Broker genes in human disease

Variation in Pleiotropic Hub Gene Expression Is Associated with Interspecific Differences in Head Shape and Eye Size in Drosophila

Revealing the mechanisms underlying the breathtaking morphological diversity observed in nature is a major challenge in Biology. It has been established that recurrent mutations in hotspot genes cause the repeated evolution of morphological traits, such as body pigmentation or the gain and loss of structures. To date, however, it remains elusive whether hotspot genes contribute to natural variation in the size and shape of organs. As natural variation in head morphology is pervasive in Drosophila, we studied the molecular and developmental basis of differences in compound eye size and head shape in two closely related Drosophila species. We show differences in the progression of retinal differentiation between species and we applied comparative transcriptomics and chromatin accessibility data to identify the GATA transcription factor Pannier (Pnr) as central factor associated with these differences. Although the genetic manipulation of Pnr affected multiple aspects of dorsal head development, the effect of natural variation is restricted to a subset of the phenotypic space. We present data suggesting that this developmental constraint is caused by the coevolution of expression of pnr and its cofactor u-shaped (ush). We propose that natural variation in expression or function of highly connected developmental regulators with pleiotropic functions is a major driver for morphological evolution and we discuss implications on gene regulatory network evolution. In comparison to previous findings, our data strongly suggest that evolutionary hotspots are not the only contributors to the repeated evolution of eye size and head shape in Drosophila.

Computational analysis of fused co-expression networks for the identification of candidate cancer gene biomarkers

The complexity of cancer has always been a huge issue in understanding the source of this disease. However, by appreciating its complexity, we can shed some light on crucial gene associations across and in specific cancer types. In this study, we develop a general framework to infer relevant gene biomarkers and their gene-to-gene associations using multiple gene co-expression networks for each cancer type. Specifically, we infer computationally and biologically interesting communities of genes from kidney renal clear cell carcinoma, liver hepatocellular carcinoma, and prostate adenocarcinoma data sets of The Cancer Genome Atlas (TCGA) database. The gene communities are extracted through a data-driven pipeline and then evaluated through both functional analyses and literature findings. Furthermore, we provide a computational validation of their relevance for each cancer type by comparing the performance of normal/cancer classification for our identified gene sets and other gene signatures, including the typically-used differentially expressed genes. The hallmark of this study is its approach based on gene co-expression networks from different similarity measures: using a combination of multiple gene networks and then fusing normal and cancer networks for each cancer type, we can have better insights on the overall structure of the cancer-type-specific network.

Boosting GWAS using biological networks: A study on susceptibility to familial breast cancer

Genome-wide association studies (GWAS) explore the genetic causes of complex diseases. However, classical approaches ignore the biological context of the genetic variants and genes under study. To address this shortcoming, one can use biological networks, which model functional relationships, to search for functionally related susceptibility loci. Many such network methods exist, each arising from different mathematical frameworks, pre-processing steps, and assumptions about the network properties of the susceptibility mechanism. Unsurprisingly, this results in disparate solutions. To explore how to exploit these heterogeneous approaches, we selected six network methods and applied them to GENESIS, a nationwide French study on familial breast cancer. First, we verified that network methods recovered more interpretable results than a standard GWAS. We addressed the heterogeneity of their solutions by studying their overlap, computing what we called the consensus. The key gene in this consensus solution was COPS5, a gene related to multiple cancer hallmarks. Another issue we observed was that network methods were unstable, selecting very different genes on different subsamples of GENESIS. Therefore, we proposed a stable consensus solution formed by the 68 genes most consistently selected across multiple subsamples. This solution was also enriched in genes known to be associated with breast cancer susceptibility (BLM, CASP8, CASP10, DNAJC1, FGFR2, MRPS30, and SLC4A7, P-value = 3 × 10-4). The most connected gene was CUL3, a regulator of several genes linked to cancer progression. Lastly, we evaluated the biases of each method and the impact of their parameters on the outcome. In general, network methods preferred highly connected genes, even after random rewirings that stripped the connections of any biological meaning. In conclusion, we present the advantages of network-guided GWAS, characterize their shortcomings, and provide strategies to address them. To compute the consensus networks, implementations of all six methods are available at https://github.com/hclimente/gwas-tools.

Where are the breaks in translation from theory to clinical practice (and back) in addressing depression? An empirical graph-theoretic approach

BACKGROUND

Research in depression has progressed rapidly over the past four decades. Yet depression rates are not subsiding and treatment success is not improving. We examine the extent to which the gap between science and practice is associated with the level of integration in how depression is considered in research and stakeholder-relevant documents.

METHODS

We used a network-science perspective to analyze similar uses of depression relevant terms in the Google News corpus (approximately 1 billion words) and the Web of Science database (120 000 documents).

RESULTS

These analyses yielded consistent pictures of insular modules associated with: (1) patient/providers, (2) academics, and (3) industry. Within academia insular modules associated with psychology, general medical, and psychiatry/neuroscience/biology were also detected.

CONCLUSIONS

These analyses suggest that the domain of depression is fragmented, and that advancements of relevance to one stakeholder group (academics, industry, or patients) may not translate to the others. We consider potential causes and associated responses to this fragmentation that could help to unify and advance translation from research on depression to the clinic, largely involving harmonizing employed language, bridging conceptual domains, and increasing communication across stakeholder groups.

A big data pipeline: Identifying dynamic gene regulatory networks from time-course Gene Expression Omnibus data with applications to influenza infection

A biological host response to an external stimulus or intervention such as a disease or infection is a dynamic process, which is regulated by an intricate network of many genes and their products. Understanding the dynamics of this gene regulatory network allows us to infer the mechanisms involved in a host response to an external stimulus, and hence aids the discovery of biomarkers of phenotype and biological function. In this article, we propose a modeling/analysis pipeline for dynamic gene expression data, called Pipeline4DGEData, which consists of a series of statistical modeling techniques to construct dynamic gene regulatory networks from the large volumes of high-dimensional time-course gene expression data that are freely available in the Gene Expression Omnibus repository. This pipeline has a consistent and scalable structure that allows it to simultaneously analyze a large number of time-course gene expression data sets, and then integrate the results across different studies. We apply the proposed pipeline to influenza infection data from nine studies and demonstrate that interesting biological findings can be discovered with its implementation.

Applications of molecular networks in biomedicine

Due to the large interdependence between the molecular components of living systems, many phenomena, including those related to pathologies, cannot be explained in terms of a single gene or a small number of genes. Molecular networks, representing different types of relationships between molecular entities, embody these large sets of interdependences in a framework that allow their mining from a systemic point of view to obtain information. These networks, often generated from high-throughput omics datasets, are used to study the complex phenomena of human pathologies from a systemic point of view. Complementing the reductionist approach of molecular biology, based on the detailed study of a small number of genes, systemic approaches to human diseases consider that these are better reflected in large and intricate networks of relationships between genes. These networks, and not the single genes, provide both better markers for diagnosing diseases and targets for treating them. Network approaches are being used to gain insight into the molecular basis of complex diseases and interpret the large datasets associated with them, such as genomic variants. Network formalism is also suitable for integrating large, heterogeneous and multilevel datasets associated with diseases from the molecular level to organismal and epidemiological scales. Many of these approaches are available to nonexpert users through standard software packages.

Identification of critical connectors in the directed reaction-centric graphs of microbial metabolic networks

BACKGROUND

Detection of central nodes in asymmetrically directed biological networks depends on centrality metrics quantifying individual nodes' importance in a network. In topological analyses on metabolic networks, various centrality metrics have been mostly applied to metabolite-centric graphs. However, centrality metrics including those not depending on high connections are largely unexplored for directed reaction-centric graphs.

RESULTS

We applied directed versions of centrality metrics to directed reaction-centric graphs of microbial metabolic networks. To investigate the local role of a node, we developed a novel metric, cascade number, considering how many nodes are closed off from information flow when a particular node is removed. High modularity and scale-freeness were found in the directed reaction-centric graphs and betweenness centrality tended to belong to densely connected modules. Cascade number and bridging centrality identified cascade subnetworks controlling local information flow and irreplaceable bridging nodes between functional modules, respectively. Reactions highly ranked with bridging centrality and cascade number tended to be essential, compared to reactions that other central metrics detected.

CONCLUSIONS

We demonstrate that cascade number and bridging centrality are useful to identify key reactions controlling local information flow in directed reaction-centric graphs of microbial metabolic networks. Knowledge about the local flow connectivity and connections between local modules will contribute to understand how metabolic pathways are assembled.

Network Medicine in the Age of Biomedical Big Data

Network medicine is an emerging area of research dealing with molecular and genetic interactions, network biomarkers of disease, and therapeutic target discovery. Large-scale biomedical data generation offers a unique opportunity to assess the effect and impact of cellular heterogeneity and environmental perturbations on the observed phenotype. Marrying the two, network medicine with biomedical data provides a framework to build meaningful models and extract impactful results at a network level. In this review, we survey existing network types and biomedical data sources. More importantly, we delve into ways in which the network medicine approach, aided by phenotype-specific biomedical data, can be gainfully applied. We provide three paradigms, mainly dealing with three major biological network archetypes: protein-protein interaction, expression-based, and gene regulatory networks. For each of these paradigms, we discuss a broad overview of philosophies under which various network methods work. We also provide a few examples in each paradigm as a test case of its successful application. Finally, we delineate several opportunities and challenges in the field of network medicine. We hope this review provides a lexicon for researchers from biological sciences and network theory to come on the same page to work on research areas that require interdisciplinary expertise. Taken together, the understanding gained from combining biomedical data with networks can be useful for characterizing disease etiologies and identifying therapeutic targets, which, in turn, will lead to better preventive medicine with translational impact on personalized healthcare.

PNME - A gene-gene parallel network module extraction method

In the domain of gene-gene network analysis, construction of co-expression networks and extraction of network modules have opened up enormous possibilities for exploring the role of genes in biological processes. Through such analysis, one can extract interesting behaviour of genes and would help in the discovery of genes participating in a common biological process. However, such network analysis methods in sequential processing mode often have been found time-consuming even for a moderately sized dataset.

It is observed that most existing network construction techniques are capable of handling only positive correlations in gene-expression data whereas biologically-significant genes exhibit both positive and negative correlations. To address these problems, we propose a faster method for construction and analysis of gene-gene network and extraction of modules using a similarity measure which can identify both negatively and positively correlated co-expressed patterns. Our method utilizes General-purpose computing on graphics processing units (GPGPU) to provide fast, efficient and parallel extraction of biologically relevant network modules to support biomarker identification for breast cancer. The modules extracted are validated using p-value and q-value for both metastasis and non-metastasis stages of breast cancer. PNME has been found capable of identifying interesting biomarkers for this critical disease. We identified six genes with the interesting behaviours which have been found to cause breast cancer in homo-sapiens.

Comparison of gene co-networks analysis provide a systems view of rice (Oryza sativa L.) response to Tilletia horrida infection

The biotrophic soil-borne fungus Tilletia horrida causes rice kernel smut, an important disease affecting the production of rice male sterile lines in most hybrid rice growing regions of the world. There are no successful ways of controlling this disease and there has been little study of mechanisms of resistance to T. horrida. Based on transcriptional data of different infection time points, we found 23, 782 and 23, 718 differentially expressed genes (fragments per kilobase of transcript sequence per million, FPKM >1) in Jiangcheng 3A (resistant to T. horrida) and 9311A (susceptible to T. horrida), respectively. In order to illuminate the differential responses of the two rice male sterile lines to T. horrida, we identified gene co-expression modules using the method of weighted gene co-expression network analysis (WGCNA) and compared the different biological functions of gene co-expression networks in key modules at different infection time points. The results indicated that gene co-expression networks in the two rice genotypes were different and that genes contained in some modules of the two groups may play important roles in resistance to T. horrida, such as DTH8 and OsHop/Sti1a. Furthermore, these results provide a global view of the responses of two different phenotypes to T. horrida, and assist our understanding of the regulation of expression changes after T. horrida infection.

A Comprehensive Survey of Immune Cytolytic Activity-Associated Gene Co-Expression Networks across 17 Tumor and Normal Tissue Types

Cytolytic immune activity in solid tissue can be quantified by transcript levels of two genes, GZMA and PRF1, which is named the CYT score. A previous study has investigated the molecular and genetic properties of tumors associated CYT, but a systematic exploration of how co-expression networks across different tumors are shaped by anti-tumor immunity is lacking. Here, we examined the connectivity and biological themes of CYT-associated modules in gene co-expression networks of 14 tumor and 3 matched normal tissues constructed from the RNA-Seq data of the "The Cancer Genome Atlas" project. We first found that tumors networks have more diverse CYT-correlated modules than normal networks. We next identified and investigated tissue-specific CYT-associated modules across 14 tumor types. Finally, a common CYT-associated network across 14 tumor types was constructed. Two common modules have mixed signs of correlation with CYT in different tumors. Given the tumors and normal tissues surveyed, our study presents a systematic view of the regulation of cytolytic immune activity across multiple tumor tissues.

Preservation affinity in consensus modules among stages of HIV-1 progression

Background

Analysis of gene expression data provides valuable insights into disease mechanism. Investigating relationship among co-expression modules of different stages is a meaningful tool to understand the way in which a disease progresses. Identifying topological preservation of modular structure also contributes to that understanding.

Methods

HIV-1 disease provides a well-documented progression pattern through three stages of infection: acute, chronic and non-progressor. In this article, we have developed a novel framework to describe the relationship among the consensus (or shared) co-expression modules for each pair of HIV-1 infection stages. The consensus modules are identified to assess the preservation of network properties. We have investigated the preservation patterns of co-expression networks during HIV-1 disease progression through an eigengene-based approach.

Results

We discovered that the expression patterns of consensus modules have a strong preservation during the transitions of three infection stages. In particular, it is noticed that between acute and non-progressor stages the preservation is slightly more than the other pair of stages. Moreover, we have constructed eigengene networks for the identified consensus modules and observed the preservation structure among them. Some consensus modules are marked as preserved in two pairs of stages and are analyzed further to form a higher order meta-network consisting of a group of preserved modules. Additionally, we observed that module membership (MM) values of genes within a module are consistent with the preservation characteristics. The MM values of genes within a pair of preserved modules show strong correlation patterns across two infection stages.

Conclusions

We have performed an extensive analysis to discover preservation pattern of co-expression network constructed from microarray gene expression data of three different HIV-1 progression stages. The preservation pattern is investigated through identification of consensus modules in each pair of infection stages. It is observed that the preservation of the expression pattern of consensus modules remains more prominent during the transition of infection from acute stage to non-progressor stage. Additionally, we observed that the module membership values of genes are coherent with preserved modules across the HIV-1 progression stages.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-017-1590-3) contains supplementary material, which is available to authorized users.

FocusHeuristics - expression-data-driven network optimization and disease gene prediction

To identify genes contributing to disease phenotypes remains a challenge for bioinformatics. Static knowledge on biological networks is often combined with the dynamics observed in gene expression levels over disease development, to find markers for diagnostics and therapy, and also putative disease-modulatory drug targets and drugs. The basis of current methods ranges from a focus on expression-levels (Limma) to concentrating on network characteristics (PageRank, HITS/Authority Score), and both (DeMAND, Local Radiality). We present an integrative approach (the FocusHeuristics) that is thoroughly evaluated based on public expression data and molecular disease characteristics provided by DisGeNet. The FocusHeuristics combines three scores, i.e. the log fold change and another two, based on the sum and difference of log fold changes of genes/proteins linked in a network. A gene is kept when one of the scores to which it contributes is above a threshold. Our FocusHeuristics is both, a predictor for gene-disease-association and a bioinformatics method to reduce biological networks to their disease-relevant parts, by highlighting the dynamics observed in expression data. The FocusHeuristics is slightly, but significantly better than other methods by its more successful identification of disease-associated genes measured by AUC, and it delivers mechanistic explanations for its choice of genes.

Integrative Variation Analysis Reveals that a Complex Genotype May Specify Phenotype in Siblings with Syndromic Autism Spectrum Disorder

It has been proposed that copy number variations (CNVs) are associated with increased risk of autism spectrum disorder (ASD) and, in conjunction with other genetic changes, contribute to the heterogeneity of ASD phenotypes. Array comparative genomic hybridization (aCGH) and exome sequencing, together with systems genetics and network analyses, are being used as tools for the study of complex disorders of unknown etiology, especially those characterized by significant genetic and phenotypic heterogeneity. Therefore, to characterize the complex genotype-phenotype relationship, we performed aCGH and sequenced the exomes of two affected siblings with ASD symptoms, dysmorphic features, and intellectual disability, searching for de novo CNVs, as well as for de novo and rare inherited point variations—single nucleotide variants (SNVs) or small insertions and deletions (indels)—with probable functional impacts. With aCGH, we identified, in both siblings, a duplication in the 4p16.3 region and a deletion at 8p23.3, inherited by a paternal balanced translocation, t(4, 8) (p16; p23). Exome variant analysis found a total of 316 variants, of which 102 were shared by both siblings, 128 were in the male sibling exome data, and 86 were in the female exome data. Our integrative network analysis showed that the siblings’ shared translocation could explain their similar syndromic phenotype, including overgrowth, macrocephaly, and intellectual disability. However, exome data aggregate genes to those already connected from their translocation, which are important to the robustness of the network and contribute to the understanding of the broader spectrum of psychiatric symptoms. This study shows the importance of using an integrative approach to explore genotype-phenotype variability.

A Systems Biology Comparison of Ovarian Cancers Implicates Putative Somatic Driver Mutations through Protein-Protein Interaction Models

Ovarian carcinomas can be aggressive with a high mortality rate (e.g., high-grade serous ovarian carcinomas, or HGSOCs), or indolent with much better long-term outcomes (e.g., low-malignant-potential, or LMP, serous ovarian carcinomas). By comparing LMP and HGSOC tumors, we can gain insight into the mechanisms underlying malignant progression in ovarian cancer. However, previous studies of the two subtypes have been focused on gene expression analysis. Here, we applied a systems biology approach, integrating gene expression profiles derived from two independent data sets containing both LMP and HGSOC tumors with protein-protein interaction data. Genes and related networks implicated by both data sets involved both known and novel disease mechanisms and highlighted the different roles of BRCA1 and CREBBP in the two tumor types. In addition, the incorporation of somatic mutation data revealed that amplification of PAK4 is associated with poor survival in patients with HGSOC. Thus, perturbations in protein interaction networks demonstrate differential trafficking of network information between malignant and benign ovarian cancers. The novel network-based molecular signatures identified here may be used to identify new targets for intervention and to improve the treatment of invasive ovarian cancer as well as early diagnosis.

Network biology concepts in complex disease comorbidities

The co-occurrence of diseases can inform the underlying network biology of shared and multifunctional genes and pathways. In addition, comorbidities help to elucidate the effects of external exposures, such as diet, lifestyle and patient care. With worldwide health transaction data now often being collected electronically, disease co-occurrences are starting to be quantitatively characterized. Linking network dynamics to the real-life, non-ideal patient in whom diseases co-occur and interact provides a valuable basis for generating hypotheses on molecular disease mechanisms, and provides knowledge that can facilitate drug repurposing and the development of targeted therapeutic strategies.

The Somatic Nature of Cancer Allows It to Affect Highly Constrained Genes

Cancer is special among genetic disorders in two major ways: first, cancer is a disease of the most basic of cellular functions, such as cell proliferation, differentiation, and the maintenance of genomic integrity. Second, in contrast to most genetic disorders that are mediated by germline (hereditary) mutations, cancer is largely a somatic disease. Here we show that these two traits are not detached and that it is the somatic nature of cancer that allows it to affect the most basic of cellular functions. We begin by demonstrating that cancer genes are both more functionally central (as measured by their patterns of expression and protein interaction) and more evolutionarily constrained than non-cancer genetic disease genes. We then compare genes that are only modified somatically in cancer (hereinafter referred to as “somatic cancer genes”) to those that can also be modified in a hereditary manner, contributing to cancer development (hereinafter referred to as “hereditary cancer genes”). We show that both somatic and hereditary cancer genes are much more functionally central than genes contributing to non-cancer genetic disorders. At the same time, hereditary cancer genes are only as constrained as non-cancer hereditary disease genes, while somatic cancer genes tend to be much more constrained in evolution. Thus, it appears that it is the somatic nature of cancer that allows it to modify the most constrained genes and, therefore, affect the most basic of cellular functions.

Uncovering disease mechanisms through network biology in the era of Next Generation Sequencing

Characterizing the behavior of disease genes in the context of biological networks has the potential to shed light on disease mechanisms, and to reveal both new candidate disease genes and therapeutic targets. Previous studies addressing the network properties of disease genes have produced contradictory results. Here we have explored the causes of these discrepancies and assessed the relationship between the network roles of disease genes and their tolerance to deleterious germline variants in human populations leveraging on: the abundance of interactome resources, a comprehensive catalog of disease genes and exome variation data. We found that the most salient network features of disease genes are driven by cancer genes and that genes related to different types of diseases play network roles whose centrality is inversely correlated to their tolerance to likely deleterious germline mutations. This proved to be a multiscale signature, including global, mesoscopic and local network centrality features. Cancer driver genes, the most sensitive to deleterious variants, occupy the most central positions, followed by dominant disease genes and then by recessive disease genes, which are tolerant to variants and isolated within their network modules.

An integrative approach to investigate the respective roles of single-nucleotide variants and copy-number variants in Attention-Deficit/Hyperactivity Disorder

Many studies have attempted to investigate the genetic susceptibility of Attention-Deficit/Hyperactivity Disorder (ADHD), but without much success. The present study aimed to analyze both single-nucleotide and copy-number variants contributing to the genetic architecture of ADHD. We generated exome data from 30 Brazilian trios with sporadic ADHD. We also analyzed a Brazilian sample of 503 children/adolescent controls from a High Risk Cohort Study for the Development of Childhood Psychiatric Disorders, and also previously published results of five CNV studies and one GWAS meta-analysis of ADHD involving children/adolescents. The results from the Brazilian trios showed that cases with de novo SNVs tend not to have de novo CNVs and vice-versa. Although the sample size is small, we could also see that various comorbidities are more frequent in cases with only inherited variants. Moreover, using only genes expressed in brain, we constructed two "in silico" protein-protein interaction networks, one with genes from any analysis, and other with genes with hits in two analyses. Topological and functional analyses of genes in this network uncovered genes related to synapse, cell adhesion, glutamatergic and serotoninergic pathways, both confirming findings of previous studies and capturing new genes and genetic variants in these pathways.

Whole-exome sequencing in obsessive-compulsive disorder identifies rare mutations in immunological and neurodevelopmental pathways

Studies of rare genetic variation have identified molecular pathways conferring risk for developmental neuropsychiatric disorders. To date, no published whole-exome sequencing studies have been reported in obsessive-compulsive disorder (OCD). We sequenced all the genome coding regions in 20 sporadic OCD cases and their unaffected parents to identify rare de novo (DN) single-nucleotide variants (SNVs). The primary aim of this pilot study was to determine whether DN variation contributes to OCD risk. To this aim, we evaluated whether there is an elevated rate of DN mutations in OCD, which would justify this approach toward gene discovery in larger studies of the disorder. Furthermore, to explore functional molecular correlations among genes with nonsynonymous DN SNVs in OCD probands, a protein-protein interaction (PPI) network was generated based on databases of direct molecular interactions. We applied Degree-Aware Disease Gene Prioritization (DADA) to rank the PPI network genes based on their relatedness to a set of OCD candidate genes from two OCD genome-wide association studies (Stewart et al., 2013; Mattheisen et al., 2014). In addition, we performed a pathway analysis with genes from the PPI network. The rate of DN SNVs in OCD was 2.51 × 10(-8) per base per generation, significantly higher than a previous estimated rate in unaffected subjects using the same sequencing platform and analytic pipeline. Several genes harboring DN SNVs in OCD were highly interconnected in the PPI network and ranked high in the DADA analysis. Nearly all the DN SNVs in this study are in genes expressed in the human brain, and a pathway analysis revealed enrichment in immunological and central nervous system functioning and development. The results of this pilot study indicate that further investigation of DN variation in larger OCD cohorts is warranted to identify specific risk genes and to confirm our preliminary finding with regard to PPI network enrichment for particular biological pathways and functions.

Integration of structural dynamics and molecular evolution via protein interaction networks: a new era in genomic medicine

Sequencing technologies are revealing many new non-synonymous single nucleotide variants (nsSNVs) in each personal exome. To assess their functional impacts, comparative genomics is frequently employed to predict if they are benign or not. However, evolutionary analysis alone is insufficient, because it misdiagnoses many disease-associated nsSNVs, such as those at positions involved in protein interfaces, and because evolutionary predictions do not provide mechanistic insights into functional change or loss. Structural analyses can aid in overcoming both of these problems by incorporating conformational dynamics and allostery in nSNV diagnosis. Finally, protein-protein interaction networks using systems-level methodologies shed light onto disease etiology and pathogenesis. Bridging these network approaches with structurally resolved protein interactions and dynamics will advance genomic medicine.

Gene co-expression network analysis reveals common system-level properties of prognostic genes across cancer types

Prognostic genes are key molecules informative for cancer prognosis and treatment. Previous studies have focused on the properties of individual prognostic genes, but have lacked a global view of their system-level properties. Here we examined their properties in gene co-expression networks for four cancer types using data from The Cancer Genome Atlas. We found that prognostic mRNA genes tend not to be hub genes (genes with an extremely high connectivity), and this pattern is unique to the corresponding cancer-type specific network. In contrast, the prognostic genes are enriched in modules (., a group of highly interconnected genes), especially in module genes conserved across different cancer co-expression networks. The target genes of prognostic miRNA genes show similar patterns. We identified the modules enriched in various prognostic genes, some of which show cross-tumor conservation. Given the cancer types surveyed, our study presents a view of emergent properties of prognostic genes.

Molecular Architecture of Spinal Cord Injury Protein Interaction Network

Spinal cord injury (SCI) is associated with complex pathophysiological processes that follow the primary traumatic event and determine the extent of secondary damage and functional recovery. Numerous reports have used global and hypothesis-driven approaches to identify protein changes that contribute to the overall pathology of SCI in an effort to identify potential therapeutic interventions. In this study, we use a semi-automatic annotation approach to detect terms referring to genes or proteins dysregulated in the SCI literature and develop a curated SCI interactome. Network analysis of the SCI interactome revealed the presence of a rich-club organization corresponding to a “powerhouse” of highly interacting hub-proteins. Studying the modular organization of the network have shown that rich-club proteins cluster into modules that are specifically enriched for biological processes that fall under the categories of cell death, inflammation, injury recognition and systems development. Pathway analysis of the interactome and the rich-club revealed high similarity indicating the role of the rich-club proteins as hubs of the most prominent pathways in disease pathophysiology and illustrating the centrality of pro-and anti-survival signal competition in the pathology of SCI. In addition, evaluation of centrality measures of single nodes within the rich-club have revealed that neuronal growth factor (NGF), caspase 3, and H-Ras are the most central nodes and potentially an interesting targets for therapy. Our integrative approach uncovers the molecular architecture of SCI interactome, and provide an essential resource for evaluating significant therapeutic candidates.

Community structure analysis of transcriptional networks reveals distinct molecular pathways for early- and late-onset temporal lobe epilepsy with childhood febrile seizures

Age at epilepsy onset has a broad impact on brain plasticity and epilepsy pathomechanisms. Prolonged febrile seizures in early childhood (FS) constitute an initial precipitating insult (IPI) commonly associated with mesial temporal lobe epilepsy (MTLE). FS-MTLE patients may have early disease onset, i.e. just after the IPI, in early childhood, or late-onset, ranging from mid-adolescence to early adult life. The mechanisms governing early (E) or late (L) disease onset are largely unknown. In order to unveil the molecular pathways underlying E and L subtypes of FS-MTLE we investigated global gene expression in hippocampal CA3 explants of FS-MTLE patients submitted to hippocampectomy. Gene coexpression networks (GCNs) were obtained for the E and L patient groups. A network-based approach for GCN analysis was employed allowing: i) the visualization and analysis of differentially expressed (DE) and complete (CO) - all valid GO annotated transcripts - GCNs for the E and L groups; ii) the study of interactions between all the system's constituents based on community detection and coarse-grained community structure methods. We found that the E-DE communities with strongest connection weights harbor highly connected genes mainly related to neural excitability and febrile seizures, whereas in L-DE communities these genes are not only involved in network excitability but also playing roles in other epilepsy-related processes. Inversely, in E-CO the strongly connected communities are related to compensatory pathways (seizure inhibition, neuronal survival and responses to stress conditions) while in L-CO these communities harbor several genes related to pro-epileptic effects, seizure-related mechanisms and vulnerability to epilepsy. These results fit the concept, based on fMRI and behavioral studies, that early onset epilepsies, although impacting more severely the hippocampus, are associated to compensatory mechanisms, while in late MTLE development the brain is less able to generate adaptive mechanisms, what has implications for epilepsy management and drug discovery.

Innate immune response is differentially dysregulated between bipolar disease and schizophrenia

Schizophrenia (SZ) and bipolar disorder (BD) are severe psychiatric conditions with a neurodevelopmental component. Genetic findings indicate the existence of an overlap in genetic susceptibility across the disorders. Also, image studies provide evidence for a shared neurobiological basis, contributing to a dimensional diagnostic approach. This study aimed to identify the molecular mechanisms that differentiate SZ and BD patients from health controls but also that distinguish both from health individuals. Comparison of gene expression profiling in post-mortem brains of both disorders and health controls (30 cases), followed by a further comparison between 29 BD and 29 SZ revealed 28 differentially expressed genes. These genes were used in co-expression analysesthat revealed the pairs CCR1/SERPINA1, CCR5/HCST, C1QA/CD68, CCR5/S100A11 and SERPINA1/TLR1 as presenting the most significant difference in co-expression between SZ and BD. Next, a protein-protein interaction (PPI) network using the 28 differentially expressed genes as seeds revealed CASP4, TYROBP, CCR1, SERPINA1, CCR5 and C1QA as having a central role in the diseases manifestation. Both co-expression and network topological analyses pointed to genes related to microglia functions. Based on this data, we suggest that differences between SZ and BP are due to genes involved with response to stimulus, defense response, immune system process and response to stress biological processes, all having a role in the communication of environmental factors to the cells and associated to microglia.

Gene age predicts the strength of purifying selection acting on gene expression variation in humans

Gene expression levels can be subject to selection. We hypothesized that the age of gene origin is associated with expression constraints, given that it affects the level of gene integration into the functional cellular environment. By studying the genetic variation affecting gene expression levels (cis expression quantitative trait loci [cis-eQTLs]) and protein levels (cis protein QTLs [cis-pQTLs]), we determined that young, primate-specific genes are enriched in cis-eQTLs and cis-pQTLs. Compared to cis-eQTLs of old genes originating before the zebrafish divergence, cis-eQTLs of young genes have a higher effect size, are located closer to the transcription start site, are more significant, and tend to influence genes in multiple tissues and populations. These results suggest that the expression constraint of each gene increases throughout its lifespan. We also detected a positive correlation between expression constraints (approximated by cis-eQTL properties) and coding constraints (approximated by Ka/Ks) and observed that this correlation might be driven by gene age. To uncover factors associated with the increase in gene-age-related expression constraints, we demonstrated that gene connectivity, gene involvement in complex regulatory networks, gene haploinsufficiency, and the strength of posttranscriptional regulation increase with gene age. We also observed an increase in heritability of gene expression levels with age, implying a reduction of the environmental component. In summary, we show that gene age shapes key gene properties during evolution and is therefore an important component of genome function.

Complex network-driven view of genomic mechanisms underlying Parkinson's disease: analyses in dorsal motor vagal nucleus, locus coeruleus, and substantia nigra

Parkinson's disease (PD)—classically characterized by severe loss of dopaminergic neurons in the substantia nigra pars compacta—has a caudal-rostral progression, beginning in the dorsal motor vagal nucleus and, in a less extent, in the olfactory system, progressing to the midbrain and eventually to the basal forebrain and the neocortex. About 90% of the cases are idiopathic. To study the molecular mechanisms involved in idiopathic PD we conducted a comparative study of transcriptional interaction networks in the dorsal motor vagal nucleus (VA), locus coeruleus (LC), and substantia nigra (SN) of idiopathic PD in Braak stages 4-5 (PD) and disease-free controls (CT) using postmortem samples. Gene coexpression networks (GCNs) for each brain region (patients and controls) were obtained to identify highly connected relevant genes (hubs) and densely interconnected gene sets (modules). GCN analyses showed differences in topology and module composition between CT and PD networks for each anatomic region. In CT networks, VA, LC, and SN hub modules are predominantly associated with neuroprotection and homeostasis in the ageing brain, whereas in the patient's group, for the three brain regions, hub modules are mostly related to stress response and neuron survival/degeneration mechanisms.

Predicting disease-related proteins based on clique backbone in protein-protein interaction network

Network biology integrates different kinds of data, including physical or functional networks and disease gene sets, to interpret human disease. A clique (maximal complete subgraph) in a protein-protein interaction network is a topological module and possesses inherently biological significance. A disease-related clique possibly associates with complex diseases. Fully identifying disease components in a clique is conductive to uncovering disease mechanisms. This paper proposes an approach of predicting disease proteins based on cliques in a protein-protein interaction network. To tolerate false positive and negative interactions in protein networks, extending cliques and scoring predicted disease proteins with gene ontology terms are introduced to the clique-based method. Precisions of predicted disease proteins are verified by disease phenotypes and steadily keep to more than 95%. The predicted disease proteins associated with cliques can partly complement mapping between genotype and phenotype, and provide clues for understanding the pathogenesis of serious diseases.

Nonparametric network design and analysis of disease genes in oral cancer progression

Biological networks in living organisms can be seen as the ultimate means of understanding the underlying mechanisms in complex diseases, such as oral cancer. During the last decade, many algorithms based on high-throughput genomic data have been developed to unravel the complexity of gene network construction and their progression in time. However, the small size of samples compared to the number of observed genes makes the inference of the network structure quite challenging. In this study, we propose a framework for constructing and analyzing gene networks from sparse experimental temporal data and investigate its potential in oral cancer. We use two network models based on partial correlations and kernel density estimation, in order to capture the genetic interactions. Using this network construction framework on real clinical data of the tissue and blood at different time stages, we identified common disease-related structures that may decipher the association between disease state and biological processes in oral cancer. Our study emphasizes an altered MET (hepatocyte growth factor receptor) network during oral cancer progression. In addition, we demonstrate that the functional changes of gene interactions during oral cancer progression might be particularly useful for patient categorization at the time of diagnosis and/or at follow-up periods.

Identifying disease associated genes by network propagation

Background

Genome-wide association studies have identified many individual genes associated with complex traits. However, pathway and network information have not been fully exploited in searches for genetic determinants, and including this information may increase our understanding of the underlying biology of common diseases.

Results

In this study, we propose a framework to address this problem in a principled way, with the underlying hypothesis that complex disease operates through multiple connected genes. Associations inferred from GWAS are translated into prior scores for vertices in a protein-protein interaction network, and these scores are propagated through the network. Permutation is used to select genes that are guilty-by-association and thus consistently obtain high scores after network propagation. We apply the approach to data of Crohn's disease and call candidate genes that have been reported by other independent GWAS, but not in the analysed data set. A prediction model based on these candidate genes show good predictive power as measured by Area Under the Receiver Operating Curve (AUC) in 10 fold cross-validations.

Conclusions

Our network propagation method applied to a genome-wide association study increases association findings over other approaches.

Early gene expression changes in skeletal muscle from SOD1(G93A) amyotrophic lateral sclerosis animal model

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by loss of motor neurons. Familial ALS is strongly associated to dominant mutations in the gene for Cu/Zn superoxide dismutase (SOD1). Recent evidences point to skeletal muscle as a primary target in the ALS mouse model. Wnt/PI3 K signaling pathways and epithelial-mesenchymal transition (EMT) have important roles in maintenance and repair of skeletal muscle. Wnt/PI3 K pathways and EMT gene expression profile were investigated in gastrocnemius muscle from SOD1(G93A) mouse model and age-paired wild-type control in the presymptomatic ages of 40 and 80 days aiming the early neuromuscular abnormalities that precede motor neuron death in ALS. A customized cDNA microarray platform containing 326 genes of Wnt/PI3 K and EMT was used and results revealed eight up-regulated (Loxl2, Pik4ca, Fzd9, Cul1, Ctnnd1, Snf1lk, Prkx, Dner) and nine down-regulated (Pik3c2a, Ripk4, Id2, C1qdc1, Eif2ak2, Rac3, Cds1, Inppl1, Tbl1x) genes at 40 days, and also one up-regulated (Pik3ca) and five down-regulated (Cd44, Eef2 k, Fzd2, Crebbp, Piki3r1) genes at 80 days. Also, protein-protein interaction networks grown from the differentially expressed genes of 40 and 80 days old mice have identified Grb2 and Src genes in both presymptomatic ages, thus playing a potential central role in the disease mechanisms. mRNA and protein levels for Grb2 and Src were found to be increased in 80 days old ALS mice. Gene expression changes in the skeletal muscle of transgenic ALS mice at presymptomatic periods of disease gave further evidence of early neuromuscular abnormalities that precede motor neuron death. The results were discussed in terms of initial triggering for neuronal degeneration and muscle adaptation to keep function before the onset of symptoms.

The genomic and functional characteristics of disease genes

Increasing evidence indicates that genes containing disease causal variation have distinct functional and genomic properties. The importance of understanding these properties is highlighted by efforts to filter lists of variants from next-generation sequencing studies, where the number of potentially deleterious variants, which are in fact unrelated to disease, may be large. Available evidence indicates that the majority of disease genes are 'non-essential' and their products occupy functionally peripheral positions in protein networks. They tend to be intermediate between genes that have core biological functions, particularly low mutation rates and low haplotype diversity, and genes for which high haplotype diversity and high mutation rates are advantageous (such as those involved in sensory perception and some immune system functions). Evidence presented here supports these conclusions through analysis of integrated data sets incorporating the latest mutational profiles, linkage disequilibrium structure and other genomic properties of individual genes. The analysis highlights the contrasting functions of genes predicted as least and most likely to contain disease variation and provides a basis for filtering gene variant lists to exclude the least plausible disease candidates.

Complex network analysis of CA3 transcriptome reveals pathogenic and compensatory pathways in refractory temporal lobe epilepsy

We previously described - studying transcriptional signatures of hippocampal CA3 explants - that febrile (FS) and afebrile (NFS) forms of refractory mesial temporal lobe epilepsy constitute two distinct genomic phenotypes. That network analysis was based on a limited number (hundreds) of differentially expressed genes (DE networks) among a large set of valid transcripts (close to two tens of thousands). Here we developed a methodology for complex network visualization (3D) and analysis that allows the categorization of network nodes according to distinct hierarchical levels of gene-gene connections (node degree) and of interconnection between node neighbors (concentric node degree). Hubs are highly connected nodes, VIPs have low node degree but connect only with hubs, and high-hubs have VIP status and high overall number of connections. Studying the whole set of CA3 valid transcripts we: i) obtained complete transcriptional networks (CO) for FS and NFS phenotypic groups; ii) examined how CO and DE networks are related; iii) characterized genomic and molecular mechanisms underlying FS and NFS phenotypes, identifying potential novel targets for therapeutic interventions. We found that: i) DE hubs and VIPs are evenly distributed inside the CO networks; ii) most DE hubs and VIPs are related to synaptic transmission and neuronal excitability whereas most CO hubs, VIPs and high hubs are related to neuronal differentiation, homeostasis and neuroprotection, indicating compensatory mechanisms. Complex network visualization and analysis is a useful tool for systems biology approaches to multifactorial diseases. Network centrality observed for hubs, VIPs and high hubs of CO networks, is consistent with the network disease model, where a group of nodes whose perturbation leads to a disease phenotype occupies a central position in the network. Conceivably, the chance for exerting therapeutic effects through the modulation of particular genes will be higher if these genes are highly interconnected in transcriptional networks.

SBEToolbox: A Matlab Toolbox for Biological Network Analysis

We present SBEToolbox (Systems Biology and Evolution Toolbox), an open-source Matlab toolbox for biological network analysis. It takes a network file as input, calculates a variety of centralities and topological metrics, clusters nodes into modules, and displays the network using different graph layout algorithms. Straightforward implementation and the inclusion of high-level functions allow the functionality to be easily extended or tailored through developing custom plugins. SBEGUI, a menu-driven graphical user interface (GUI) of SBEToolbox, enables easy access to various network and graph algorithms for programmers and non-programmers alike. All source code and sample data are freely available at https://github.com/biocoder/SBEToolbox/releases.

How old is my gene?

Gene functions, interactions, disease associations, and ecological distributions are all correlated with gene age. However, it is challenging to estimate the intricate series of evolutionary events leading to a modern-day gene and then to reduce this history to a single age estimate. Focusing on eukaryotic gene families, we introduce a framework that can be used to compare current strategies for quantifying gene age, discuss key differences between these methods, and highlight several common problems. We argue that genes with complex evolutionary histories do not have a single well-defined age. As a result, care must be taken to articulate the goals and assumptions of any analysis that uses gene age estimates. Recent algorithmic advances offer the promise of gene age estimates that are fast, accurate, and consistent across gene families. This will enable a shift to integrated genome-wide analyses of all events in gene evolutionary histories in the near future.

Human monogenic disease genes have frequently functionally redundant paralogs

Mendelian disorders are often caused by mutations in genes that are not lethal but induce functional distortions leading to diseases. Here we study the extent of gene duplicates that might compensate genes causing monogenic diseases. We provide evidence for pervasive functional redundancy of human monogenic disease genes (MDs) by duplicates by manifesting 1) genes involved in human genetic disorders are enriched in duplicates and 2) duplicated disease genes tend to have higher functional similarities with their closest paralogs in contrast to duplicated non-disease genes of similar age. We propose that functional compensation by duplication of genes masks the phenotypic effects of deleterious mutations and reduces the probability of purging the defective genes from the human population; this functional compensation could be further enhanced by higher purification selection between disease genes and their duplicates as well as their orthologous counterpart compared to non-disease genes. However, due to the intrinsic expression stochasticity among individuals, the deleterious mutations could still be present as genetic diseases in some subpopulations where the duplicate copies are expressed at low abundances. Consequently the defective genes are linked to genetic disorders while they continue propagating within the population. Our results provide insight into the molecular basis underlying the spreading of duplicated disease genes.

Global analysis of the human pathophenotypic similarity gene network merges disease module components

The molecular complexity of genetic diseases requires novel approaches to break it down into coherent biological modules. For this purpose, many disease network models have been created and analyzed. We highlight two of them, "the human diseases networks" (HDN) and "the orphan disease networks" (ODN). However, in these models, each single node represents one disease or an ambiguous group of diseases. In these cases, the notion of diseases as unique entities reduces the usefulness of network-based methods. We hypothesize that using the clinical features (pathophenotypes) to define pathophenotypic connections between disease-causing genes improve our understanding of the molecular events originated by genetic disturbances. For this, we have built a pathophenotypic similarity gene network (PSGN) and compared it with the unipartite projections (based on gene-to-gene edges) similar to those used in previous network models (HDN and ODN). Unlike these disease network models, the PSGN uses semantic similarities. This pathophenotypic similarity has been calculated by comparing pathophenotypic annotations of genes (human abnormalities of HPO terms) in the "Human Phenotype Ontology". The resulting network contains 1075 genes (nodes) and 26197 significant pathophenotypic similarities (edges). A global analysis of this network reveals: unnoticed pairs of genes showing significant pathophenotypic similarity, a biological meaningful re-arrangement of the pathological relationships between genes, correlations of biochemical interactions with higher similarity scores and functional biases in metabolic and essential genes toward the pathophenotypic specificity and the pleiotropy, respectively. Additionally, pathophenotypic similarities and metabolic interactions of genes associated with maple syrup urine disease (MSUD) have been used to merge into a coherent pathological module.Our results indicate that pathophenotypes contribute to identify underlying co-dependencies among disease-causing genes that are useful to describe disease modularity.

Disentangling function from topology to infer the network properties of disease genes

BACKGROUND

The topological features of disease genes within interaction networks are the subject of intense study, as they shed light on common mechanisms of pathology and are useful for uncovering additional disease genes. Computational analyses typically try to uncover whether disease genes exhibit distinct network features, as compared to all genes.

RESULTS

We demonstrate that the functional composition of disease gene sets is an important confounding factor in these types of analyses. We consider five disease sets and show that while they indeed have distinct topological features, they are also enriched in functions that a priori exhibit distinct network properties. To address this, we develop a computational framework to assess the network properties of disease genes based on a sampling algorithm that generates control gene sets that are functionally similar to the disease set. Using our function-constrained sampling approach, we demonstrate that for most of the topological properties studied, disease genes are more similar to sets of genes with similar functional make-up than they are to randomly selected genes; this suggests that these observed differences in topological properties reflect not only the distinguishing network features of disease genes but also their functional composition. Nevertheless, we also highlight many cases where disease genes have distinct topological properties even when accounting for function.

CONCLUSIONS

Our approach is an important first step in extracting the residual topological differences in disease genes when accounting for function, and leads to new insights into the network properties of disease genes.

Differences in transcriptional effects of 1α,25 dihydroxyvitamin D3 on fibroblasts associated to breast carcinomas and from paired normal breast tissues

The effects of 1α,25 dihydroxyvitamin D3 (1,25D) on breast carcinoma associated fibroblasts (CAFs) are still unknown. This study aimed to identify genes whose expression was altered after 1,25D treatment in CAFs and matched adjacent normal mammary associated fibroblasts (NAFs). CAFs and NAFs (from 5 patients) were cultured with or without (control) 1,25D 100 nM. Both CAF and NAF expressed vitamin D receptor (VDR) and 1,25D induction of the genomic pathway was detected through up-regulation of the target gene CYP24A1. Microarray analysis showed that despite presenting 50% of overlapping genes, CAFs and NAFs exhibited distinct transcriptional profiles after 1,25D treatment (FDR<0.05). Functional analysis revealed that in CAFs, genes associated with proliferation (NRG1, WNT5A, PDGFC) were down regulated and those involved in immune modulation (NFKBIA, TREM-1) were up regulated, consistent with anti tumor activities of 1,25D in breast cancer. In NAFs, a distinct subset of genes was induced by 1,25D, involved in anti apoptosis, detoxification, antibacterial defense system and protection against oxidative stress, which may limit carcinogenesis. Co-expression network and interactome analysis of genes commonly regulated by 1,25D in NAFs and CAFs revealed differences in their co-expression values, suggesting that 1,25D effects in NAFs are distinct from those triggered in CAFs.

Biomarker identification for prostate cancer and lymph node metastasis from microarray data and protein interaction network using gene prioritization method

Finding a genetic disease-related gene is not a trivial task. Therefore, computational methods are needed to present clues to the biomedical community to explore genes that are more likely to be related to a specific disease as biomarker. We present biomarker identification problem using gene prioritization method called gene prioritization from microarray data based on shortest paths, extended with structural and biological properties and edge flux using voting scheme (GP-MIDAS-VXEF). The method is based on finding relevant interactions on protein interaction networks, then scoring the genes using shortest paths and topological analysis, integrating the results using a voting scheme and a biological boosting. We applied two experiments, one is prostate primary and normal samples and the other is prostate primary tumor with and without lymph nodes metastasis. We used 137 truly prostate cancer genes as benchmark. In the first experiment, GP-MIDAS-VXEF outperforms all the other state-of-the-art methods in the benchmark by retrieving the truest related genes from the candidate set in the top 50 scores found. We applied the same technique to infer the significant biomarkers in prostate cancer with lymph nodes metastasis which is not established well.

Do cancer proteins really interact strongly in the human protein-protein interaction network?

Protein-protein interaction (PPI) network analysis has been widely applied in the investigation of the mechanisms of diseases, especially cancer. Recent studies revealed that cancer proteins tend to interact more strongly than other categories of proteins, even essential proteins, in the human interactome. However, it remains unclear whether this observation was introduced by the bias towards more cancer studies in humans. Here, we examined this important issue by uniquely comparing network characteristics of cancer proteins with three other sets of proteins in four organisms, three of which (fly, worm, and yeast) whose interactomes are essentially not biased towards cancer or other diseases. We confirmed that cancer proteins had stronger connectivity, shorter distance, and larger betweenness centrality than non-cancer disease proteins, essential proteins, and control proteins. Our statistical evaluation indicated that such observations were overall unlikely attributed to random events. Considering the large size and high quality of the PPI data in the four organisms, the conclusion that cancer proteins interact strongly in the PPI networks is reliable and robust. This conclusion suggests that perturbation of cancer proteins might cause major changes of cellular systems and result in abnormal cell function leading to cancer.

Mining susceptibility gene modules and disease risk genes from SNP data by combining network topological properties with support vector regression

Genome-wide association study is a powerful approach to identify disease risk loci. However, the molecular regulatory mechanisms for most complex diseases are still not well understood. Therefore, further investigating the interplay between genetic factors and biological networks is important for elucidating the molecular mechanisms of complex diseases. Here, we proposed a novel framework to identify susceptibility gene modules and disease risk genes by combining network topological properties with support vector regression from single nucleotide polymorphism (SNP) level. We assigned risk SNPs to genes using the University of California at Santa Cruz (UCSC) genome database, and then mapped these genes to protein-protein interaction (PPI) networks. The gene modules implicated by hub genes were extracted using the PPI networks and the topological property was analyzed for these gene modules. For each gene module, risk feature genes were determined by topological property analysis and support vector regression. As a result, five shared risk feature genes, CD80, EGFR, FN1, GSK3B and TRAF6 were found and proven to be associated with rheumatoid arthritis by previous reports. Our approach showed a good performance in comparison with other approaches and can be used for prioritizing candidate genes associated with complex diseases.