A curated census of pathogenic and likely pathogenic UTR variants and evaluation of deep learning models for variant effect prediction.

Introduction: Variants in 5' and 3' untranslated regions (UTR) contribute to rare disease. While predictive algorithms to assist in classifying pathogenicity can potentially be highly valuable, the utility of these tools is often unclear, as it depends on carefully selected training and validation conditions. To address this, we developed a high confidence set of pathogenic (P) and likely pathogenic (LP) variants and assessed deep learning (DL) models for predicting their molecular effects. Methods: 3' and 5' UTR variants documented as P or LP (P/LP) were obtained from ClinVar and refined by reviewing the annotated variant effect and reassessing evidence of pathogenicity following published guidelines. Prediction scores from sequence-based DL models were compared between three groups: P/LP variants acting though the mechanism for which the model was designed (model-matched), those operating through other mechanisms (model-mismatched), and putative benign variants. PhyloP was used to compare conservation scores between P/LP and putative benign variants. Results: 295 3' and 188 5' UTR variants were obtained from ClinVar, of which 26 3' and 68 5' UTR variants were classified as P/LP. Predictions by DL models achieved statistically significant differences when comparing modelmatched P/LP variants to both putative benign variants and modelmismatched P/LP variants, as well as when comparing all P/LP variants to putative benign variants. PhyloP conservation scores were significantly higher among P/LP compared to putative benign variants for both the 3' and 5' UTR. Discussion: In conclusion, we present a high-confidence set of P/LP 3' and 5' UTR variants spanning a range of mechanisms and supported by detailed pathogenicity and molecular mechanism evidence curation. Predictions from DL models further substantiate these classifications. These datasets will support further development and validation of DL algorithms designed to predict the functional impact of variants that may be implicated in rare disease.

A resource of variant effect predictions of single nucleotide variants in model organisms.

The effect of single nucleotide variants (SNVs) in coding and noncoding regions is of great interest in genetics. Although many computational methods aim to elucidate the effects of SNVs on cellular mechanisms, it is not straightforward to comprehensively cover different molecular effects. To address this, we compiled and benchmarked sequence and structure-based variant effect predictors and we computed the impact of nearly all possible amino acid and nucleotide variants in the reference genomes of Homo sapiens, Saccharomyces cerevisiae and Escherichia coli Studied mechanisms include protein stability, interaction interfaces, post-translational modifications and transcription factor binding sites. We apply this resource to the study of natural and disease coding variants. We also show how variant effects can be aggregated to generate protein complex burden scores that uncover protein complex to phenotype associations based on a set of newly generated growth profiles of 93 sequenced S. cerevisiae strains in 43 conditions. This resource is available through mutfunc (www.mutfunc.com), a tool by which users can query precomputed predictions by providing amino acid or nucleotide-level variants.

Phenotype inference in an Escherichia coli strain panel.

Understanding how genetic variation contributes to phenotypic differences is a fundamental question in biology. Combining high-throughput gene function assays with mechanistic models of the impact of genetic variants is a promising alternative to genome-wide association studies. Here we have assembled a large panel of 696 Escherichia coli strains, which we have genotyped and measured their phenotypic profile across 214 growth conditions. We integrated variant effect predictors to derive gene-level probabilities of loss of function for every gene across all strains. Finally, we combined these probabilities with information on conditional gene essentiality in the reference K-12 strain to compute the growth defects of each strain. Not only could we reliably predict these defects in up to 38% of tested conditions, but we could also directly identify the causal variants that were validated through complementation assays. Our work demonstrates the power of forward predictive models and the possibility of precision genetic interventions.

ggseqlogo: a versatile R package for drawing sequence logos.

SUMMARY: Sequence logos have become a crucial visualization method for studying underlying sequence patterns in the genome. Despite this, there remains a scarcity of software packages that provide the versatility often required for such visualizations. ggseqlogo is an R package built on the ggplot2 package that aims to address this issue. ggseqlogo offers native illustration of publication-ready DNA, RNA and protein sequence logos in a highly customizable fashion with features including multi-logo plots, qualitative and quantitative colour schemes, annotation of logos and integration with other plots. The package is intuitive to use and seamlessly integrates into R analysis pipelines. AVAILABILITY AND IMPLEMENTATION: ggseqlogo is released under the GNU licence and is freely available via CRAN-The Comprehensive R Archive Network https://cran.r-project.org/web/packages/ggseqlogo. A detailed tutorial can be found at https://omarwagih.github.io/ggseqlogo. CONTACT: wagih@ebi.ac.uk.

Identifications of Putative PKA Substrates with Quantitative Phosphoproteomics and Primary-Sequence-Based Scoring.

Protein kinase A (PKA or cAMP-dependent protein kinase) is a serine/threonine kinase that plays essential roles in the regulation of proliferation, differentiation, and apoptosis. To better understand the functions of PKA, it is necessary to elucidate the direct interplay between PKA and their substrates in living human cells. To identify kinase target substrates in a high-throughput manner, we first quantified the change of phosphoproteome in the cells of which PKA activity was perturbed by drug stimulations. LC-MS/MS analyses identified 2755 and 3191 phosphopeptides from experiments with activator or inhibitor of PKA. To exclude potential indirect targets of PKA, we built a computational model to characterize the kinase sequence specificity toward the substrate target site based on known kinase-substrate relationships. Finally, by combining the sequence recognition model with the quantitative changes in phosphorylation measured in the two drug perturbation experiments, we identified 29 reliable candidates of PKA targeting residues in living cells including 8 previously known substrates. Moreover, 18 of these sites were confirmed to be site-specifically phosphorylated in vitro. Altogether this study proposed a confident list of PKA substrate candidates, expanding our knowledge of PKA signaling network.

Microevolution of Serial Clinical Isolates of Cryptococcus neoformans var. grubii and C. gattii.

The pathogenic species of Cryptococcus are a major cause of mortality owing to severe infections in immunocompromised as well as immunocompetent individuals. Although antifungal treatment is usually effective, many patients relapse after treatment, and in such cases, comparative analyses of the genomes of incident and relapse isolates may reveal evidence of determinative, microevolutionary changes within the host. Here, we analyzed serial isolates cultured from cerebrospinal fluid specimens of 18 South African patients with recurrent cryptococcal meningitis. The time between collection of the incident isolates and collection of the relapse isolates ranged from 124 days to 290 days, and the analyses revealed that, during this period within the patients, the isolates underwent several genetic and phenotypic changes. Considering the vast genetic diversity of cryptococcal isolates in sub-Saharan Africa, it was not surprising to find that the relapse isolates had acquired different genetic and correlative phenotypic changes. They exhibited various mechanisms for enhancing virulence, such as growth at 39°C, adaptation to stress, and capsule production; a remarkable amplification of ERG11 at the native and unlinked locus may provide stable resistance to fluconazole. Our data provide a deeper understanding of the microevolution of Cryptococcus species under pressure from antifungal chemotherapy and host immune responses. This investigation clearly suggests a promising strategy to identify novel targets for improved diagnosis, therapy, and prognosis.IMPORTANCE Opportunistic infections caused by species of the pathogenic yeast Cryptococcus lead to chronic meningoencephalitis and continue to ravage thousands of patients with HIV/AIDS. Despite receiving antifungal treatment, over 10% of patients develop recurrent disease. In this study, we collected isolates of Cryptococcus from cerebrospinal fluid specimens of 18 patients at the time of their diagnosis and when they relapsed several months later. We then sequenced and compared the genomic DNAs of each pair of initial and relapse isolates. We also tested the isolates for several key properties related to cryptococcal virulence as well as for their susceptibility to the antifungal drug fluconazole. These analyses revealed that the relapsing isolates manifested multiple genetic and chromosomal changes that affected a variety of genes implicated in the pathogenicity of Cryptococcus or resistance to fluconazole. This application of comparative genomics to serial clinical isolates provides a blueprint for identifying the mechanisms whereby pathogenic microbes adapt within patients to prolong disease.

Systematic Analysis of Transcriptional and Post-transcriptional Regulation of Metabolism in Yeast.

Cells react to extracellular perturbations with complex and intertwined responses. Systematic identification of the regulatory mechanisms that control these responses is still a challenge and requires tailored analyses integrating different types of molecular data. Here we acquired time-resolved metabolomics measurements in yeast under salt and pheromone stimulation and developed a machine learning approach to explore regulatory associations between metabolism and signal transduction. Existing phosphoproteomics measurements under the same conditions and kinase-substrate regulatory interactions were used to in silico estimate the enzymatic activity of signalling kinases. Our approach identified informative associations between kinases and metabolic enzymes capable of predicting metabolic changes. We extended our analysis to two studies containing transcriptomics, phosphoproteomics and metabolomics measurements across a comprehensive panel of kinases/phosphatases knockouts and time-resolved perturbations to the nitrogen metabolism. Changes in activity of transcription factors, kinases and phosphatases were estimated in silico and these were capable of building predictive models to infer the metabolic adaptations of previously unseen conditions across different dynamic experiments. Time-resolved experiments were significantly more informative than genetic perturbations to infer metabolic adaptation. This difference may be due to the indirect nature of the associations and of general cellular states that can hinder the identification of causal relationships. This work provides a novel genome-scale integrative analysis to propose putative transcriptional and post-translational regulatory mechanisms of metabolic processes.

Quantitative CRISPR interference screens in yeast identify chemical-genetic interactions and new rules for guide RNA design.

BACKGROUND: Genome-scale CRISPR interference (CRISPRi) has been used in human cell lines; however, the features of effective guide RNAs (gRNAs) in different organisms have not been well characterized. Here, we define rules that determine gRNA effectiveness for transcriptional repression in Saccharomyces cerevisiae. RESULTS: We create an inducible single plasmid CRISPRi system for gene repression in yeast, and use it to analyze fitness effects of gRNAs under 18 small molecule treatments. Our approach correctly identifies previously described chemical-genetic interactions, as well as a new mechanism of suppressing fluconazole toxicity by repression of the ERG25 gene. Assessment of multiple target loci across treatments using gRNA libraries allows us to determine generalizable features associated with gRNA efficacy. Guides that target regions with low nucleosome occupancy and high chromatin accessibility are clearly more effective. We also find that the best region to target gRNAs is between the transcription start site (TSS) and 200 bp upstream of the TSS. Finally, unlike nuclease-proficient Cas9 in human cells, the specificity of truncated gRNAs (18 nt of complementarity to the target) is not clearly superior to full-length gRNAs (20 nt of complementarity), as truncated gRNAs are generally less potent against both mismatched and perfectly matched targets. CONCLUSIONS: Our results establish a powerful functional and chemical genomics screening method and provide guidelines for designing effective gRNAs, which consider chromatin state and position relative to the target gene TSS. These findings will enable effective library design and genome-wide programmable gene repression in many genetic backgrounds.

Conditional Epistatic Interaction Maps Reveal Global Functional Rewiring of Genome Integrity Pathways in Escherichia coli.

As antibiotic resistance is increasingly becoming a public health concern, an improved understanding of the bacterial DNA damage response (DDR), which is commonly targeted by antibiotics, could be of tremendous therapeutic value. Although the genetic components of the bacterial DDR have been studied extensively in isolation, how the underlying biological pathways interact functionally remains unclear. Here, we address this by performing systematic, unbiased, quantitative synthetic genetic interaction (GI) screens and uncover widespread changes in the GI network of the entire genomic integrity apparatus of Escherichia coli under standard and DNA-damaging growth conditions. The GI patterns of untreated cultures implicated two previously uncharacterized proteins (YhbQ and YqgF) as nucleases, whereas reorganization of the GI network after DNA damage revealed DDR roles for both annotated and uncharacterized genes. Analyses of pan-bacterial conservation patterns suggest that DDR mechanisms and functional relationships are near universal, highlighting a modular and highly adaptive genomic stress response.

Uncovering Phosphorylation-Based Specificities through Functional Interaction Networks.

Protein kinases are an important class of enzymes involved in the phosphorylation of their targets, which regulate key cellular processes and are typically mediated by a specificity for certain residues around the target phospho-acceptor residue. While efforts have been made to identify such specificities, only ∼30% of human kinases have a significant number of known binding sites. We describe a computational method that utilizes functional interaction data and phosphorylation data to predict specificities of kinases. We applied this method to human kinases to predict substrate preferences for 57% of all known kinases and show that we are able to reconstruct well-known specificities. We used an in vitro mass spectrometry approach to validate four understudied kinases and show that predicted models closely resemble true specificities. We show that this method can be applied to different organisms and can be extended to other phospho-recognition domains. Applying this approach to different types of posttranslational modifications (PTMs) and binding domains could uncover specificities of understudied PTM recognition domains and provide significant insight into the mechanisms of signaling networks.

Genetic Interaction Scoring Procedure for Bacterial Species.

A genetic interaction occurs when the phenotype of an organism carrying two mutant genes differs from what should have been observed given their independent influence. Such unexpected outcome indicates a mechanistic connection between the perturbed genes, providing a key source of functional information about the cell. Large-scale screening for genetic interactions involves measuring phenotypes of single and double mutants, which for microorganisms is usually done by automated analysis of images of ordered colonies. Obtaining accurate colony sizes, and using them to identify genetic interactions from such screens remains a challenging and time-consuming task. Here, we outline steps to compute genetic interaction scores in E. coli by measuring colony sizes from plate images, performing normalisation, and quantifying the strength of the effect.

Prediction of Functionally Important Phospho-Regulatory Events in Xenopus laevis Oocytes.

The African clawed frog Xenopus laevis is an important model organism for studies in developmental and cell biology, including cell-signaling. However, our knowledge of X. laevis protein post-translational modifications remains scarce. Here, we used a mass spectrometry-based approach to survey the phosphoproteome of this species, compiling a list of 2636 phosphosites. We used structural information and phosphoproteomic data for 13 other species in order to predict functionally important phospho-regulatory events. We found that the degree of conservation of phosphosites across species is predictive of sites with known molecular function. In addition, we predicted kinase-protein interactions for a set of cell-cycle kinases across all species. The degree of conservation of kinase-protein interactions was found to be predictive of functionally relevant regulatory interactions. Finally, using comparative protein structure models, we find that phosphosites within structured domains tend to be located at positions with high conformational flexibility. Our analysis suggests that a small class of phosphosites occurs in positions that have the potential to regulate protein conformation.

MIMP: predicting the impact of mutations on kinase-substrate phosphorylation.

Protein phosphorylation is important in cellular pathways and altered in disease. We developed MIMP (http://mimp.baderlab.org/), a machine learning method to predict the impact of missense single-nucleotide variants (SNVs) on kinase-substrate interactions. MIMP analyzes kinase sequence specificities and predicts whether SNVs disrupt existing phosphorylation sites or create new sites. This helps discover mutations that modify protein function by altering kinase networks and provides insight into disease biology and therapy development.

Evolutionary constraint and disease associations of post-translational modification sites in human genomes.

Interpreting the impact of human genome variation on phenotype is challenging. The functional effect of protein-coding variants is often predicted using sequence conservation and population frequency data, however other factors are likely relevant. We hypothesized that variants in protein post-translational modification (PTM) sites contribute to phenotype variation and disease. We analyzed fraction of rare variants and non-synonymous to synonymous variant ratio (Ka/Ks) in 7,500 human genomes and found a significant negative selection signal in PTM regions independent of six factors, including conservation, codon usage, and GC-content, that is widely distributed across tissue-specific genes and function classes. PTM regions are also enriched in known disease mutations, suggesting that PTM variation is more likely deleterious. PTM constraint also affects flanking sequence around modified residues and increases around clustered sites, indicating presence of functionally important short linear motifs. Using target site motifs of 124 kinases, we predict that at least ∼180,000 motif-breaker amino acid residues that disrupt PTM sites when substituted, and highlight kinase motifs that show specific negative selection and enrichment of disease mutations. We provide this dataset with corresponding hypothesized mechanisms as a community resource. As an example of our integrative approach, we propose that PTPN11 variants in Noonan syndrome aberrantly activate the protein by disrupting an uncharacterized cluster of phosphorylation sites. Further, as PTMs are molecular switches that are modulated by drugs, we study mutated binding sites of PTM enzymes in disease genes and define a drug-disease network containing 413 novel predicted disease-gene links.

Quantitative genome-wide genetic interaction screens reveal global epistatic relationships of protein complexes in Escherichia coli.

Large-scale proteomic analyses in Escherichia coli have documented the composition and physical relationships of multiprotein complexes, but not their functional organization into biological pathways and processes. Conversely, genetic interaction (GI) screens can provide insights into the biological role(s) of individual gene and higher order associations. Combining the information from both approaches should elucidate how complexes and pathways intersect functionally at a systems level. However, such integrative analysis has been hindered due to the lack of relevant GI data. Here we present a systematic, unbiased, and quantitative synthetic genetic array screen in E. coli describing the genetic dependencies and functional cross-talk among over 600,000 digenic mutant combinations. Combining this epistasis information with putative functional modules derived from previous proteomic data and genomic context-based methods revealed unexpected associations, including new components required for the biogenesis of iron-sulphur and ribosome integrity, and the interplay between molecular chaperones and proteases. We find that functionally-linked genes co-conserved among γ-proteobacteria are far more likely to have correlated GI profiles than genes with divergent patterns of evolution. Overall, examining bacterial GIs in the context of protein complexes provides avenues for a deeper mechanistic understanding of core microbial systems.

gitter: a robust and accurate method for quantification of colony sizes from plate images.

Colony-based screens that quantify the fitness of clonal populations on solid agar plates are perhaps the most important source of genome-scale functional information in microorganisms. The images of ordered arrays of mutants produced by such experiments can be difficult to process because of laboratory-specific plate features, morphed colonies, plate edges, noise, and other artifacts. Most of the tools developed to address this problem are optimized to handle a single setup and do not work out of the box in other settings. We present gitter, an image analysis tool for robust and accurate processing of images from colony-based screens. gitter works by first finding the grid of colonies from a preprocessed image and then locating the bounds of each colony separately. We show that gitter produces comparable colony sizes to other tools in simple cases but outperforms them by being able to handle a wider variety of screens and more accurately quantify colony sizes from difficult images. gitter is freely available as an R package from http://cran.r-project.org/web/packages/gitter under the LGPL. Tutorials and demos can be found at http://omarwagih.github.io/gitter.

The mutational landscape of phosphorylation signaling in cancer.

Somatic mutations in cancer genomes include drivers that provide selective advantages to tumor cells and passengers present due to genome instability. Discovery of pan-cancer drivers will help characterize biological systems important in multiple cancers and lead to development of better therapies. Driver genes are most often identified by their recurrent mutations across tumor samples. However, some mutations are more important for protein function than others. Thus considering the location of mutations with respect to functional protein sites can predict their mechanisms of action and improve the sensitivity of driver gene detection. Protein phosphorylation is a post-translational modification central to cancer biology and treatment, and frequently altered by driver mutations. Here we used our ActiveDriver method to analyze known phosphorylation sites mutated by single nucleotide variants (SNVs) in The Cancer Genome Atlas Research Network (TCGA) pan-cancer dataset of 3,185 genomes and 12 cancer types. Phosphorylation-related SNVs (pSNVs) occur in ~90% of tumors, show increased conservation and functional mutation impact compared to other protein-coding mutations, and are enriched in cancer genes and pathways. Gene-centric analysis found 150 known and candidate cancer genes with significant pSNV recurrence. Using a novel computational method, we predict that 29% of these mutations directly abolish phosphorylation or modify kinase target sites to rewire signaling pathways. This analysis shows that incorporation of information about protein signaling sites will improve computational pipelines for variant function prediction.

SGAtools: one-stop analysis and visualization of array-based genetic interaction screens.

Screening genome-wide sets of mutants for fitness defects provides a simple but powerful approach for exploring gene function, mapping genetic networks and probing mechanisms of drug action. For yeast and other microorganisms with global mutant collections, genetic or chemical-genetic interactions can be effectively quantified by growing an ordered array of strains on agar plates as individual colonies, and then scoring the colony size changes in response to a genetic or environmental perturbation. To do so, requires efficient tools for the extraction and analysis of quantitative data. Here, we describe SGAtools (http://sgatools.ccbr.utoronto.ca), a web-based analysis system for designer genetic screens. SGAtools outlines a series of guided steps that allow the user to quantify colony sizes from images of agar plates, correct for systematic biases in the observations and calculate a fitness score relative to a control experiment. The data can also be visualized online to explore the colony sizes on individual plates, view the distribution of resulting scores, highlight genes with the strongest signal and perform Gene Ontology enrichment analysis.