Genome-wide association studies have uncovered thousands of common variants associated with human disease, but the contribution of rare variants to common disease remains relatively unexplored. The UK Biobank contains detailed phenotypic data linked to medical records for approximately 500,000 participants, offering an unprecedented opportunity to evaluate the effect of rare variation on a broad collection of traits^1,2. Here we study the relationships between rare protein-coding variants and 17,361 binary and 1,419 quantitative phenotypes using exome sequencing data from 269,171 UK Biobank participants of European ancestry. Gene-based collapsing analyses revealed 1,703 statistically significant gene–phenotype associations for binary traits, with a median odds ratio of 12.4. Furthermore, 83% of these associations were undetectable via single-variant association tests, emphasizing the power of gene-based collapsing analysis in the setting of high allelic heterogeneity. Gene–phenotype associations were also significantly enriched for loss-of-function-mediated traits and approved drug targets. Finally, we performed ancestry-specific and pan-ancestry collapsing analyses using exome sequencing data from 11,933 UK Biobank participants of African, East Asian or South Asian ancestry. Our results highlight a significant contribution of rare variants to common disease. Summary statistics are publicly available through an interactive portal (http://azphewas.com/).

Despite the recognized clinical value of exome-based diagnostics, methods for comprehensive genomic interpretation remain immature. Diagnoses are based on known or presumed pathogenic variants in genes already associated with a similar phenotype. Here, we extend this paradigm by evaluating novel bioinformatics approaches to aid identification of new gene–disease associations. We analyzed 119 trios to identify both diagnostic genotypes in known genes and candidate genotypes in novel genes. We considered qualifying genotypes based on their population frequency and in silico predicted effects we also characterized the patterns of genotypes enriched among this collection of patients. We obtained a genetic diagnosis for 29 (24%) of our patients. We showed that patients carried an excess of damaging de novo mutations in intolerant genes, particularly those shown to be essential in mice (P = 3.4 × 10⁻⁸). This enrichment is only partially explained by mutations found in known disease-causing genes. This work indicates that the application of appropriate bioinformatics analyses to clinical sequence data can also help implicate novel disease genes and suggest expanded phenotypes for known disease genes. These analyses further suggest that some cases resolved by whole-exome sequencing will have direct therapeutic implications. Genet Med 17 10, 774–781.

Integrating human genomics and proteomics can help elucidate disease mechanisms, identify clinical biomarkers and discover drug targets^1–4. Because previous proteogenomic studies have focused on common variation via genome-wide association studies, the contribution of rare variants to the plasma proteome remains largely unknown. Here we identify associations between rare protein-coding variants and 2,923 plasma protein abundances measured in 49,736 UK Biobank individuals. Our variant-level exome-wide association study identified 5,433 rare genotype–protein associations, of which 81% were undetected in a previous genome-wide association study of the same cohort⁵. We then looked at aggregate signals using gene-level collapsing analysis, which revealed 1,962 gene–protein associations. Of the 691 gene-level signals from protein-truncating variants, 99.4% were associated with decreased protein levels. STAB1 and STAB2, encoding scavenger receptors involved in plasma protein clearance, emerged as pleiotropic loci, with 77 and 41 protein associations, respectively. We demonstrate the utility of our publicly accessible resource through several applications. These include detailing an allelic series in NLRC4, identifying potential biomarkers for a fatty liver disease-associated variant in HSD17B13 and bolstering phenome-wide association studies by integrating protein quantitative trait loci with protein-truncating variants in collapsing analyses. Finally, we uncover distinct proteomic consequences of clonal haematopoiesis (CH), including an association between TET2-CH and increased FLT3 levels. Our results highlight a considerable role for rare variation in plasma protein abundance and the value of proteogenomics in therapeutic discovery.

The emergence of biobank-level datasets offers new opportunities to discover novel biomarkers and develop predictive algorithms for human disease. Here, we present an ensemble machine-learning framework (machine learning with phenotype associations, MILTON) utilizing a range of biomarkers to predict 3,213 diseases in the UK Biobank. Leveraging the UK Biobank’s longitudinal health record data, MILTON predicts incident disease cases undiagnosed at time of recruitment, largely outperforming available polygenic risk scores. We further demonstrate the utility of MILTON in augmenting genetic association analyses in a phenome-wide association study of 484,230 genome-sequenced samples, along with 46,327 samples with matched plasma proteomics data. This resulted in improved signals for 88 known (P < 1 × 10⁻⁸) gene–disease relationships alongside 182 gene–disease relationships that did not achieve genome-wide significance in the nonaugmented baseline cohorts. We validated these discoveries in the FinnGen biobank alongside two orthogonal machine-learning methods built for gene–disease prioritization. All extracted gene–disease associations and incident disease predictive biomarkers are publicly available (http://milton.public.cgr.astrazeneca.com).

Elucidating functionality in non-coding regions is a key challenge in human genomics. It has been shown that intolerance to variation of coding and proximal non-coding sequence is a strong predictor of human disease relevance. Here, we integrate intolerance to variation, functional genomic annotations and primary genomic sequence to build JARVIS: a comprehensive deep learning model to prioritize non-coding regions, outperforming other human lineage-specific scores. Despite being agnostic to evolutionary conservation, JARVIS performs comparably or outperforms conservation-based scores in classifying pathogenic single-nucleotide and structural variants. In constructing JARVIS, we introduce the genome-wide residual variation intolerance score (gwRVIS), applying a sliding-window approach to whole genome sequencing data from 62,784 individuals. gwRVIS distinguishes Mendelian disease genes from more tolerant CCDS regions and highlights ultra-conserved non-coding elements as the most intolerant regions in the human genome. Both JARVIS and gwRVIS capture previously inaccessible human-lineage constraint information and will enhance our understanding of the non-coding genome.

The druggability of targets is a crucial consideration in drug target selection. Here, we adopt a stochastic semi-supervised ML framework to develop DrugnomeAI, which estimates the druggability likelihood for every protein-coding gene in the human exome. DrugnomeAI integrates gene-level properties from 15 sources resulting in 324 features. The tool generates exome-wide predictions based on labelled sets of known drug targets (median AUC: 0.97), highlighting features from protein-protein interaction networks as top predictors. DrugnomeAI provides generic as well as specialised models stratified by disease type or drug therapeutic modality. The top-ranking DrugnomeAI genes were significantly enriched for genes previously selected for clinical development programs (p value < 1 × 10⁻³⁰⁸) and for genes achieving genome-wide significance in phenome-wide association studies of 450 K UK Biobank exomes for binary (p value = 1.7 × 10⁻⁵) and quantitative traits (p value = 1.6 × 10⁻⁷). We accompany our method with a web application (http://drugnomeai.public.cgr.astrazeneca.com) to visualise the druggability predictions and the key features that define gene druggability, per disease type and modality.