Tissue-aware interpretation of genetic variants advances the etiology of rare diseases.

Pathogenic variants underlying Mendelian diseases often disrupt the normal physiology of a few tissues and organs. However, variant effect prediction tools that aim to identify pathogenic variants are typically oblivious to tissue contexts. Here we report a machine-learning framework, denoted "Tissue Risk Assessment of Causality by Expression for variants" (TRACEvar, https://netbio.bgu.ac.il/TRACEvar/ ), that offers two advancements. First, TRACEvar predicts pathogenic variants that disrupt the normal physiology of specific tissues. This was achieved by creating 14 tissue-specific models that were trained on over 14,000 variants and combined 84 attributes of genetic variants with 495 attributes derived from tissue omics. TRACEvar outperformed 10 well-established and tissue-oblivious variant effect prediction tools. Second, the resulting models are interpretable, thereby illuminating variants' mode of action. Application of TRACEvar to variants of 52 rare-disease patients highlighted pathogenicity mechanisms and relevant disease processes. Lastly, the interpretation of all tissue models revealed that top-ranking determinants of pathogenicity included attributes of disease-affected tissues, particularly cellular process activities. Collectively, these results show that tissue contexts and interpretable machine-learning models can greatly enhance the etiology of rare diseases.

sNucConv: A bulk RNA-seq deconvolution method trained on single-nucleus RNA-seq data to estimate cell-type composition of human adipose tissues.

Deconvolution algorithms mostly rely on single-cell RNA-sequencing (scRNA-seq) data applied onto bulk RNA-sequencing (bulk RNA-seq) to estimate tissues' cell-type composition, with performance accuracy validated on deposited databases. Adipose tissues' cellular composition is highly variable, and adipocytes can only be captured by single-nucleus RNA-sequencing (snRNA-seq). Here we report the development of sNucConv, a Scaden deep-learning-based deconvolution tool, trained using 5 hSAT and 7 hVAT snRNA-seq-based data corrected by (i) snRNA-seq/bulk RNA-seq highly correlated genes and (ii) individual cell-type regression models. Applying sNucConv on our bulk RNA-seq data resulted in cell-type proportion estimation of 15 and 13 cell types, with accuracy of R = 0.93 (range: 0.76-0.97) and R = 0.95 (range: 0.92-0.98) for hVAT and hSAT, respectively. This performance level was further validated on an independent set of samples (5 hSAT; 5 hVAT). The resulting model was depot specific, reflecting depot differences in gene expression patterns. Jointly, sNucConv provides proof-of-concept for producing validated deconvolution models for tissues un-amenable to scRNA-seq.

Machine-learning analysis reveals an important role for negative selection in shaping cancer aneuploidy landscapes.

BACKGROUND: Aneuploidy, an abnormal number of chromosomes within a cell, is a hallmark of cancer. Patterns of aneuploidy differ across cancers, yet are similar in cancers affecting closely related tissues. The selection pressures underlying aneuploidy patterns are not fully understood, hindering our understanding of cancer development and progression. RESULTS: Here, we apply interpretable machine learning methods to study tissue-selective aneuploidy patterns. We define 20 types of features corresponding to genomic attributes of chromosome-arms, normal tissues, primary tumors, and cancer cell lines (CCLs), and use them to model gains and losses of chromosome arms in 24 cancer types. To reveal the factors that shape the tissue-specific cancer aneuploidy landscapes, we interpret the machine learning models by estimating the relative contribution of each feature to the models. While confirming known drivers of positive selection, our quantitative analysis highlights the importance of negative selection for shaping aneuploidy landscapes. This is exemplified by tumor suppressor gene density being a better predictor of gain patterns than oncogene density, and vice versa for loss patterns. We also identify the importance of tissue-selective features and demonstrate them experimentally, revealing KLF5 as an important driver for chr13q gain in colon cancer. Further supporting an important role for negative selection in shaping the aneuploidy landscapes, we find compensation by paralogs to be among the top predictors of chromosome arm loss prevalence and demonstrate this relationship for one paralog interaction. Similar factors shape aneuploidy patterns in human CCLs, demonstrating their relevance for aneuploidy research. CONCLUSIONS: Our quantitative, interpretable machine learning models improve the understanding of the genomic properties that shape cancer aneuploidy landscapes.

Affected cell types for hundreds of Mendelian diseases revealed by analysis of human and mouse single-cell data.

Mendelian diseases tend to manifest clinically in certain tissues, yet their affected cell types typically remain elusive. Single-cell expression studies showed that overexpression of disease-associated genes may point to the affected cell types. Here, we developed a method that infers disease-affected cell types from the preferential expression of disease-associated genes in cell types (PrEDiCT). We applied PrEDiCT to single-cell expression data of six human tissues, to infer the cell types affected in Mendelian diseases. Overall, we inferred the likely affected cell types for 328 diseases. We corroborated our findings by literature text-mining, expert validation, and recapitulation in mouse corresponding tissues. Based on these findings, we explored characteristics of disease-affected cell types, showed that diseases manifesting in multiple tissues tend to affect similar cell types, and highlighted cases where gene functions could be used to refine inference. Together, these findings expand the molecular understanding of disease mechanisms and cellular vulnerability.

Multicellular communities are perturbed in the aging human brain and Alzheimer's disease.

The role of different cell types and their interactions in Alzheimer's disease (AD) is a complex and open question. Here, we pursued this question by assembling a high-resolution cellular map of the aging frontal cortex using single-nucleus RNA sequencing of 24 individuals with a range of clinicopathologic characteristics. We used this map to infer the neocortical cellular architecture of 638 individuals profiled by bulk RNA sequencing, providing the sample size necessary for identifying statistically robust associations. We uncovered diverse cell populations associated with AD, including a somatostatin inhibitory neuronal subtype and oligodendroglial states. We further identified a network of multicellular communities, each composed of coordinated subpopulations of neuronal, glial and endothelial cells, and we found that two of these communities are altered in AD. Finally, we used mediation analyses to prioritize cellular changes that might contribute to cognitive decline. Thus, our deconstruction of the aging neocortex provides a roadmap for evaluating the cellular microenvironments underlying AD and dementia.

ProAct: quantifying the differential activity of biological processes in tissues, cells, and user-defined contexts.

The distinct functions and phenotypes of human tissues and cells derive from the activity of biological processes that varies in a context-dependent manner. Here, we present the Process Activity (ProAct) webserver that estimates the preferential activity of biological processes in tissues, cells, and other contexts. Users can upload a differential gene expression matrix measured across contexts or cells, or use a built-in matrix of differential gene expression in 34 human tissues. Per context, ProAct associates gene ontology (GO) biological processes with estimated preferential activity scores, which are inferred from the input matrix. ProAct visualizes these scores across processes, contexts, and process-associated genes. ProAct also offers potential cell-type annotations for cell subsets, by inferring them from the preferential activity of 2001 cell-type-specific processes. Thus, ProAct output can highlight the distinct functions of tissues and cell types in various contexts, and can enhance cell-type annotation efforts. The ProAct webserver is available at https://netbio.bgu.ac.il/ProAct/.

Predicting molecular mechanisms of hereditary diseases by using their tissue-selective manifestation.

How do aberrations in widely expressed genes lead to tissue-selective hereditary diseases? Previous attempts to answer this question were limited to testing a few candidate mechanisms. To answer this question at a larger scale, we developed "Tissue Risk Assessment of Causality by Expression" (TRACE), a machine learning approach to predict genes that underlie tissue-selective diseases and selectivity-related features. TRACE utilized 4,744 biologically interpretable tissue-specific gene features that were inferred from heterogeneous omics datasets. Application of TRACE to 1,031 disease genes uncovered known and novel selectivity-related features, the most common of which was previously overlooked. Next, we created a catalog of tissue-associated risks for 18,927 protein-coding genes (https://netbio.bgu.ac.il/trace/). As proof-of-concept, we prioritized candidate disease genes identified in 48 rare-disease patients. TRACE ranked the verified disease gene among the patient's candidate genes significantly better than gene prioritization methods that rank by gene constraint or tissue expression. Thus, tissue selectivity combined with machine learning enhances genetic and clinical understanding of hereditary diseases.

The TissueNet v.3 Database: Protein-protein Interactions in Adult and Embryonic Human Tissue contexts.

Tissue contexts are extremely valuable when studying protein functions and their associated phenotypes. Recently, the study of proteins in tissue contexts was greatly facilitated by the availability of thousands of tissue transcriptomes. To provide access to these data we developed the TissueNet integrative database that displays protein-protein interactions (PPIs) in tissue contexts. Through TissueNet, users can create tissue-sensitive network views of the PPI landscape of query proteins. Unlike other tools, TissueNet output networks highlight tissue-specific and broadly expressed proteins, as well as over- and under-expressed proteins per tissue. The TissueNet v.3 upgrade has a much larger dataset of proteins and PPIs, and represents 125 adult tissues and seven embryonic tissues. Thus, TissueNet provides an extensive, quantitative, and user-friendly interface to study the roles of human proteins in adulthood and embryonic stages. TissueNet v3 is freely available at https://netbio.bgu.ac.il/tissuenet3.

The Organ-Disease Annotations (ODiseA) Database of Hereditary Diseases and Inflicted Tissues.

Hereditary diseases tend to manifest clinically in few selected tissues. Knowledge of those tissues is important for better understanding of disease mechanisms, which often remain elusive. However, information on the tissues inflicted by each disease is not easily obtainable. Well-established resources, such as the Online Mendelian Inheritance in Man (OMIM) database and Human Phenotype Ontology (HPO), report on a spectrum of disease manifestations, yet do not highlight the main inflicted tissues. The Organ-Disease Annotations (ODiseA) database contains 4,357 thoroughly-curated annotations for 2,181 hereditary diseases and 45 inflicted tissues. Additionally, ODiseA reports 692 annotations of 635 diseases and the pathogenic tissues where they emerge. ODiseA can be queried by disease, disease gene, or inflicted tissue. Owing to its expansive, high-quality annotations, ODiseA serves as a valuable and unique tool for biomedical and computational researchers studying genotype-phenotype relationships of hereditary diseases. ODiseA is available at https://netbio.bgu.ac.il/odisea.

The differential activity of biological processes in tissues and cell subsets can illuminate disease-related processes and cell-type identities.

MOTIVATION: The distinct functionalities of human tissues and cell types underlie complex phenotype-genotype relationships, yet often remain elusive. Harnessing the multitude of bulk and single-cell human transcriptomes while focusing on processes can help reveal these distinct functionalities. RESULTS: The Tissue-Process Activity (TiPA) method aims to identify processes that are preferentially active or under-expressed in specific contexts, by comparing the expression levels of process genes between contexts. We tested TiPA on 1579 tissue-specific processes and bulk tissue transcriptomes, finding that it performed better than another method. Next, we used TiPA to ask whether the activity of certain processes could underlie the tissue-specific manifestation of 1233 hereditary diseases. We found that 21% of the disease-causing genes indeed participated in such processes, thereby illuminating their genotype-phenotype relationships. Lastly, we applied TiPA to single-cell transcriptomes of 108 human cell types, revealing that process activities often match cell-type identities and can thus aid annotation efforts. Hence, differential activity of processes can highlight the distinct functionality of tissues and cells in a robust and meaningful manner. AVAILABILITY AND IMPLEMENTATION: TiPA code is available in GitHub (https://github.com/moranshar/TiPA). In addition, all data are available as part of the Supplementary Material. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Unraveling the hidden role of a uORF-encoded peptide as a kinase inhibitor of PKCs.

Approximately 40% of human messenger RNAs (mRNAs) contain upstream open reading frames (uORFs) in their 5' untranslated regions. Some of these uORF sequences, thought to attenuate scanning ribosomes or lead to mRNA degradation, were recently shown to be translated, although the function of the encoded peptides remains unknown. Here, we show a uORF-encoded peptide that exhibits kinase inhibitory functions. This uORF, upstream of the protein kinase C-eta (PKC-η) main ORF, encodes a peptide (uPEP2) containing the typical PKC pseudosubstrate motif present in all PKCs that autoinhibits their kinase activity. We show that uPEP2 directly binds to and selectively inhibits the catalytic activity of novel PKCs but not of classical or atypical PKCs. The endogenous deletion of uORF2 or its overexpression in MCF-7 cells revealed that the endogenously translated uPEP2 reduces the protein levels of PKC-η and other novel PKCs and restricts cell proliferation. Functionally, treatment of breast cancer cells with uPEP2 diminished cell survival and their migration and synergized with chemotherapy by interfering with the response to DNA damage. Furthermore, in a xenograft of MDA-MB-231 breast cancer tumor in mice models, uPEP2 suppressed tumor progression, invasion, and metastasis. Tumor histology showed reduced proliferation, enhanced cell death, and lower protein expression levels of novel PKCs along with diminished phosphorylation of PKC substrates. Hence, our study demonstrates that uORFs may encode biologically active peptides beyond their role as translation regulators of their downstream ORFs. Together, we point to a unique function of a uORF-encoded peptide as a kinase inhibitor, pertinent to cancer therapy.

Small heat-shock protein HSPB3 promotes myogenesis by regulating the lamin B receptor.

One of the critical events that regulates muscle cell differentiation is the replacement of the lamin B receptor (LBR)-tether with the lamin A/C (LMNA)-tether to remodel transcription and induce differentiation-specific genes. Here, we report that localization and activity of the LBR-tether are crucially dependent on the muscle-specific chaperone HSPB3 and that depletion of HSPB3 prevents muscle cell differentiation. We further show that HSPB3 binds to LBR in the nucleoplasm and maintains it in a dynamic state, thus promoting the transcription of myogenic genes, including the genes to remodel the extracellular matrix. Remarkably, HSPB3 overexpression alone is sufficient to induce the differentiation of two human muscle cell lines, LHCNM2 cells, and rhabdomyosarcoma cells. We also show that mutant R116P-HSPB3 from a myopathy patient with chromatin alterations and muscle fiber disorganization, forms nuclear aggregates that immobilize LBR. We find that R116P-HSPB3 is unable to induce myoblast differentiation and instead activates the unfolded protein response. We propose that HSPB3 is a specialized chaperone engaged in muscle cell differentiation and that dysfunctional HSPB3 causes neuromuscular disease by deregulating LBR.

The landscape of molecular chaperones across human tissues reveals a layered architecture of core and variable chaperones.

The sensitivity of the protein-folding environment to chaperone disruption can be highly tissue-specific. Yet, the organization of the chaperone system across physiological human tissues has received little attention. Through computational analyses of large-scale tissue transcriptomes, we unveil that the chaperone system is composed of core elements that are uniformly expressed across tissues, and variable elements that are differentially expressed to fit with tissue-specific requirements. We demonstrate via a proteomic analysis that the muscle-specific signature is functional and conserved. Core chaperones are significantly more abundant across tissues and more important for cell survival than variable chaperones. Together with variable chaperones, they form tissue-specific functional networks. Analysis of human organ development and aging brain transcriptomes reveals that these functional networks are established in development and decline with age. In this work, we expand the known functional organization of de novo versus stress-inducible eukaryotic chaperones into a layered core-variable architecture in multi-cellular organisms.

Dosage-sensitive molecular mechanisms are associated with the tissue-specificity of traits and diseases.

Hereditary diseases and complex traits often manifest in specific tissues, whereas their causal genes are expressed in many tissues that remain unaffected. Among the mechanisms that have been suggested for this enigmatic phenomenon is dosage-sensitive compensation by paralogs of causal genes. Accordingly, tissue-selectivity stems from dosage imbalance between causal genes and paralogs that occurs particularly in disease-susceptible tissues. Here, we used a large-scale dataset of thousands of tissue transcriptomes and applied a linear mixed model (LMM) framework to assess this and other dosage-sensitive mechanisms. LMM analysis of 382 hereditary diseases consistently showed evidence for dosage-sensitive compensation by paralogs across diseases subsets and susceptible tissues. LMM analysis of 135 candidate genes that are strongly associated with 16 tissue-selective complex traits revealed a similar tendency among half of the trait-associated genes. This suggests that dosage-sensitive compensation by paralogs affects the tissue-selectivity of complex traits, and can be used to illuminate candidate genes' modes of action. Next, we applied LMM to analyze dosage imbalance between causal genes and three classes of genetic modifiers, including regulatory micro-RNAs, pseudogenes, and genetic interactors. Our results propose modifiers as a fundamental axis in tissue-selectivity of diseases and traits, and demonstrates the power of LMM as a statistical framework for discovering treatment avenues.

Mechanisms of tissue and cell-type specificity in heritable traits and diseases.

Hundreds of heritable traits and diseases that are caused by germline aberrations in ubiquitously expressed genes manifest in a remarkably limited number of cell types and tissues across the body. Unravelling mechanisms that govern their tissue-specific manifestations is critical for our understanding of disease aetiologies and may direct efforts to develop treatments. Owing to recent advances in high-throughput technologies and open resources, data and tools are now available to approach this enigmatic phenomenon at large scales, both computationally and experimentally. Here, we discuss the large prevalence of tissue-selective traits and diseases, describe common molecular mechanisms underlying their tissue-selective manifestation and present computational strategies and publicly available resources for elucidating the molecular basis of their genotype-phenotype relationships.

Differential network analysis of multiple human tissue interactomes highlights tissue-selective processes and genetic disorder genes.

MOTIVATION: Differential network analysis, designed to highlight network changes between conditions, is an important paradigm in network biology. However, differential network analysis methods have been typically designed to compare between two conditions and were rarely applied to multiple protein interaction networks (interactomes). Importantly, large-scale benchmarks for their evaluation have been lacking. RESULTS: Here, we present a framework for assessing the ability of differential network analysis of multiple human tissue interactomes to highlight tissue-selective processes and disorders. For this, we created a benchmark of 6499 curated tissue-specific Gene Ontology biological processes. We applied five methods, including four differential network analysis methods, to construct weighted interactomes for 34 tissues. Rigorous assessment of this benchmark revealed that differential analysis methods perform well in revealing tissue-selective processes (AUCs of 0.82-0.9). Next, we applied differential network analysis to illuminate the genes underlying tissue-selective hereditary disorders. For this, we curated a dataset of 1305 tissue-specific hereditary disorders and their manifesting tissues. Focusing on subnetworks containing the top 1% differential interactions in disease-relevant tissue interactomes revealed significant enrichment for disorder-causing genes in 18.6% of the cases, with a significantly high success rate for blood, nerve, muscle and heart diseases. SUMMARY: Altogether, we offer a framework that includes expansive manually curated datasets of tissue-selective processes and disorders to be used as benchmarks or to illuminate tissue-selective processes and genes. Our results demonstrate that differential analysis of multiple human tissue interactomes is a powerful tool for highlighting processes and genes with tissue-selective functionality and clinical impact. AVAILABILITY AND IMPLEMENTATION: Datasets are available as part of the Supplementary data. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Aging promotes reorganization of the CD4 T cell landscape toward extreme regulatory and effector phenotypes.

Age-associated changes in CD4 T-cell functionality have been linked to chronic inflammation and decreased immunity. However, a detailed characterization of CD4 T cell phenotypes that could explain these dysregulated functional properties is lacking. We used single-cell RNA sequencing and multidimensional protein analyses to profile thousands of CD4 T cells obtained from young and old mice. We found that the landscape of CD4 T cell subsets differs markedly between young and old mice, such that three cell subsets-exhausted, cytotoxic, and activated regulatory T cells (aTregs)-appear rarely in young mice but gradually accumulate with age. Most unexpected were the extreme pro- and anti-inflammatory phenotypes of cytotoxic CD4 T cells and aTregs, respectively. These findings provide a comprehensive view of the dynamic reorganization of the CD4 T cell milieu with age and illuminate dominant subsets associated with chronic inflammation and immunity decline, suggesting new therapeutic avenues for age-related diseases.

ResponseNet v.3: revealing signaling and regulatory pathways connecting your proteins and genes across human tissues.

ResponseNet v.3 is an enhanced version of ResponseNet, a web server that is designed to highlight signaling and regulatory pathways connecting user-defined proteins and genes by using the ResponseNet network optimization approach (http://netbio.bgu.ac.il/respnet). Users run ResponseNet by defining source and target sets of proteins, genes and/or microRNAs, and by specifying a molecular interaction network (interactome). The output of ResponseNet is a sparse, high-probability interactome subnetwork that connects the two sets, thereby revealing additional molecules and interactions that are involved in the studied condition. In recent years, massive efforts were invested in profiling the transcriptomes of human tissues, enabling the inference of human tissue interactomes. ResponseNet v.3 expands ResponseNet2.0 by harnessing ∼11,600 RNA-sequenced human tissue profiles made available by the Genotype-Tissue Expression consortium, to support context-specific analysis of 44 human tissues. Thus, ResponseNet v.3 allows users to illuminate the signaling and regulatory pathways potentially active in the context of a specific tissue, and to compare them with active pathways in other tissues. In the era of precision medicine, such analyses open the door for tissue- and patient-specific analyses of pathways and diseases.

Large-scale analysis of human gene expression variability associates highly variable drug targets with lower drug effectiveness and safety.

MOTIVATION: The effectiveness of drugs tends to vary between patients. One of the well-known reasons for this phenomenon is genetic polymorphisms in drug target genes among patients. Here, we propose that differences in expression levels of drug target genes across individuals can also contribute to this phenomenon. RESULTS: To explore this hypothesis, we analyzed the expression variability of protein-coding genes, and particularly drug target genes, across individuals. For this, we developed a novel variability measure, termed local coefficient of variation (LCV), which ranks the expression variability of each gene relative to genes with similar expression levels. Unlike commonly used methods, LCV neutralizes expression levels biases without imposing any distribution over the variation and is robust to data incompleteness. Application of LCV to RNA-sequencing profiles of 19 human tissues and to target genes of 1076 approved drugs revealed that drug target genes were significantly more variable than protein-coding genes. Analysis of 113 drugs with available effectiveness scores showed that drugs targeting highly variable genes tended to be less effective in the population. Furthermore, comparison of approved drugs to drugs that were withdrawn from the market showed that withdrawn drugs targeted significantly more variable genes than approved drugs. Last, upon analyzing gender differences we found that the variability of drug target genes was similar between men and women. Altogether, our results suggest that expression variability of drug target genes could contribute to the variable responsiveness and effectiveness of drugs, and is worth considering during drug treatment and development. AVAILABILITY AND IMPLEMENTATION: LCV is available as a python script in GitHub (https://github.com/eyalsim/LCV). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Integrating Rio1 activities discloses its nutrient-activated network in Saccharomyces cerevisiae.

The Saccharomyces cerevisiae kinase/adenosine triphosphatase Rio1 regulates rDNA transcription and segregation, pre-rRNA processing and small ribosomal subunit maturation. Other roles are unknown. When overexpressed, human ortholog RIOK1 drives tumor growth and metastasis. Likewise, RIOK1 promotes 40S ribosomal subunit biogenesis and has not been characterized globally. We show that Rio1 manages directly and via a series of regulators, an essential signaling network at the protein, chromatin and RNA levels. Rio1 orchestrates growth and division depending on resource availability, in parallel to the nutrient-activated Tor1 kinase. To define the Rio1 network, we identified its physical interactors, profiled its target genes/transcripts, mapped its chromatin-binding sites and integrated our data with yeast's protein-protein and protein-DNA interaction catalogs using network computation. We experimentally confirmed network components and localized Rio1 also to mitochondria and vacuoles. Via its network, Rio1 commands protein synthesis (ribosomal gene expression, assembly and activity) and turnover (26S proteasome expression), and impinges on metabolic, energy-production and cell-cycle programs. We find that Rio1 activity is conserved to humans and propose that pathological RIOK1 may fuel promiscuous transcription, ribosome production, chromosomal instability, unrestrained metabolism and proliferation; established contributors to cancer. Our study will advance the understanding of numerous processes, here revealed to depend on Rio1 activity.