HIPred: an integrative approach to predicting haploinsufficient genes

Intratumor heterogeneity of HPV integration in HPV-associated head and neck cancer

Integration of human papillomavirus (HPV) into the host genome drives HPV-positive head and neck squamous cell carcinoma (HPV+ HNSCC). Whole-genome sequencing of 51 tumors revealed intratumor heterogeneity of HPV integration, with 44% of breakpoints subclonal, and a biased distribution of integration breakpoints across the HPV genome. Four HPV physical states were identified, with at least 49% of tumors progressing without integration. HPV integration was associated with APOBEC-induced broad genomic instability and focal genomic instability, including structural variants at integration sites. HPV+ HNSCCs exhibited almost no smoking-induced mutational signatures. Heterozygous loss of ataxia-telangiectasia mutated (ATM) was observed in 67% of tumors, with its downregulation confirmed by single-cell RNA sequencing and immunohistochemistry, suggesting ATM haploinsufficiency contributes to carcinogenesis. PI3K activation was the major oncogenic mutation, with JAK-STAT activation in tumors with clonal integration and NF-kappa B activation in those without. These findings provide valuable insights into HPV integration in HPV+ HNSCC.

Proteome-scale prediction of molecular mechanisms underlying dominant genetic diseases

Many dominant genetic disorders result from protein-altering mutations, acting primarily through dominant-negative (DN), gain-of-function (GOF), and loss-of-function (LOF) mechanisms. Deciphering the mechanisms by which dominant diseases exert their effects is often experimentally challenging and resource intensive, but is essential for developing appropriate therapeutic approaches. Diseases that arise via a LOF mechanism are more amenable to be treated by conventional gene therapy, whereas DN and GOF mechanisms may require gene editing or targeting by small molecules. Moreover, pathogenic missense mutations that act via DN and GOF mechanisms are more difficult to identify than those that act via LOF using nearly all currently available variant effect predictors. Here, we introduce a tripartite statistical model made up of support vector machine binary classifiers trained to predict whether human protein coding genes are likely to be associated with DN, GOF, or LOF molecular disease mechanisms. We test the utility of the predictions by examining biologically and clinically meaningful properties known to be associated with the mechanisms. Our results strongly support that the models are able to generalise on unseen data and offer insight into the functional attributes of proteins associated with different mechanisms. We hope that our predictions will serve as a springboard for researchers studying novel variants and those of uncertain clinical significance, guiding variant interpretation strategies and experimental characterisation. Predictions for the human UniProt reference proteome are available at https://osf.io/z4dcp/.

DrivR-Base: a feature extraction toolkit for variant effect prediction model construction

MOTIVATION

Recent advancements in sequencing technologies have led to the discovery of numerous variants in the human genome. However, understanding their precise roles in diseases remains challenging due to their complex functional mechanisms. Various methodologies have emerged to predict the pathogenic significance of these genetic variants. Typically, these methods employ an integrative approach, leveraging diverse data sources that provide important insights into genomic function. Despite the abundance of publicly available data sources and databases, the process of navigating, extracting, and pre-processing features for machine learning models can be highly challenging and time-consuming. Furthermore, researchers often invest substantial effort in feature extraction, only to later discover that these features lack informativeness.

RESULTS

In this article, we introduce DrivR-Base, an innovative resource that efficiently extracts and integrates molecular information (features) related to single nucleotide variants. These features encompass information about the genomic positions and the associated protein positions of a variant. They are derived from a wide array of databases and tools, including structural properties obtained from AlphaFold, regulatory information sourced from ENCODE, and predicted variant consequences from Variant Effect Predictor. DrivR-Base is easily deployable via a Docker container to ensure reproducibility and ease of access across diverse computational environments. The resulting features can be used as input for machine learning models designed to predict the pathogenic impact of human genome variants in disease. Moreover, these feature sets have applications beyond this, including haploinsufficiency prediction and the development of drug repurposing tools. We describe the resource's development, practical applications, and potential for future expansion and enhancement.

AVAILABILITY AND IMPLEMENTATION

DrivR-Base source code is available at https://github.com/amyfrancis97/DrivR-Base.

Heterozygous Nexmif female mice demonstrate mosaic NEXMIF expression, autism-like behaviors, and abnormalities in dendritic arborization and synaptogenesis

Autism spectrum disorder (ASD) is a neurodevelopmental disorder with a strong genetic basis. ASDs are commonly characterized by impairments in language, restrictive and repetitive behaviors, and deficits in social interactions. Although ASD is a highly heterogeneous disease with many different genes implicated in its etiology, many ASD-associated genes converge on common cellular defects, such as aberrant neuronal morphology and synapse dysregulation. Our previous work revealed that, in mice, complete loss of the ASD-associated X-linked gene NEXMIF results in a reduction in dendritic complexity, a decrease in spine and synapse density, altered synaptic transmission, and ASD-like behaviors. Interestingly, human females of NEXMIF haploinsufficiency have recently been reported to demonstrate autistic features; however, the cellular and molecular basis for this haploinsufficiency-caused ASD remains unclear. Here we report that in the brains of Nexmif± female mice, NEXMIF shows a mosaic pattern in its expression in neurons. Heterozygous female mice demonstrate behavioral impairments similar to those of knockout male mice. In the mosaic mixture of neurons from Nexmif± mice, cells that lack NEXMIF have impairments in dendritic arborization and spine development. Remarkably, the NEXMIF-expressing neurons from Nexmif± mice also demonstrate similar defects in dendritic growth and spine formation. These findings establish a novel mouse model of NEXMIF haploinsufficiency and provide new insights into the pathogenesis of NEXMIF-dependent ASD.

Nemaline myopathy: reclassification of previously reported variants according to ACMG guidelines, and report of novel genetic variants

Nemaline myopathy (NM) is a heterogeneous genetic neuromuscular disorder characterized by rod bodies in muscle fibers resulting in multiple complications due to muscle weakness. NM patients and their families could benefit from genetic analysis for early diagnosis, carrier and prenatal testing; however, clinical classification of variants is subject to change as further information becomes available. Reclassification can significantly alter the clinical management of patients and their families. We used the newly published data and ACMG/AMP guidelines to reassess NM-associated variants previously reported by clinical laboratories (ClinVar). Our analyses on rare variants that were not canonical loss-of-function (LOF) resulted in the downgrading of ~29% (28/97) of variants from pathogenic or likely-pathogenic (P/LP) to variants of uncertain significance (VUS). In addition, we analyzed the splicing effect of variants identified in NM patients by clinical laboratories or research, using an accurate in silico prediction tool that applies a deep-learning network. We identified 55 rare variants that may impact splicing (cryptic splicing). We also analyzed six new NM families and identified eight variants in NEB and ACTA1, including three novel variants: homozygous pathogenic c.164A > G (p.Tyr55Cys), and homozygous likely pathogenic c.980T > C (p.Met327Thr) in ACTA1, and heterozygous VUS c.18694-3T > G in NEB. This study demonstrates the importance of reclassifying variants to facilitate more definitive "calls" on causality or no causality in clinical genetic testing of patients with NM. Reclassification of ~150 variants is now available for improved clinical management, risk counseling and screening of NM patients.

Large-scale across species transcriptomic analysis identifies genetic selection signatures associated with longevity in mammals

Lifespan varies significantly among mammals, with more than 100-fold difference between the shortest and longest living species. This natural difference may uncover the evolutionary forces and molecular features that define longevity. To understand the relationship between gene expression variation and longevity, we conducted a comparative transcriptomics analysis of liver, kidney, and brain tissues of 103 mammalian species. We found that few genes exhibit common expression patterns with longevity in the three organs analyzed. However, pathways related to translation fidelity, such as nonsense-mediated decay and eukaryotic translation elongation, correlated with longevity across mammals. Analyses of selection pressure found that selection intensity related to the direction of longevity-correlated genes is inconsistent across organs. Furthermore, expression of methionine restriction-related genes correlated with longevity and was under strong selection in long-lived mammals, suggesting that a common strategy is utilized by natural selection and artificial intervention to control lifespan. Our results indicate that lifespan regulation via gene expression is driven through polygenic and indirect natural selection.

Distinct longevity mechanisms across and within species and their association with aging

Lifespan varies within and across species, but the general principles of its control remain unclear. Here, we conducted multi-tissue RNA-seq analyses across 41 mammalian species, identifying longevity signatures and examining their relationship with transcriptomic biomarkers of aging and established lifespan-extending interventions. An integrative analysis uncovered shared longevity mechanisms within and across species, including downregulated Igf1 and upregulated mitochondrial translation genes, and unique features, such as distinct regulation of the innate immune response and cellular respiration. Signatures of long-lived species were positively correlated with age-related changes and enriched for evolutionarily ancient essential genes, involved in proteolysis and PI3K-Akt signaling. Conversely, lifespan-extending interventions counteracted aging patterns and affected younger, mutable genes enriched for energy metabolism. The identified biomarkers revealed longevity interventions, including KU0063794, which extended mouse lifespan and healthspan. Overall, this study uncovers universal and distinct strategies of lifespan regulation within and across species and provides tools for discovering longevity interventions.

Germline loss-of-function PAM variants are enriched in subjects with pituitary hypersecretion

Introduction

Pituitary adenomas (PAs) are common, usually benign tumors of the anterior pituitary gland which, for the most part, have no known genetic cause. PAs are associated with major clinical effects due to hormonal dysregulation and tumoral impingement on vital brain structures. PAM encodes a multifunctional protein responsible for the essential C-terminal amidation of secreted peptides.

Methods

Following the identification of a loss-of-function variant (p.Arg703Gln) in the peptidylglycine a-amidating monooxygenase (PAM) gene in a family with pituitary gigantism, we investigated 299 individuals with sporadic PAs and 17 familial isolated PA kindreds for PAM variants. Genetic screening was performed by germline and tumor sequencing and germline copy number variation (CNV) analysis.

Results

In germline DNA, we detected seven heterozygous, likely pathogenic missense, truncating, and regulatory SNVs. These SNVs were found in sporadic subjects with growth hormone excess (p.Gly552Arg and p.Phe759Ser), pediatric Cushing disease (c.-133T>C and p.His778fs), or different types of PAs (c.-361G>A, p.Ser539Trp, and p.Asp563Gly). The SNVs were functionally tested in vitro for protein expression and trafficking by Western blotting, splicing by minigene assays, and amidation activity in cell lysates and serum samples. These analyses confirmed a deleterious effect on protein expression and/or function. By interrogating 200,000 exomes from the UK Biobank, we confirmed a significant association of the PAM gene and rare PAM SNVs with diagnoses linked to pituitary gland hyperfunction.

Conclusion

The identification of PAM as a candidate gene associated with pituitary hypersecretion opens the possibility of developing novel therapeutics based on altering PAM function.

Buffering of genetic dominance by allele-specific protein complex assembly

Protein complex assembly often occurs while subunits are being translated, resulting in complexes whose subunits were translated from the same mRNA in an allele-specific manner. It has thus been hypothesized that such cotranslational assembly may counter the assembly-mediated dominant-negative effect, whereby co-assembly of mutant and wild-type subunits "poisons" complex activity. Here, we show that cotranslationally assembling subunits are much less likely to be associated with autosomal dominant relative to recessive disorders, and that subunits with dominant-negative disease mutations are significantly depleted in cotranslational assembly compared to those associated with loss-of-function mutations. We also find that complexes with known dominant-negative effects tend to expose their interfaces late during translation, lessening the likelihood of cotranslational assembly. Finally, by combining complex properties with other features, we trained a computational model for predicting proteins likely to be associated with non-loss-of-function disease mechanisms, which we believe will be of considerable utility for protein variant interpretation.

Transfer learning enables predictions in network biology

Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.

EpiTEAmDNA: Sequence feature representation via transfer learning and ensemble learning for identifying multiple DNA epigenetic modification types across species

Methylation is a major DNA epigenetic modification for regulating the biological processes without altering the DNA sequence, and multiple types of DNA methylations have been discovered, including 6mA, 5hmC, and 4mC. Multiple computational approaches were developed to automatically identify the DNA methylation residues using machine learning or deep learning algorithms. The machine learning (ML) based methods are difficult to be transferred to the other predicting tasks of the DNA methylation sites using additional knowledge. Deep learning (DL) may facilitate the transfer learning of knowledge from similar tasks, but they are often ineffective on small datasets. This study proposes an integrated feature representation framework EpiTEAmDNA based on the strategies of transfer learning and ensemble learning, which is evaluated on multiple DNA methylation types across 15 species. EpiTEAmDNA integrates convolutional neural network (CNN) and conventional machine learning methods, and shows improved performances than the existing DL-based methods on small datasets when no additional knowledge is available. The experimental data suggests that the EpiTEAmDNA models may be further improved via transfer learning based on additional knowledge. The evaluation experiments on the independent test datasets also suggest that the proposed EpiTEAmDNA framework outperforms the existing models in most prediction tasks of the 3 DNA methylation types across 15 species. The source code, pre-trained global model, and the EpiTEAmDNA feature representation framework are freely available at http://www.healthinformaticslab.org/supp/.

Faster and more accurate pathogenic combination predictions with VarCoPP2.0

BACKGROUND

The prediction of potentially pathogenic variant combinations in patients remains a key task in the field of medical genetics for the understanding and detection of oligogenic/multilocus diseases. Models tailored towards such cases can help shorten the gap of missing diagnoses and can aid researchers in dealing with the high complexity of the derived data. The predictor VarCoPP (Variant Combinations Pathogenicity Predictor) that was published in 2019 and identified potentially pathogenic variant combinations in gene pairs (bilocus variant combinations), was the first important step in this direction. Despite its usefulness and applicability, several issues still remained that hindered a better performance, such as its False Positive (FP) rate, the quality of its training set and its complex architecture.

RESULTS

We present VarCoPP2.0: the successor of VarCoPP that is a simplified, faster and more accurate predictive model identifying potentially pathogenic bilocus variant combinations. Results from cross-validation and on independent data sets reveal that VarCoPP2.0 has improved in terms of both sensitivity (95% in cross-validation and 98% during testing) and specificity (5% FP rate). At the same time, its running time shows a significant 150-fold decrease due to the selection of a simpler Balanced Random Forest model. Its positive training set now consists of variant combinations that are more confidently linked with evidence of pathogenicity, based on the confidence scores present in OLIDA, the Oligogenic Diseases Database ( https://olida.ibsquare.be ). The improvement of its performance is also attributed to a more careful selection of up-to-date features identified via an original wrapper method. We show that the combination of different variant and gene pair features together is important for predictions, highlighting the usefulness of integrating biological information at different levels.

CONCLUSIONS

Through its improved performance and faster execution time, VarCoPP2.0 enables a more accurate analysis of larger data sets linked to oligogenic diseases. Users can access the ORVAL platform ( https://orval.ibsquare.be ) to apply VarCoPP2.0 on their data.

Germline loss-of-function PAM variants are enriched in subjects with pituitary hypersecretion

Pituitary adenomas (PAs) are common, usually benign tumors of the anterior pituitary gland which, for the most part, have no known genetic cause. PAs are associated with major clinical effects due to hormonal dysregulation and tumoral impingement on vital brain structures. Following the identification of a loss-of-function variant (p.Arg703Gln) in the PAM gene in a family with pituitary gigantism, we investigated 299 individuals with sporadic PAs and 17 familial isolated pituitary adenomas kindreds for PAM variants. PAM encodes a multifunctional protein responsible for the essential C-terminal amidation of secreted peptides. Genetic screening was performed by germline and tumor sequencing and germline copy number variation (CNV) analysis. No germline CNVs or somatic single nucleotide variants (SNVs) were identified. We detected seven likely pathogenic heterozygous missense, truncating, and regulatory SNVs. These SNVs were found in sporadic subjects with GH excess (p.Gly552Arg and p.Phe759Ser), pediatric Cushing disease (c.-133T>C and p.His778fs), or with different types of PAs (c.-361G>A, p.Ser539Trp, and p.Asp563Gly). The SNVs were functionally tested in vitro for protein expression and trafficking by Western blotting, for splicing by minigene assays, and for amidation activity in cell lysates and serum samples. These analyses confirmed a deleterious effect on protein expression and/or function. By interrogating 200,000 exomes from the UK Biobank, we confirmed a significant association of the PAM gene and rare PAM SNVs to diagnoses linked to pituitary gland hyperfunction. Identification of PAM as a candidate gene associated with pituitary hypersecretion opens the possibility of developing novel therapeutics based on altering PAM function.

ERI1: A case report of an autosomal recessive syndrome associated with developmental delay and distal limb abnormalities

ERI1 is an evolutionary conserved 3'-5' exonuclease with an important function in multiple RNA processing pathways. Although the molecular mechanisms in which ERI1 is involved have been studied extensively in model organisms, the pathology associated with ERI1 variants in humans has remained elusive because no case has been reported so far. Here, we present a case of a female patient with a homozygous nonsense variant in ERI1 gene. The patient exhibits mild intellectual disability, eyelid ptosis, and anomalies in her hands and feet (brachydactyly, clinodactyly, dysplastic/short nail of halluces, brachytelephalangy, short metacarpals, and toe syndactyly). This case report is the first of its kind and is invaluable for understanding ERI1 pathology in humans.

Proteotype coevolution and quantitative diversity across 11 mammalian species

Evolutionary profiling has been largely limited to the nucleotide level. Using consistent proteomic methods, we quantified proteomic and phosphoproteomic layers in fibroblasts from 11 common mammalian species, with transcriptomes as reference. Covariation analysis indicates that transcript and protein expression levels and variabilities across mammals remarkably follow functional role, with extracellular matrix-associated expression being the most variable, demonstrating strong transcriptome-proteome coevolution. The biological variability of gene expression is universal at both interindividual and interspecies scales but to a different extent. RNA metabolic processes particularly show higher interspecies versus interindividual variation. Our results further indicate that while the ubiquitin-proteasome system is strongly conserved in mammals, lysosome-mediated protein degradation exhibits remarkable variation between mammalian lineages. In addition, the phosphosite profiles reveal a phosphorylation coevolution network independent of protein abundance.

Neurodevelopmental Disorders in Patients With Complex Phenotypes and Potential Complex Genetic Basis Involving Non-Coding Genes, and Double CNVs

Neurodevelopmental disorders (NDDs) are a heterogeneous class of brain diseases, with a complex genetic basis estimated to account for up to 50% of cases. Nevertheless, genetic diagnostic yield is about 20%. Array-comparative genomic hybridization (array-CGH) is an established first-level diagnostic test able to detect pathogenic copy number variants (CNVs), however, most identified variants remain of uncertain significance (VUS). Failure of interpretation of VUSs may depend on various factors, including complexity of clinical phenotypes and inconsistency of genotype-phenotype correlations. Indeed, although most NDD-associated CNVs are de novo, transmission from unaffected parents to affected children of CNVs with high risk for NDDs has been observed. Moreover, variability of genetic components overlapped by CNVs, such as long non-coding genes, genomic regions with long-range effects, and additive effects of multiple CNVs can make CNV interpretation challenging. We report on 12 patients with complex phenotypes possibly explained by complex genetic mechanisms, including involvement of antisense genes and boundaries of topologically associating domains. Eight among the 12 patients carried two CNVs, either de novo or inherited, respectively, by each of their healthy parents, that could additively contribute to the patients’ phenotype. CNVs overlapped either known NDD-associated or novel candidate genes (PTPRD, BUD13, GLRA3, MIR4465, ABHD4, and WSCD2). Bioinformatic enrichment analyses showed that genes overlapped by the co-occurring CNVs have synergistic roles in biological processes fundamental in neurodevelopment. Double CNVs could concur in producing deleterious effects, according to a two-hit model, thus explaining the patients’ phenotypes and the incomplete penetrance, and variable expressivity, associated with the single variants. Overall, our findings could contribute to the knowledge on clinical and genetic diagnosis of complex forms of NDD.

Integration of Data from Liquid-Liquid Phase Separation Databases Highlights Concentration and Dosage Sensitivity of LLPS Drivers

Liquid–liquid phase separation (LLPS) is a molecular process that leads to the formation of membraneless organelles, representing functionally specialized liquid-like cellular condensates formed by proteins and nucleic acids. Integrating the data on LLPS-associated proteins from dedicated databases revealed only modest agreement between them and yielded a high-confidence dataset of 89 human LLPS drivers. Analysis of the supporting evidence for our dataset uncovered a systematic and potentially concerning difference between protein concentrations used in a good fraction of the in vitro LLPS experiments, a key parameter that governs the phase behavior, and the proteomics-derived cellular abundance levels of the corresponding proteins. Closer scrutiny of the underlying experimental data enabled us to offer a sound rationale for this systematic difference, which draws on our current understanding of the cellular organization of the proteome and the LLPS process. In support of this rationale, we find that genes coding for our human LLPS drivers tend to be dosage-sensitive, suggesting that their cellular availability is tightly regulated to preserve their functional role in direct or indirect relation to condensate formation. Our analysis offers guideposts for increasing agreement between in vitro and in vivo studies, probing the roles of proteins in LLPS.

Case Report: Whole Exome Sequencing Revealed Disease-Causing Variants in Two Genes in a Patient With Autism Spectrum Disorder, Intellectual Disability, Hyperactivity, Sleep and Gastrointestinal Disturbances

Autism Spectrum Disorder (ASD) refers to a broad range of conditions characterized by difficulties in communication, social interaction and behavior, and may be accompanied by other medical or psychiatric conditions. Patients with ASD and comorbidities are often difficult to diagnose because of the tendency to consider the multiple symptoms as the presentation of a complicated syndromic form. This view influences variant filtering which might ignore causative variants for specific clinical features shown by the patient. Here we report on a male child diagnosed with ASD, showing cognitive and motor impairments, stereotypies, hyperactivity, sleep, and gastrointestinal disturbances. The analysis of whole exome sequencing (WES) data with bioinformatic tools for oligogenic diseases helped us to identify two major previously unreported pathogenetic variants: a maternally inherited missense variant (p.R4122H) in HUWE1, an ubiquitin protein ligase associated to X-linked intellectual disability and ASD; and a de novo stop variant (p.Q259X) in TPH2, encoding the tryptophan hydroxylase 2 enzyme involved in serotonin synthesis and associated with susceptibility to attention deficit-hyperactivity disorder (ADHD). TPH2, expressed in central and peripheral nervous tissues, modulates various physiological functions, including gut motility and sleep. To the best of our knowledge, this is the first case presenting with ASD, cognitive impairment, sleep, and gastrointestinal disturbances linked to both HUWE1 and TPH2 genes. Our findings could contribute to the existing knowledge on clinical and genetic diagnosis of patients with ASD presentation with comorbidities.

Of numbers and movement - understanding transcription factor pathogenesis by advanced microscopy

Transcription factors (TFs) are life-sustaining and, therefore, the subject of intensive research. By regulating gene expression, TFs control a plethora of developmental and physiological processes, and their abnormal function commonly leads to various developmental defects and diseases in humans. Normal TF function often depends on gene dosage, which can be altered by copy-number variation or loss-of-function mutations. This explains why TF haploinsufficiency (HI) can lead to disease. Since aberrant TF numbers frequently result in pathogenic abnormalities of gene expression, quantitative analyses of TFs are a priority in the field. In vitro single-molecule methodologies have significantly aided the identification of links between TF gene dosage and transcriptional outcomes. Additionally, advances in quantitative microscopy have contributed mechanistic insights into normal and aberrant TF function. However, to understand TF biology, TF-chromatin interactions must be characterised in vivo, in a tissue-specific manner and in the context of both normal and altered TF numbers. Here, we summarise the advanced microscopy methodologies most frequently used to link TF abundance to function and dissect the molecular mechanisms underlying TF HIs. Increased application of advanced single-molecule and super-resolution microscopy modalities will improve our understanding of how TF HIs drive disease.

Transcriptome and translatome co-evolution in mammals

Gene expression programs define shared and species-specific phenotypes, but their evolution remains largely uncharacterized beyond the transcriptome layer¹. Here we report an analysis of the co-evolution of translatomes and transcriptomes using ribosome-profling and matched RNA-sequencing data for three organs (brain, liver and testis) in fve mammals (human, macaque, mouse, opossum and platypus) and a bird (chicken). Our within-species analyses reveal that translational regulation is widespread in the diferent organs, in particular across the spermatogenic cell types of the testis. The between-species divergence in gene expression is around 20% lower at the translatome layer than at the transcriptome layer owing to extensive buffering between the expression layers, which especially preserved old, essential and housekeeping genes. Translational upregulation specifcally counterbalanced global dosage reductions during the evolution of sex chromosomes and the efects of meiotic sex-chromosome inactivation during spermatogenesis. Despite the overall prevalence of bufering, some genes evolved faster at the translatome layer—potentially indicating adaptive changes in expression; testis tissue shows the highest fraction of such genes. Further analyses incorporating mass spectrometry proteomics data establish that the co-evolution of transcriptomes and translatomes is refected at the proteome layer. Together, our work uncovers co-evolutionary patterns and associated selective forces across the expression layers, and provides a resource for understanding their interplay in mammalian organs.

Promoter CpG Density Predicts Downstream Gene Loss-of-Function Intolerance

The aggregation and joint analysis of large numbers of exome sequences has recently made it possible to derive estimates of intolerance to loss-of-function (LoF) variation for human genes. Here, we demonstrate strong and widespread coupling between genic LoF intolerance and promoter CpG density across the human genome. Genes downstream of the most CpG-rich promoters (top 10% CpG density) have a 67.2% probability of being highly LoF intolerant, using the LOEUF metric from gnomAD. This is in contrast to 7.4% of genes downstream of the most CpG-poor (bottom 10% CpG density) promoters. Combining promoter CpG density with exonic and promoter conservation explains 33.4% of the variation in LOEUF, and the contribution of CpG density exceeds the individual contributions of exonic and promoter conservation. We leverage this to train a simple and easily interpretable predictive model that outperforms other existing predictors and allows us to classify 1,760 genes-which are currently unascertained in gnomAD-as highly LoF intolerant or not. These predictions have the potential to aid in the interpretation of novel variants in the clinical setting. Moreover, our results reveal that high CpG density is not merely a generic feature of human promoters but is preferentially encountered at the promoters of the most selectively constrained genes, calling into question the prevailing view that CpG islands are not subject to selection.

Mutation severity spectrum of rare alleles in the human genome is predictive of disease type

The human genome harbors a variety of genetic variations. Single-nucleotide changes that alter amino acids in protein-coding regions are one of the major causes of human phenotypic variation and diseases. These single-amino acid variations (SAVs) are routinely found in whole genome and exome sequencing. Evaluating the functional impact of such genomic alterations is crucial for diagnosis of genetic disorders. We developed DeepSAV, a deep-learning convolutional neural network to differentiate disease-causing and benign SAVs based on a variety of protein sequence, structural and functional properties. Our method outperforms most stand-alone programs, and the version incorporating population and gene-level information (DeepSAV+PG) has similar predictive power as some of the best available. We transformed DeepSAV scores of rare SAVs in the human population into a quantity termed "mutation severity measure" for each human protein-coding gene. It reflects a gene's tolerance to deleterious missense mutations and serves as a useful tool to study gene-disease associations. Genes implicated in cancer, autism, and viral interaction are found by this measure as intolerant to mutations, while genes associated with a number of other diseases are scored as tolerant. Among known disease-associated genes, those that are mutation-intolerant are likely to function in development and signal transduction pathways, while those that are mutation-tolerant tend to encode metabolic and mitochondrial proteins.

Characterization of Human Dosage-Sensitive Transcription Factor Genes

Copy number changes in protein-coding genes are detrimental if the consequent changes in protein concentrations disrupt essential cellular functions. The dosage sensitivity of transcription factor (TF) genes is particularly interesting because their products are essential in regulating the expression of genetic information. From four recently curated data sets of dosage-sensitive genes (genes with conserved copy numbers across mammals, ohnologs, and two data sets of haploinsufficient genes), we compiled a data set of the most reliable dosage-sensitive (MRDS) genes and a data set of the most reliable dosage-insensitive (MRDIS) genes. The MRDS genes were those present in all four data sets, while the MRDIS genes were those absent from any one of the four data sets and with the probability of being loss of function-intolerant (pLI) values < 0.5 in both of the haploinsufficient gene data sets. Enrichment analysis of TF genes among the MRDS and MRDIS gene data sets showed that TF genes are more likely to be dosage-sensitive than other genes in the human genome. The nuclear receptor family was the most enriched TF family among the dosage-sensitive genes. TF families with very few members were also deemed more likely to be dosage-sensitive than TF families with more members. In addition, we found a certain number of dosage-insensitive TFs. The most typical were the Krüppel-associated box domain-containing zinc-finger proteins (KZFPs). Gene ontology (GO) enrichment analysis showed that the MRDS TFs were enriched for many more terms than the MRDIS TFs; however, the proteins interacting with these two groups of TFs did not show such sharp differences. Furthermore, we found that the MRDIS KZFPs were not significantly enriched for any GO terms, whereas their interacting proteins were significantly enriched for thousands of GO terms. Further characterizations revealed significant differences between MRDS TFs and MRDIS TFs in the lengths and nucleotide compositions of DNA-binding sites as well as in expression level, protein size, and selective force.

FATHMM-XF: accurate prediction of pathogenic point mutations via extended features

Summary

We present FATHMM-XF, a method for predicting pathogenic point mutations in the human genome. Drawing on an extensive feature set, FATHMM-XF outperforms competitors on benchmark tests, particularly in non-coding regions where the majority of pathogenic mutations are likely to be found.

Availability and implementation

The FATHMM-XF web server is available at http://fathmm.biocompute.org.uk/fathmm-xf/, and as tracks on the Genome Tolerance Browser: http://gtb.biocompute.org.uk. Predictions are provided for human genome version GRCh37/hg19. The data used for this project can be downloaded from: http://fathmm.biocompute.org.uk/fathmm-xf/

Supplementary information

Supplementary data are available at Bioinformatics online.

Epigenetic and Cellular Diversity in the Brain through Allele-Specific Effects

The benefits of diploidy are considered to involve masking partially recessive mutations and increasing genetic diversity. Here, we review new studies showing evidence for diverse allele-specific expression and epigenetic states in mammalian brain cells, which suggest that diploidy expands the landscape of gene regulatory and expression programs in cells. Allele-specific expression has been thought to be restricted to a few specific classes of genes. However, new studies show novel genomic imprinting effects that are brain-region-, cell-type- and age-dependent. In addition, novel forms of random monoallelic expression that impact many autosomal genes have been described in vitro and in vivo. We discuss the implications for understanding the benefits of diploidy, and the mechanisms shaping brain development, function, and disease.

CScape: a tool for predicting oncogenic single-point mutations in the cancer genome

For somatic point mutations in coding and non-coding regions of the genome, we propose CScape, an integrative classifier for predicting the likelihood that mutations are cancer drivers. Tested on somatic mutations, CScape tends to outperform alternative methods, reaching 91% balanced accuracy in coding regions and 70% in non-coding regions, while even higher accuracy may be achieved using thresholds to isolate high-confidence predictions. Positive predictions tend to cluster in genomic regions, so we apply a statistical approach to isolate coding and non-coding regions of the cancer genome that appear enriched for high-confidence predicted disease-drivers. Predictions and software are available at http://CScape.biocompute.org.uk/ .

Machine learning for epigenetics and future medical applications

Understanding epigenetic processes holds immense promise for medical applications. Advances in Machine Learning (ML) are critical to realize this promise. Previous studies used epigenetic data sets associated with the germline transmission of epigenetic transgenerational inheritance of disease and novel ML approaches to predict genome-wide locations of critical epimutations. A combination of Active Learning (ACL) and Imbalanced Class Learning (ICL) was used to address past problems with ML to develop a more efficient feature selection process and address the imbalance problem in all genomic data sets. The power of this novel ML approach and our ability to predict epigenetic phenomena and associated disease is suggested. The current approach requires extensive computation of features over the genome. A promising new approach is to introduce Deep Learning (DL) for the generation and simultaneous computation of novel genomic features tuned to the classification task. This approach can be used with any genomic or biological data set applied to medicine. The application of molecular epigenetic data in advanced machine learning analysis to medicine is the focus of this review.