are significantly and specifically enriched in neurodevelopmental disorders. We identified a significant twofold enrichment of recurrently mutated elements. We estimate that, genome-wide, 1-3% of patients without a diagnostic coding variant carry pathogenic de novo mutations in fetal brain-active regulatory elements and that only 0.15% of all possible mutations within highly conserved fetal brain-active elements cause neurodevelopmental disorders with a dominant mechanism. Our findings represent a robust estimate of the contribution of de novo mutations in regulatory elements to this genetically heterogeneous set of disorders, and emphasize the importance of combining functional and evolutionary evidence to identify regulatory causes of genetic disorders.

Making new genetic diagnoses with old data: iterative reanalysis and reporting from genome-wide data in 1,133 families with developmental disorders.

PURPOSE: Given the rapid pace of discovery in rare disease genomics, it is likely that improvements in diagnostic yield can be made by systematically reanalyzing previously generated genomic sequence data in light of new knowledge. METHODS: We tested this hypothesis in the United Kingdom-wide Deciphering Developmental Disorders study, where in 2014 we reported a diagnostic yield of 27% through whole-exome sequencing of 1,133 children with severe developmental disorders and their parents. We reanalyzed existing data using improved variant calling methodologies, novel variant detection algorithms, updated variant annotation, evidence-based filtering strategies, and newly discovered disease-associated genes. RESULTS: We are now able to diagnose an additional 182 individuals, taking our overall diagnostic yield to 454/1,133 (40%), and another 43 (4%) have a finding of uncertain clinical significance. The majority of these new diagnoses are due to novel developmental disorder-associated genes discovered since our original publication. CONCLUSION: This study highlights the importance of coupling large-scale research with clinical practice, and of discussing the possibility of iterative reanalysis and recontact with patients and health professionals at an early stage. We estimate that implementing parent-offspring whole-exome sequencing as a first-line diagnostic test for developmental disorders would diagnose >50% of patients.

Clinical and molecular consequences of disease-associated de novo mutations in SATB2.

PURPOSE: To characterize features associated with de novo mutations affecting SATB2 function in individuals ascertained on the basis of intellectual disability. METHODS: Twenty previously unreported individuals with 19 different SATB2 mutations (11 loss-of-function and 8 missense variants) were studied. Fibroblasts were used to measure mutant protein production. Subcellular localization and mobility of wild-type and mutant SATB2 were assessed using fluorescently tagged protein. RESULTS: Recurrent clinical features included neurodevelopmental impairment (19/19), absent/near absent speech (16/19), normal somatic growth (17/19), cleft palate (9/19), drooling (12/19), and dental anomalies (8/19). Six of eight missense variants clustered in the first CUT domain. Sibling recurrence due to gonadal mosaicism was seen in one family. A nonsense mutation in the last exon resulted in production of a truncated protein retaining all three DNA-binding domains. SATB2 nuclear mobility was mutation-dependent; p.Arg389Cys in CUT1 increased mobility and both p.Gly515Ser in CUT2 and p.Gln566Lys between CUT2 and HOX reduced mobility. The clinical features in individuals with missense variants were indistinguishable from those with loss of function. CONCLUSION: SATB2 haploinsufficiency is a common cause of syndromic intellectual disability. When mutant SATB2 protein is produced, the protein appears functionally inactive with a disrupted pattern of chromatin or matrix association.Genet Med advance online publication 02 February 2017.

Distinct genetic architectures for syndromic and nonsyndromic congenital heart defects identified by exome sequencing.

Congenital heart defects (CHDs) have a neonatal incidence of 0.8-1% (refs. 1,2). Despite abundant examples of monogenic CHD in humans and mice, CHD has a low absolute sibling recurrence risk (∼2.7%), suggesting a considerable role for de novo mutations (DNMs) and/or incomplete penetrance. De novo protein-truncating variants (PTVs) have been shown to be enriched among the 10% of 'syndromic' patients with extra-cardiac manifestations. We exome sequenced 1,891 probands, including both syndromic CHD (S-CHD, n = 610) and nonsyndromic CHD (NS-CHD, n = 1,281). In S-CHD, we confirmed a significant enrichment of de novo PTVs but not inherited PTVs in known CHD-associated genes, consistent with recent findings. Conversely, in NS-CHD we observed significant enrichment of PTVs inherited from unaffected parents in CHD-associated genes. We identified three genome-wide significant S-CHD disorders caused by DNMs in CHD4, CDK13 and PRKD1. Our study finds evidence for distinct genetic architectures underlying the low sibling recurrence risk in S-CHD and NS-CHD.