Improved pathogenicity prediction for rare human missense variants.

The success of personalized genomic medicine depends on our ability to assess the pathogenicity of rare human variants, including the important class of missense variation. There are many challenges in training accurate computational systems, e.g., in finding the balance between quantity, quality, and bias in the variant sets used as training examples and avoiding predictive features that can accentuate the effects of bias. Here, we describe VARITY, which judiciously exploits a larger reservoir of training examples with uncertain accuracy and representativity. To limit circularity and bias, VARITY excludes features informed by variant annotation and protein identity. To provide a rationale for each prediction, we quantified the contribution of features and feature combinations to the pathogenicity inference of each variant. VARITY outperformed all previous computational methods evaluated, identifying at least 10% more pathogenic variants at thresholds achieving high (90% precision) stringency.

Shifting landscapes of human MTHFR missense-variant effects.

Most rare clinical missense variants cannot currently be classified as pathogenic or benign. Deficiency in human 5,10-methylenetetrahydrofolate reductase (MTHFR), the most common inherited disorder of folate metabolism, is caused primarily by rare missense variants. Further complicating variant interpretation, variant impacts often depend on environment. An important example of this phenomenon is the MTHFR variant p.Ala222Val (c.665C>T), which is carried by half of all humans and has a phenotypic impact that depends on dietary folate. Here we describe the results of 98,336 variant functional-impact assays, covering nearly all possible MTHFR amino acid substitutions in four folinate environments, each in the presence and absence of p.Ala222Val. The resulting atlas of MTHFR variant effects reveals many complex dependencies on both folinate and p.Ala222Val. MTHFR atlas scores can distinguish pathogenic from benign variants and, among individuals with severe MTHFR deficiency, correlate with age of disease onset. Providing a powerful tool for understanding structure-function relationships, the atlas suggests a role for a disordered loop in retaining cofactor at the active site and identifies variants that enable escape of inhibition by S-adenosylmethionine. Thus, a model based on eight MTHFR variant effect maps illustrates how shifting landscapes of environment- and genetic-background-dependent missense variation can inform our clinical, structural, and functional understanding of MTHFR deficiency.

A web application and service for imputing and visualizing missense variant effect maps.

SUMMARY: The promise of personalized genomic medicine depends on our ability to assess the functional impact of rare sequence variation. Multiplexed assays can experimentally measure the functional impact of missense variants on a massive scale. However, even after such assays, many missense variants remain poorly measured. Here we describe a software pipeline and application to impute missing information in experimentally determined variant effect maps. AVAILABILITY AND IMPLEMENTATION: http://impute.varianteffect.org source code: https://github.com/joewuca/imputation. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Proteome-wide, Structure-Based Prediction of Protein-Protein Interactions/New Molecular Interactions Viewer.

Determining the complete Arabidopsis (Arabidopsis thaliana) protein-protein interaction network is essential for understanding the functional organization of the proteome. Numerous small-scale studies and a couple of large-scale ones have elucidated a fraction of the estimated 300,000 binary protein-protein interactions in Arabidopsis. In this study, we provide evidence that a docking algorithm has the ability to identify real interactions using both experimentally determined and predicted protein structures. We ranked 0.91 million interactions generated by all possible pairwise combinations of 1,346 predicted structure models from an Arabidopsis predicted "structure-ome" and found a significant enrichment of real interactions for the top-ranking predicted interactions, as shown by cosubcellular enrichment analysis and yeast two-hybrid validation. Our success rate for computationally predicted, structure-based interactions was 63% of the success rate for published interactions naively tested using the yeast two-hybrid system and 2.7 times better than for randomly picked pairs of proteins. This study provides another perspective in interactome exploration and biological network reconstruction using protein structural information. We have made these interactions freely accessible through an improved Arabidopsis Interactions Viewer and have created community tools for accessing these and ∼2.8 million other protein-protein and protein-DNA interactions for hypothesis generation by researchers worldwide. The Arabidopsis Interactions Viewer is freely available at http://bar.utoronto.ca/interactions2/.

Baseplate assembly of phage Mu: Defining the conserved core components of contractile-tailed phages and related bacterial systems.

Contractile phage tails are powerful cell puncturing nanomachines that have been co-opted by bacteria for self-defense against both bacteria and eukaryotic cells. The tail of phage T4 has long served as the paradigm for understanding contractile tail-like systems despite its greater complexity compared with other contractile-tailed phages. Here, we present a detailed investigation of the assembly of a "simple" contractile-tailed phage baseplate, that of Escherichia coli phage Mu. By coexpressing various combinations of putative Mu baseplate proteins, we defined the required components of this baseplate and delineated its assembly pathway. We show that the Mu baseplate is constructed through the independent assembly of wedges that are organized around a central hub complex. The Mu wedges are comprised of only three protein subunits rather than the seven found in the equivalent structure in T4. Through extensive bioinformatic analyses, we found that homologs of the essential components of the Mu baseplate can be identified in the majority of contractile-tailed phages and prophages. No T4-like prophages were identified. The conserved simple baseplate components were also found in contractile tail-derived bacterial apparatuses, such as type VI secretion systems, Photorhabdus virulence cassettes, and R-type tailocins. Our work highlights the evolutionary connections and similarities in the biochemical behavior of phage Mu wedge components and the TssF and TssG proteins of the type VI secretion system. In addition, we demonstrate the importance of the Mu baseplate as a model system for understanding bacterial phage tail-derived systems.

A framework for exhaustively mapping functional missense variants.

Although we now routinely sequence human genomes, we can confidently identify only a fraction of the sequence variants that have a functional impact. Here, we developed a deep mutational scanning framework that produces exhaustive maps for human missense variants by combining random codon mutagenesis and multiplexed functional variation assays with computational imputation and refinement. We applied this framework to four proteins corresponding to six human genes: UBE2I (encoding SUMO E2 conjugase), SUMO1 (small ubiquitin-like modifier), TPK1 (thiamin pyrophosphokinase), and CALM1/2/3 (three genes encoding the protein calmodulin). The resulting maps recapitulate known protein features and confidently identify pathogenic variation. Assays potentially amenable to deep mutational scanning are already available for 57% of human disease genes, suggesting that DMS could ultimately map functional variation for all human disease genes.

Benchmarking computational variant effect predictors by their ability to infer human traits.

BACKGROUND: Computational variant effect predictors offer a scalable and increasingly reliable means of interpreting human genetic variation, but concerns of circularity and bias have limited previous methods for evaluating and comparing predictors. Population-level cohorts of genotyped and phenotyped participants that have not been used in predictor training can facilitate an unbiased benchmarking of available methods. Using a curated set of human gene-trait associations with a reported rare-variant burden association, we evaluate the correlations of 24 computational variant effect predictors with associated human traits in the UK Biobank and All of Us cohorts. RESULTS: AlphaMissense outperformed all other predictors in inferring human traits based on rare missense variants in UK Biobank and All of Us participants. The overall rankings of computational variant effect predictors in these two cohorts showed a significant positive correlation. CONCLUSION: We describe a method to assess computational variant effect predictors that sidesteps the limitations of previous evaluations. This approach is generalizable to future predictors and could continue to inform predictor choice for personal and clinical genetics.

A proactive genotype-to-patient-phenotype map for cystathionine beta-synthase.

BACKGROUND: For the majority of rare clinical missense variants, pathogenicity status cannot currently be classified. Classical homocystinuria, characterized by elevated homocysteine in plasma and urine, is caused by variants in the cystathionine beta-synthase (CBS) gene, most of which are rare. With early detection, existing therapies are highly effective. METHODS: Damaging CBS variants can be detected based on their failure to restore growth in yeast cells lacking the yeast ortholog CYS4. This assay has only been applied reactively, after first observing a variant in patients. Using saturation codon-mutagenesis, en masse growth selection, and sequencing, we generated a comprehensive, proactive map of CBS missense variant function. RESULTS: Our CBS variant effect map far exceeds the performance of computational predictors of disease variants. Map scores correlated strongly with both disease severity (Spearman's ϱ = 0.9) and human clinical response to vitamin B6 (ϱ = 0.93). CONCLUSIONS: We demonstrate that highly multiplexed cell-based assays can yield proactive maps of variant function and patient response to therapy, even for rare variants not previously seen in the clinic.

A predicted protein interactome for rice.

BACKGROUND: Protein-protein interactions (PPIs) create the steps in signaling and regulatory networks central to most fundamental biological processes. It is possible to predict these interactions by making use of experimentally determined orthologous interactions in other species. RESULTS: In this study, prediction of PPIs in rice was carried out by the interolog method of mapping deduced orthologous genes to protein interactions supported by experimental evidence from reference organisms. We predicted 37112 interactions for 4567 rice proteins, including 1671 predicted self interactions (homo-interactions) and 35441 predicted interactions between different proteins (hetero-interactions). These matched 168 of 675 experimentally-determined interactions in rice. Interacting proteins were significantly more co-expressed than expected by chance, which is typical of experimentally-determined interactomes. The rice interacting proteins were divided topologically into 981 free ends (proteins with single interactions), 499 pipes (proteins with two interactions) and 3087 hubs of different sizes ranging from three to more than 100 interactions. CONCLUSIONS: This predicted rice interactome extends known pathways and improves functional annotation of unknown rice proteins and networks in rice, and is easily explored with software tools presented here.