Modern high-throughput experiments provide us with numerous potential associations between genes and diseases. Experimental validation of all the discovered associations, let alone all the possible interactions between them, is time-consuming and expensive. To facilitate the discovery of causative genes, various approaches for prioritization of genes according to their relevance for a given disease have been developed. In this article, we explain the gene prioritization problem and provide an overview of computational tools for gene prioritization. Among about a hundred of published gene prioritization tools, we select and briefly describe 14 most up-to-date and user-friendly. Also, we discuss the advantages and disadvantages of existing tools, challenges of their validation, and the directions for future research.
Gene prioritization problem emerged together with the growth of popularity of genetic linkage analysis. Genetic mapping yielded large loci containing many candidate genes, only a few of which were indeed associated with the investigated phenotype. In order to determine causative gene variants, dozens of candidate genes from those loci had to be examined and it was economically reasonable to start from genes more likely to impact the observed phenotype. Selection of the most promising candidates can be done on the basis of previous knowledge about these genes, e.g. functional associations or similarity with known disease genes. In the early 2000s, aggregation of publicly available sequence data and growth of functional annotation of the human genome enabled automatization of gene prioritization . In 2006 Aerts et al.  prioritized 58 candidate genes from a 2-Mb region of chromosome 22 according to their similarity with known disease genes in ten distinct evidence sources. They predicted YPEL1 as a novel gene involved in atypical DiGeorge syndrome (DGS) and validated this prediction in vivo: knock-down of YPEL1 homolog in Zebrafish embryos led to craniofacial defects and confirmed its role in pharyngeal arch morphogenesis .
With the advent of high throughput methods, genetic linkage analysis was replaced by Genome-wide association studies (GWAS)  allowing cheaper, faster and more precise genetic mapping. However, the necessity of ranking long list of candidate genes according to their relevance to phenotypes had not disappeared. Modern high-throughput experiments, such as genome-wide association studies (GWAS) or differential expression studies generate hundreds or thousands of potential associations, requiring further exploration. Some of these associations may appear by chance or due to systematic biases and therefore may be poorly reproducible. In parallel with the simplification of the candidate gene search, the amount of available information about genes increased. This information includes the data on gene-gene interactions or interactions of genes with other biological entities, their involvement in various biological processes, such as the development of disorders or other phenotypic traits. The emergence of various biological databases and the explosive growth of relevant scientific publications further complicated manual evaluation of candidate genes and stimulated the development of computational methods and tools for gene prioritization. Gene prioritization tools were extensively applied for prediction of genes involved in Mendelian , , ,  and complex diseases , , ,  and other polygenic traits , . In addition to the evaluation of gene relevance for single diseases, gene prioritization was used for the selection of genes potentially responsible for the comorbidity between two complex diseases – asthma and hypertension . Moreover, taking into account the predicted importance of candidate genes, i.e. score assigned in the result of prioritization, improved the results of pathway enrichment analysis , , , enhanced models for drug response , and disease outcome predictions from gene expression profiles .
At the moment, hundreds of research papers on gene prioritization have been published  and about a hundred of them describe computational tools. In the course of development of this field, many reviews , , , , , ,  and benchmark , ,  works have been published. However, either they cover a very small number of tools and are more focused on theoretical aspects  or describe outdated programs no longer supported. Recently, Seyyedrazzagi and Navimipour  published a comprehensive literature review where they selected 19 gene prioritization methods for comparison. However, only 7 of those 19 works published between 2011 and 2015 provided links to either web service or code. We managed to access five of them and successfully run only one tool, ProphNet , which became unavailable during the preparation of this work. Therefore, in contrast with Seyyedrazzagi and Navimipour , we focused on articles describing computational tools and inspected their availability and usability. This review is aimed to give an updated view of existing gene prioritization tools with a specific focus on the most competitive, user-friendly and up-to-date tools. Here, we provide the classification of gene prioritization techniques, develop criteria for selection of the most up-to date and user-friendly tools and briefly describe selected tools.
2 Overview of Gene Prioritization Tools
Gene prioritization task could be formulated as follows: arrange candidate genes in order of their potential to be truly associated with the disease decreasing on the basis of prior knowledge about these genes and the disease. A typical gene prioritization tool is composed of two parts: a collection of evidence sources (i.e. databases of associations between genes, diseases and other biological entities) and a prioritization module (Figure 1). Prioritization module takes two inputs: training data, which is used to define a phenotype of interest and testing data, a set of user-defined candidate genes to prioritize. After that, it extracts information about given genes or terms from evidence sources and calculates a score that reflects “likelihood” of each gene to be responsible for the phenotype. Training data could be represented either by genes, that were previously linked with a phenotype (seed genes). Alternatively to seed genes, some tools (PolySearch2 , PhenoRank , Open Targets  and others , , , , ) require phenotype or disease terms defining relevant gene-disease associations. The second part of the input is a set of candidate genes to prioritize or in some cases, the whole genome , , , , , , . Some tools, such as Génie , Open Targets and Endeavour , can automatically construct a set of candidate genes, belonging to a specific class, e.g. all protein-coding genes , all receptors or enzymes , or use GO or pathway terms to define a functional group of genes , . The output of the program is a list of candidate genes arranged according to calculated scores or p-values. Every gene prioritization tool represents a unique combination of evidence sources, prioritization strategy and input requirements.
2.1 Prioritization Strategies
In previous works, gene prioritization tools have been classified based on the scope of their application (generic or disease-specific) , types of evidence sources used , approaches (filter-based selection or ranking) ,  and method types (network analysis, similarity profiling or text-mining) , . In this review, we propose two classifications of gene prioritization tools: (i) by assumptions they rely on, and (ii) by data repesentay use. It is important to note that the proposed categories are not mutually exclusive, and the purpose of this chapter is to provide the reader with a general idea on the most popular gene prioritization approaches.
The existing gene prioritization approaches rely on two major assumptions. First, genes may be directly associated with a disease, if they are systematically altered in the disease compared to controls (e.g. carry a disease-specific variant). Although various associations may have different strengths and qualities, it is assumed that association, supported by multiple independent studies is more likely to be true. Second, genes can be associated with a disease indirectly, via guilt-by-association principle, assuming that the most probable candidates are somehow linked with genes or other biological entities, that were previously shown to impact the phenotype of interest.
Two types of gene prioritization strategies can be distinguished, depending on the assumption they rely on and, consequently, on the kind of prior knowledge used to solve the gene prioritization problem. Strategies of the first type integrate for each candidate all evidence supporting its association with the query disease and compute the overall score. Such tools require a user to provide keywords or ontology terms specifying the disease and then integrate gene-disease associations of various kinds (subsection 2.2.1). In contrast, strategies of the second type reduce the gene prioritization problem to the task of finding genes closely related to known disease genes and, instead of specifying the disease explicitly, accept a set of seed genes, implicitly defining the disease. These tools integrate and analyze associations between genes (subsection 2.2.2), direct and indirect, and prioritize candidates by their similarity and/or proximity to a set of seeds.
Although the majority of tools follow exclusively one of these two strategies, some tools implement a combination of them. For example, PhenoRank  and Phenolyzer  accept disease keywords, automatically construct a scored list of seed genes, and rank the rest of genes such that genes associated with high-scored seeds also get higher ranks. Another example is NetworkPrioritizer, which retrieves genes associated with a query disease, builds disease-specific network and identifies the most relevant genes based on the network topology .
2.1.2 Data Representation Models and Prioritization Approaches
The structure of evidence sources utilized by a gene prioritization tool can be either relational (Figure 2A), when data sources are represented by a collection of tables, containing an association of a particular kind, or network (Figure 2B), where nodes correspond to genes (or other entities) and edges represent relationships between them. Although these two data representations models are interchangeable, organization of evidence sources is always consistent with the prioritization algorithm. Accordingly, most of the existing approaches can be classified as score aggregation or network analysis methods, or represent their combination , . Since the proposed classification is very general, we discuss the methods of network analysis and score aggregation below.
Network is a natural representation of pairwise entity relationships, widely used to describe similarities or interactions between biological objects. Many independent studies agree that disease-associated proteins tend to cluster on the network of protein-protein interactions , ,  (PPI). This observation became the basis of many gene prioritization approaches assuming that proteins, encoded by unknown disease genes and more tightly connected with known disease proteins than irrelevant proteins. Moreover, disease proteins were shown to possess special network properties, for example, they tend to be more central in the disease-specific networks , . This observation is used in some network-based gene prioritization workflows , , ,  including also those omitting seed genes , . Briefly, these methods retrieve all genes somehow associated with a query disease, build disease-specific interaction network and determine the essentiality of all nodes considering the network topology.
The majority of network-based tools require seed genes and rank candidate according to their proximity in the network toseeds. The distance from a node to a group of nodes in the network may be defined in numerous ways. MaxLink ,  ranks first neighbours of known genes according to a number of direct links to them. In order to reduce hub bias, it takes into account only candidates which have significantly more connection with seed genes than expected by chance. Similarly with MaxLink, DIAMOnD  applies hypergeometric test to detect candidates enreached by seed genes among their first neighbours. In contrast, DIAMOnD ranks genes according to significance of seed overrepresentation among the first neighbours. In every iteration, DIAMOnD includes the most significant candidate into a set of seeds and recalculates p-values for the remaining candidates regarding the updated set of seeds. However, MaxLink considers only first neighbours of seeds and DIAMOnD ignores indirect interactions on every iteration. Gentrepid  ranks candidates conforming to the shortest path distance to a seed node. NetShort method implemented in GUILD framework  down-weights edges connected to genes with a high score when computing shortest path length. The disadvantage of this measure it that not all the pathways are equally informative, e.g. a path going through promiscuous hub nodes may be short but unspecific to the disease mechanism. In order to address this problem, various network propagation methods, modeling information flow over the network have been developed. ToppNet , GeneWanderer , PhenoRank  and many others , , , , , ,  apply random walk-based algorithms , ,  in order to assess relative importance of a node to a group nodes considering the global network topology. Other methods mathematically related  with random walk, modelling diffusion , ,  or electric current flow  through the network have been used successfully in gene prioritization. GeneMANIA  implements Gaussian field label propagation algorithm , which redistributes seed gene scores to their neighbors, minimizing differences between both scores of neighboring genes and initial and assigned scores of seed genes. PRINCE  uses conceptually similar approach to smooths influence of disease genes over the network. It simulates the exchange of flows between genes in the network, where every node produces outcoming flows to neighbors, proportional to its score, and computes a new score summarizing incoming flows. The process starts from disease genes and stops after many iterations. In the result, candidates connected with many disease genes gain higher income flow and thus a higher score. eQED  represents the network as an electric circuit where seeds are current sources, edges are conductances, candidates are drains, and rank candidates by current flowing through them. Köhler et al. , Navlakha and Kingsford , and Shim et al.  have shown that methods considering global network topology demonstrate higher overall performance than methods based only on local network information. At the same time, methods using local network topology, e.g. direct interactions or shortest path distances, rank true top-ranked candidates higher ,  and therefore are more successful for diseases with few associated genes, tightly connected in the network .
Another important feature determining the performance of the network analysis tool is the network type used, its quality and completeness . Some network-based prioritization tools use homogeneous networks modeling only one type of interactions , , , . However, recent studies demonstrate that composite networks, composed of many various kinds of interactions and relationships, outperform any single network, possibly because individual networks contain complementary information , . Therefore, many gene prioritization tools use functional protein interaction networks such as GeneMania , FunCoup , STRING  or integrate several networks of different types , . Moreover, in previous works gene prioritization is performed on heterogeneous networks including multiple types of biological entities , , , , , .
This group includes tools implementing various strategies of aggregating all found pieces of evidence within each data source and then between different data sources into a total score. For example, Polysearch ,  recognizes sentences supporting gene-disease associations, weights them according to their reliability and summarizes weights into the total relevancy score. When relevancy scores computed for all genes, Polysearch standardizes them and uses for prioritization. Similarly, DisGeNET , ,  and Open Targets  integrate data from multiple evidence sources. For each gene, they compute a weighted sum over all individual gene-disease association scores. Each weighting coefficient is determined by the reliability of association and the type of data source it came from. Thus, strong genetic associations discovered in human make a bigger impact into the overall gene score, than less reliable associations inferred from animal models or text mining.
Tools operating with seed genes employ similar idea to summarize gene-gene associations. Initially, they score each candidate by its similarity with seeds, considering each evidence sources independently, and then combine all data source-specific scores into a total score. GPS  follows the most straightforward way to integrate multiple rankings: for each gene, it calculates a simple rank average over seven independent rankings. ToppGene ,  and Endeavour , ,  realize more sophisticated approaches to obtain the overall ranking. They convert data source-specific scores into p-values and apply meta-analysis-based techniques to compute the overall p-value for each gene.
Score aggregation approaches described above have at least two drawbacks. First, these tools favor genes top-ranked in a maximal number of evidence sources. Meanwhile, they may not consider various reliability and potential dependency of evidence sources. Second, tools from this category do not take into account the fact that the impact of independent rankings into the total score may not be additive.
These deficiencies have been partly overcome with the development of machine learning methods. Similar to ToppGene and Endeavour, machine learning-based methods represent genes as n-dimensional feature vectors, use seed genes as positive training exemplars, genes other than seeds or candidates as negative exemplars, and then classify candidates. Machine learning methods such as multiple linear , ,  and logistic ,  regressions, kernel-based approaches , , , neural networks  and others  were successfully applied for gene prioritization. Recent works have demonstrated that machine learning-based methods tend to outperform other score aggregators , , , possibly owing to their ability to capture unknown or non-linear feature relationships and tuning model parameters.
2.2 Evidence Sources
Almost all gene prioritization tools with rare exceptions ,  rely on internal databases integrating a wide variety of information sources. In this paragraph, we discuss types of evidence used for establishing of gene-disease associations and for calculation of gene similarities. Performance of gene prioritization tools strongly depends on variety data sources used , , ,  and on their novelty .
2.2.1 Gene-Disease Associations
Since many of human diseases are proven to have a genetic component, the discovery of genetic variants underlying diseases is one of the major challenges of human genetics. Impact of the individual genetic constitution on the development of the disease may be very different: from highly heritable Mendelian disorders, e.g. sickle-cell anemia (OMIM:#603903) or Duchenne muscular dystrophy (OMIM:#310200) caused by mutations in certain genes, to infectious diseases  caused by an external pathogenic agent. Several thousands of the human diseases are caused by the dysregulation of a single gene: loss  or modification  of its function. These diseases are usually rare and aggregate in families which carry a specific mutation. In addition to medical significance, such monogenic diseases allow connecting phenotypes and genes and thus providing us with clues about gene functions. Therefore, one of the main and widely used sources of knowledge for gene prioritization is OMIM . OMIM provides a constantly updating catalog of hereditary disorders and associated genes. Although OMIM initiated as a database of Mendelian diseases, now in addition to more than 5000 single-gene traits, it coves about 700 complex and 150 non-disease phenotypes. Besides OMIM, several other databases collect information about rare monogenic diseases and disease-causing variants: OrphaNet , DECIPHER , ClinVar , HGMD . In addition to comprehensive catalogs of human monogenic diseases, one can obtain genotype and phenotype information on animal models of human diseases from species-specific databases e.g. Mouse Genome Database  and Rat Genome Database .
Unlike Mendelian diseases, complex diseases cannot be explained by a single mutation with high effect but thought to be the result of interactions between multiple genetic and environmental factors. Genome-wide association studies (GWAS) discover common single nucleotide polymorphisms (SNPs), which allelic states significantly correlate with disease status. NHGRI-EBI GWAS Catalog  provides a curated and regularly updated lists of published GWAS, and contains over 60,000 of SNP-trait associations from 3411 publications. However, effects of such risk variants identified in GWAS are much weaker than effects of Mendelian variants. Moreover, linking a variant with its effect on a certain gene may be a challenging task because, in contrast with Mendelian disease variants, only a small fraction of GWAS hits has an obvious effect on the protein, e.g. missense substitution or frameshift. Most of significant GWAS variants locate in intronic or intergenic regions . Mapping them to the closest gene may not always be correct, and considering regulatory annotations for mapping of silent GWAS variants to genes appears to be a better strategy . Joehanes et al.  have shown, that about half of GWAS variants lay in expression quantitative trait loci (eQTL) – genome regions with markers correlated with expression levels of one or several genes.
Besides DNA sequence variations, many other kinds of biological evidence may be used for the inference of gene-disease associations. Since disease manifestation usually accompanied by various molecular changes, case-control omics studies allow identifying coding and non-coding transcripts , proteins , and other entities such as metabolites  and epigenetic marks  altered in disease samples compared to controls and thus potentially involved into the pathogenesis. Genes and proteins differentially expressed under various biological conditions, including diseases and developmental stages in human and other organisms can be obtained from Expression Atlas , a curated database of expression profiles in human and many other species derived from selected RNA-seq and microarray datasets. Expression Atlas uniformly processes and analyse expression data obtained from multiple sources, from individual Gene Expression Omnibus  datasets to large expression studies, such as GTEx  and Human Protein Atlas , . Also, Expression Atlas provides baseline expression levels in tissues and cell types and allows retrieving tissue- and cell type-specific genes. This information can also give a hint about gene function. For example, geneTIER  assumes that plausible candidate genes are highly expressed in tissues, affected by the disease. However, this assumption is only applicable when affected tissues or cell types are well known, which is not always the case.
Other Ways of Establishing Gene-Disease Associations
In addition to gene-disease associations identified in case-control experiments, some indirect associations can be inferred through a third biological entity, utilizing guilt-by-association principle. The next paragraph discusses indirect gene-disease associations mediated by known diseases genes in deep details. Moreover, indirect gene-disease associations can be mediated by biological entities other than genes: chemicals , tissues , and other diseases and phenotypes. For example, a gene targeted by a drug used to treat the disease, it is likely to participate in the mechanism of the disease. The same may be true for targets of drugs and toxic compounds are known to cause negative effects e.g. adverse reactions, similar to the disease. Such gene-disease associations inferred from curated pairwise associations with chemicals, which can be obtained, for example, from Comparative Toxicogenomic Database  (CTD, http://ctdbase.org/). Similarly, putative gene-disease associations can be established through the second disease, via disease correlation  or symptom similarities . Since co-occurring diseases tend to have more shared genes, than expected by chance , at least some of genes, known to be associated with one disease may be involved into the other disease. The same is true for diseases demonstrating phenotype similarities . Each human disease can be characterized by a specific combination of multiple phenotypes and, in turn, some phenotypic abnormalities can manifest in many diseases. Thus, brachydactyly syndrome (OMIM:#112410), characterized by age-dependent hypertension, shortening of both phalanges and many other abnormalities . At the same time, hypertension is a symptom of many other disorders, e.g. #613677, #614495, #500005, #218030, #602531. In order to annotate diseases with phenotypes and calculate phenotypic similarities between diseases, unified ontologies for diseases  and phenotypes , ,  were developed. Human Phenotype Ontology (HPO) project also provides the results of semi-automatic mapping between phenotypes, disease, and genes . Recent the Monarch Initiative  allows comparing human phenotypes with animal phenotypes  with a known genetic basis.
2.2.2 Associations between Genes
Gene prioritization approaches are based on guilt-by-association principle and assume that the most promising candidates are in some way associated with seed genes. Physical PPI point to potential functional interaction between these proteins and subsequently, to the association between corresponding genes. Physical PPI can be experimentally identified using high-throughput methods, such as yeast two-hybrid assay, affinity purification with mass spectrometry or confirmed in single experiments, e.g. X-ray crystallography. Primary PPI databases obtain data from curation of published literature, e.g. DIP , HPRD , BioGRID , InnateDB  or MatrixDB  or from single large-scale experiments , . Other PPI databases, such as IntAct , MINT , MENTHA , HitPredict , integrate protein interaction data from multiple primary databases and assign interaction reliability scores according to the supporting evidence. In order to facilitate an access to a large number of redundant PPI databases, a standardized query interface PSIQUIC was created.
In addition to direct physical contacts, proteins can also interact indirectly, collectively performing their function. For example, since a protein complex functions as a whole, all its members, including those non-interacting directly, are strongly functionally related. CORUM  and Complex Portal  provide curated human and animal protein complexes, their subunit composition, structure and functions.
Pathways and Regulation
Proteins participating in consequent steps of a biological pathway are also considered to be functionally related. In a broad sense, biological pathway is a chain of molecular events, such as chemical reactions, conformational changes, binding or dissociation, etc., which leads to certain changes in the cell. Pathguide  is a comprehensive catalog comprising of 702 resources related to pathways and molecular interactions in human and other organisms. Pathways are classified according to prevailing interaction type as metabolic, signaling, and regulatory. Metabolic pathways, representing chains of chemical reactions catalyzed by enzymes, can be found in MetaCyc , which is a part of BioCyc, including pathway-related information for more than 13,000 species. Signalling databases, such as OmniPath , Signor , SignaLink , PhosphoSite , contain literature-curated information on cellular signal transduction via post-translational modifications, relocation, binding or conformational changes. Genetic regulation databases contain manually curated and computationally inferred relationships between genes and transcriptional factors (TFs), e.g. JASPAR , TRANSFAC , or miRNA, e.g. miRTarBase . Large pathway databases, such as KEGG , Reactome  and ConsensusPathDB  are not specialized on a particular type of pathway or process and provide biological interaction of multiple types for human and other organisms, while the other resources have a certain focus, e.g. innate immunity  or a specific disease , .
Since biological pathways are mediated by gene products, proteins or RNAs, pathway data is the invaluable source of functional relationships between genes. However, known pathways cover only a small part of all the existing interactions and not all human genes are fully functionally annotated. Unknown gene functions and interactions can be computationally predicted on the basis of gene co-expression , sequence similarity  or interactions ,  with well-annotated genes. Genes or proteins with expression level correlated across different conditions are likely to be co-regulated and may share functions . Sequence similarity and domain composition can also give a clue about a function of an unannotated protein and help to identify its interaction partners. Recent paralogs may have the same function , but later their functions tend to diverge. Orthologs are more functionally conservative  and therefore functional annotations of genes from related species and PPI  may be transferred on their human orthologs.
The amount of knowledge regarding gene and protein roles in the cell is diverse, enormous and continuously growing. The unification and formalization of this knowledge are crucial to ensure its computational processing and analysis. Gene Ontology consortium  created in 1999, develops and maintains a controlled vocabulary of concepts describing gene functions, localizations and participation in biological processes. GO consortium provides regularly updating  whole-genome annotations, either supported by experimental evidence or computationally predicted, for multiple species, from human to bacteria, which allows within and between-species comparisons of gene functions. GO term enrichment analysis became a community standard for functional annotation of gene sets and interpretation of the experiment results. Since genes sharing GO terms are considered to be functionally related, many gene prioritization tools utilize GO as an additional source of evidence.
2.2.3 Text Mining
Yet another way of establishing putative associations between genes, diseases and other biological entities is text mining of biomedical literature. Many gene prioritization tools utilize the results of co-occurrence based text mining, assuming that frequent colocalization of two entities in biomedical texts points to their possible interaction. More sophisticated pattern-based text-mining methods use advanced weighting schemes to assign qualities to predicted associations , . Other text-mining systems, e.g. ANDsystem , apply natural language processing (NLP) algorithms allowing to differentiate between various kinds of biological entities and associations between them.
Some gene prioritization tools consider text mining-inferred associations together with curated and predicted associations from other evidence sources when some of them (e.g. Génie , GLAD4U ) completely rely on the text-mining results. The major drawbacks of text mining methods are the high rate of false positives and the lack of accuracy in the detection of associations and determination of their types. Despite that, text mining remains the only way to absorb the whole volume of relevant scientific literature, impossible to handle manually.
2.3 Selection of Related Works
Since many of research papers on gene prioritization describe the method but provide no code or link to the tool, in order to achieve better coverage of existing tools we were not confined to works found in Pubmed and Google Scholar by keywords “gene”, “prioritization” or “prioritizing” and “disease”. In addition, we searched in specialized software catalogs Gene Prioritization Portal (http://homes.esat.kuleuven.be/~bioiuser/gpp/) and OMICtools (https://omictools.com/) and included in the list all found tools which were published in scientific journals. Aside from tools, explicitly classified as gene prioritization tools by authors, we included into this review several multi-purpose tools, such as Polysearch , , able to prioritize associations between various biological entities, including, but not limited to genes and diseases. Also, some large biological data portals such as DisGeNET , , , Open Targets  and GeneMania , ,  providing engines for gene prioritization, although it is not their main purpose, were added to comparison. We did not take into account computational tools or resources which can be used for gene prioritization, but required implementation of essential parts of gene prioritization workflow, e.g. scoring functions , . Also, it is important to note that tools developed to solve related, but different problems such as variant prioritization , , , prediction of protein function  or detection of cancer driver genes  were left out of the scope of this review. We also did not consider gene prioritization tools developed specifically for non-human organisms , , or working with a very limited set of human phenotypes, e.g. only for neurodegenerative diseases  or cancers . Despite such tools often use data more relevant for a specific task, they might be less interesting for a broader audience than generic gene prioritization tools. In total we have found references to 96 published tools for gene prioritization in human disorders (Table S1).
2.4 Selection and Characterization of the most Promising Tools
Since many of gene prioritization tools were not available or not supported by their authors, we decided to focus on the most promising ones. The process of selection of the most up-to-date and user-friendly tools for gene prioritization in human disorders is shown in Figure 3. First of all, we excluded from consideration methods which implementations were not published (i.e. no web link was found), what reduced the list to 84 tools. Next, we kept only open source tools with implementations relying on freely available components and tried to run the remaining 75 tools. After exclusion of tools which were offline, failed to run or executed with errors, the list reduced down to 33 tools. Thus, only 34% of 96 gene prioritization tools published since 2002 remained available for users at the beginning of 2018. Finally, we favored tools with the most up-to-date evidence sources, because the completeness of evidence sources is crucial for gene prioritization tool performance . Therefore we checked the dates of the last database update and selected for detailed comparison 14 tools, which have updated their databases since 2016.
We categorized these 14 gene prioritization tools according to assumption and prioritization approach they use (Table 1) and types of evidence sources they rely on (Table 2). Five tools integrated gene-disease associations and accepted disease terms instead of seed genes. Seven other tools searched for genes associated with user-defined seeds and four of them applied network analysis techniques. Two remaining tools, Phenolyzer and PhenoRank, took into account both gene-gene and gene-disease associations and automatically identified most relevant genes for given diseases terms. Brief descriptions of the methods used by all 14 tools and their additional parameters are provided in Supplementary Table 2.
We also paid attention to interfaces of selected tools because they correspond to the level of expertise required from the user. All of the selected tools, except PhenoRank , had web interfaces and therefore were available for a broad audience of users without special programming skills. However, only 6 tools provided programmatic access options (e.g. web API interfaces, scripts or command line tools), advantageous for tasks requiring batch execution.
|integrate gene-disease associations||search for genes associated with seeds||score aggregation||network analysis||web interface||programmatic access||seed genes||disease or phenotype terms||candidate genes|
|DisGeNET ||+||+||+||+||no||yes||whole genome|
|GLAD4U ||+||+||+||no||yes||whole genome|
|Polysearch ||+||+||+||no||yes||whole genome|
|PhenoRank ||+||+||+||+||no||yes||whole genome|
|GeneMANIA ||+||+||+||+||yes||no||whole genome|
|MaxLink ||+||+||+||yes||no||whole genome|
|Gene Interactions||Gene Similarities||Gene-Disease associations|
|Physical PPI||Pathways||Genetic interactions||Regulation||Interologs||Co-expression||Co-localization||Functional annotations||Phenotype similarity||Shared domains||Sequence similarity||Phylogenetic profile similarity||Chemical interaction||Text Mining||Genetic associations||Differential expression||Animal models||Human phenotype similarity||Chemical information||Pathways||Text mining|
3 Evaluation of Gene Prioritization Tools
Although some individual tools may demonstrate superior performance on average, none of them outperform in all cases , , , . This may happen because of the different capability of methods to model aspects relevant to the disease mechanism, or due to the usage of distinct validation approaches and different training and testing datasets. In this chapter, we discuss approaches for validation of gene prioritization results, measures of performance, and their advantages and disadvantages.
3.1 Validation Data
Despite gene prioritization results in a ranked list of genes, in many works, it is turned into a binary classification task, where genes associated with the disease must be distinguished from unrelated ones. The assessment of classification quality requires multiple positive and negative instances, i.e. genes associated and not associated with a disease. Many studies prove the ability of gene prioritization tools to facilitate the identification of genes carrying monogenic diseases variants, e.g. , . Gene prioritization can also advance identification genes relevant for complex traits, allow selection the most promising candidates for GWAS  and selection of targets for RNA interference screening . However, none of these studies represents performance evaluation in its strict sense, with the determination of type I and type II error rates and experimental verification of gene roles.
Since experimental testing of a sufficiently large number of genes is expensive and time-consuming, most of the researchers turn to publicly available databases or scientific papers and construct golden datasets of high-confidence curated associations , . Genes without any evidence of association with the disease are considered as negative exemplars. Finally, positive and negative instances are distributed between testing and training sets (if necessary) and are used to calculate one or several performance measures.
3.2 Measures of Performance
Although many evaluations of gene prioritization tools are described in literature , , , no unified approach for evaluation of gene prioritization tool exists. Doncheva et al. , and Gill et al.  provide a good overview of performance measures applicable for evaluation of gene prioritization tools. Optimal validation approach and measure of performance depend on the nature of the disease (monogenic or polygenic) and tolerance to false positives. For monogenic or oligogenic  diseases by definition just one or few causative genes exist and their experimental validation is time-consuming and laborious. Therefore, in the case when only a few candidates can be tested, suitable performance measures are the averaged number of false candidates appeared above the right one, or success rate, when the correctly identified disease gene was ranked within n top genes.
In contrast with monogenic diseases, complex disorders are associated with multiple genes, and the expected number of true associations may vary depending on the disease. Although the mean rank ratio of true positive findings is applicable in this case, it provides little information about the overall distribution of the true positive ranks. For complex disorders, potentially associated genes are not necessarily tested individually but may be analyzed as a group. For some tasks, such as gene set or pathway enrichment analysis, individual gene ranks are not so crucial as the enrichment of true positives in the top of the list. Averaged fold-enrichment ,  shows the increase of true positive rate (TPR) in the top of the ranked list compared to the background TPR. The method achieves n/m-fold enrichment on average, if in n% of cases correctly identified disease genes are ranked among the top m% of all candidates . Besides the TPR, also known as sensitivity, another essential characteristic of performance is specificity, which shows the ability of the method to correctly classify negative exemplars (true negative rate). The relationship between sensitivity and false positive rate (FPR) plotted at variable threshold levels gives a receiver operating characteristic (ROC) curve . The cross-validated area under the ROC curve (AUC)  provides an estimate of overall tool performance , , , .
3.3 Challenges of Validation
Many methods reduce gene prioritization problem to the problem of finding genes most similar to seed genes and thus the result may strongly depend on seed gene selection. Nevertheless, no universal rule on seed selection exists, and the choice of seeds is usually made subjectively. Expert decisions on optimal number of seed genes vary from several to dozens , ,  or even over a hundred . However, since seed genes must have a proven role in the disease, for monogenic and oligogenic disorders only one or few such genes exist. Consequently, for these diseases, the methods accepting disease terms and analyzing gene-disease associations bypassing seed genes, e.g. via phenotype similarities are more suitable. In turn, multifactorial disorders have many associations of various confidence and power, but by no means all are causal. Furthermore, none of the multifactorial disorders are fully explained, therefore gold standard datasets may be incomplete or contain false associations and therefore give uncertain performance estimates. Finally, if no evidence of association between a gene and a complex disease found, we still cannot be sure whether this gene is a true negative indeed, due to the incompleteness of current knowledge.
At the same time with a lack of reliable ground truth, there are two issues related to the oversupply of biomedical knowledge and its rapid and non-uniform growth. The first problem also referred to as knowledge bias ,  is that well-characterized genes have a better chance to be ranked higher than unannotated ones, only because they have more connections with other biological entities. This effect can be at least partially eliminated by including large-scale experimental data  e.g. whole interactome, co-expression network or the results of GWAS. Cornish et al.  suggest an elegant way to reduce the effect of knowledge bias on gene ranking. They developed PhenoRank, which computes empirical p-values for each gene score, comparing scores obtained on the real disease with scores for the same genes, prioritized for simulated sets of phenotype terms. This improvement allowed PhenoRank outperforming three other gene prioritization tools .
Another problem concerning the reliability of benchmark results is a possible uncontrolled inclusion of testing associations into evidence sources, leading to overestimation of the actual performance of the tool. Tranchevent et al.  proposed to adopt the methodology  termed time-stamped benchmark, in order to reduce this “knowledge contamination”. They saved novel gene-disease association predictions made at the beginning of 2013 and checked which of them were published during the next 2 years. Although in some works researchers apply time-stamped benchmark , , while reducing “knowledge contamination”, it excludes from the validation cohort the most reliable and well-studied associations, which may also result in a biased performance estimate.
4 Future Directions
With the advance of high-throughput technologies, in the last two decades the volume of biomedical knowledge, the variety of evidence sources and their completeness have constantly increased. Large comprehensive data portals collecting gene and gene-disease associations for human and other organisms have been created. Incorporation of new or rarely used evidence sources and improvement of data quality are important directions for future work in the field of gene prioritization.
Simultaneously with extensive development of evidence sources, great progress in the development of gene prioritization algorithms has been made. Although recent studies have shown that machine learning methods tend to outperform simpler score aggregation methods based on statistical approaches , , , , no single method outperforming others exist and different tools seem to be complementary. Therefore, the analysis of the results coming from several tools relying on different principles and data sources remains beneficial.
Besides that, there is also an advancement in usability and flexibility of gene prioritization tools. Modern gene prioritization tools allow selection of the algorithm and tuning its parameters, choosing evidence sources and incorporation of user-defined data. They provide a variety of user interfaces and programmatic access options facilitating integration with other computational tools.
Despite the active development and great success of gene prioritization tools during past decades, some challenges remain not fully addressed. In many works, gene prioritization is referred as the task of ranking genes by their relevance to the disease , , , , or, more specifically, according to the probability to be causal for this disease , , , . If for monogenic diseases, one or several validated causal variants exist, complex disorders are associated with some variants, whose causal roles are not always confirmed. Furthermore, in complex disorders, many genes may act together and modify the effect of each other jointly contributing to disease development. Despite this, most of the reviewed tools rank candidate genes separately, what fits Mendelian diseases but for complex disorders may be disadvantageous. Although in some tools ,  scores of candidate genes influence each other, nearly none of them explicitly boosts scores of candidates forming a putative pathway or complex. Therefore, it might be reasonable to separate the task of gene prioritization for complex disorders from gene prioritization for monogenic disorders and reformulate it. Instead of ranking all the candidates independently, one can look for subsets of functionally interconnected candidate genes, enriched by associations with a specified disease , . The development of specialized protocols for validation of the results gene prioritization for complex diseases is also necessary.
Another aspect of the problem was not addressed by current gene prioritization tools is disease heterogeneity. Disease heterogeneity implies that the disease may be represented by latent subclasses, phenotypically similar but molecularly distinct. The evidence of heterogeneity are shown for many human diseases, complex and monogenic , . In order to take into account disease heterogeneity, one should perform gene prioritization together with analysis of patient-level experimental data. One example of such data integration is variant prioritization, which identifies likely disease-causing rare point mutations , , .
Among about a hundred of gene prioritization tools published to date, we selected and described 14 most promising ones on the basis of their availability, usability, and novelty. In this review, we classified gene prioritization tools according to underlying assumptions, methodology, and data representation models. An optimal tool for gene prioritization must be chosen considering the specificity of a particular task, namely error tolerance, type of inheritance and availability of knowledge about genes and disease of interest. Also, novelty and type of evidence sources utilized by the gene prioritization tool should be taken into account. Finally, we highlighted the limitations of existing gene prioritization tools and discussed the directions of future research.
This study was supported by the International DFG Research Training Group GRK 1906 and the Volkswagen Stiftung Trilateral Partnerships – Cooperation Projects between Scholars and Scientists from Ukraine, Russia, and Germany “In silico screening and experimental validation of new drug targets for the treatment of co-morbid multifactorial diseases” (#90335).
Conflict of interest statement: Authors state no conflict of interest. All authors have read the journal’s Publication ethics and publication malpractice statement available at the journal’s website and hereby confirm that they comply with all its parts applicable to the present scientific work.
 Thienpont B, Zhang L, Postma AV, Breckpot J, Tranchevent LC, Loo PV, et al. Haploinsufficiency of TAB2 causes congenital heart defects in humans. Am J Hum Genet 2010;86:839–49.10.1016/j.ajhg.2010.04.011Search in Google Scholar PubMed PubMed Central
 Erlich Y, Edvardson S, Hodges E, Zenvirt S, Thekkat P, Shaag A, et al. Exome sequencing and disease-network analysis of a single family implicate a mutation in KIF1A in hereditary spastic paraparesis. Genome Res 2011;21:658–64.10.1101/gr.117143.110Search in Google Scholar PubMed PubMed Central
 Pers TH, Timshel P, Ripke S, Sullivan PF, O’Donovan MC, Franke L, et al. Comprehensive analysis of schizophrenia-associated loci highlights ion channel pathways and biologically plausible candidate causal genes. Hum Mol Genet 2016;25:1247–54.10.1093/hmg/ddw007Search in Google Scholar PubMed PubMed Central
 Mahurkar S, Moldovan M, Suppiah V, O’Doherty C. Identification of shared genes and pathways: a comparative study of multiple sclerosis susceptibility, severity and response to interferon beta treatment. PLoS One 2013;8:e57655.10.1371/journal.pone.0057655Search in Google Scholar PubMed PubMed Central
 Lee I, Blom UM, Wang PI, Shim JE, Marcotte EM. Prioritizing candidate disease genes by network-based boosting of genome-wide association data. Genome Res 2011;21:1109–21.10.1101/gr.118992.110Search in Google Scholar PubMed PubMed Central
 Kobayashi M, Yokoyama K, Shimizu E, Yusa N, Ito M, Yamaguchi R, et al. Phenotype-based gene analysis allowed successful diagnosis of X-linked neutropenia associated with a novel WASp mutation. Ann Hematol 2017;97:367–9.10.1007/s00277-017-3134-3Search in Google Scholar PubMed
 van der Valk RJP, Kreiner-Møller E, Kooijman MN, Guxens M, Stergiakouli E, Sääf A, et al. A novel common variant in DCST2 is associated with length in early life and height in adulthood. Hum Mol Genet 2014;24:1155–68.10.1093/hmg/ddu510Search in Google Scholar PubMed PubMed Central
 Shungin D, Winkler TW, Croteau-Chonka DC, Ferreira T, Locke AE, et al. New genetic loci link adipose and insulin biology to body fat distribution. Nature 2015;518(7538):187–96.10.1038/nature14132Search in Google Scholar PubMed PubMed Central
 Saik OV, Demenkov PS, Ivanisenko TV, Bragina EY, Freidin MB, Goncharova IA, et al. Novel candidate genes important for asthma and hypertension comorbidity revealed from associative gene networks. BMC Med Genomics 2018;11:15.10.1186/s12920-018-0331-4Search in Google Scholar PubMed PubMed Central
 Emad A, Cairns J, Kalari KR, Wang L, Sinha S. Knowledge-guided gene prioritization reveals new insights into the mechanisms of chemoresistance. Genome Bio 2017;18. Available from: https://doi.org/10.1186/s13059-017-1282-3.10.1186/s13059-017-1282-3Search in Google Scholar PubMed PubMed Central
 Browne F, Wang H, Zheng H. A computational framework for the prioritization of disease-gene candidates. BMC Genomics 2015;16(Suppl 9):S2.10.1186/1471-2164-16-S9-S2Search in Google Scholar PubMed PubMed Central
 Seyyedrazzagi E, Navimipour NJ. Disease genes prioritizing mechanisms: a comprehensive and systematic literature review. Netw Model Anal Health Inform Bioinform. 2017;6(1). DOI: 10.1007/s13721-017-0154-9.10.1007/s13721-017-0154-9Search in Google Scholar
 Tiffin N. Conceptual thinking for in silico prioritization of candidate disease genes. In: Methods in molecular biology. USA: Humana Press, 2011:175–87. Available from: https://doi.org/10.1007/978-1-61779-176-5_11.10.1007/978-1-61779-176-5_11Search in Google Scholar PubMed
 Tranchevent LC, Capdevila FB, Nitsch D, Moor BD, Causmaecker PD, Moreau Y. A guide to web tools to prioritize candidate genes. Brief Bioinform 2010;12:22–32.10.1093/bib/bbq007Search in Google Scholar PubMed
 Piro RM, Cunto FD. Computational approaches to disease-gene prediction: rationale, classification and successes. FEBS J 2012;279:678–96.10.1111/j.1742-4658.2012.08471.xSearch in Google Scholar PubMed
 Doncheva NT, Kacprowski T, Albrecht M. Recent approaches to the prioritization of candidate disease genes. Wiley Interdiscip Rev Syst Biol Med 2012;4:429–42.10.1002/wsbm.1177Search in Google Scholar PubMed
 Börnigen D, Tranchevent LC, Bonachela-Capdevila F, Devriendt K, Moor BD, Causmaecker PD, et al. An unbiased evaluation of gene prioritization tools. Bioinformatics 2012;28:3081–8.10.1093/bioinformatics/bts581Search in Google Scholar PubMed
 Shim JE, Hwang S, Lee I. Pathway-dependent effectiveness of network algorithms for gene prioritization. PLoS One 2015;10:e0130589.10.1371/journal.pone.0130589Search in Google Scholar PubMed PubMed Central
 Guala D, Sonnhammer ELL. A large-scale benchmark of gene prioritization methods. Sci Rep 2017;7. Available from: https://doi.org/10.1038/srep4659810.1038/srep46598Search in Google Scholar PubMed PubMed Central
 Hassani-Pak K, Rawlings C. Knowledge discovery in biological databases for revealing candidate genes linked to complex phenotypes. J Integr Bioinform 2017;14. Available from: https://doi.org/10.1515/jib-2016-0002.10.1515/jib-2016-0002Search in Google Scholar PubMed PubMed Central
 Martínez V, Cano C, Blanco A. ProphNet: a generic prioritization method through propagation of information. BMC Bioinformatics 2014;15(Suppl 1):S5.10.1186/1471-2105-15-S1-S5Search in Google Scholar PubMed PubMed Central
 MacArthur J, Bowler E, Cerezo M, Gil L, Hall P, Hastings E, et al. The new NHGRI-EBI Catalog of published genome-wide association studies (GWAS Catalog). Nucleic Acids Res 2016;45:D896–901.10.1093/nar/gkw1133Search in Google Scholar PubMed PubMed Central
 Liu Y, Liang Y, Wishart D. PolySearch2: a significantly improved text-mining system for discovering associations between human diseases, genes, drugs, metabolites, toxins and more. Nucleic Acids Res 2015;43:W535–42.10.1093/nar/gkv383Search in Google Scholar PubMed PubMed Central
 Cornish AJ, David A, Sternberg MJE. PhenoRank: reducing study bias in gene prioritization through simulation. Bioinformatics 2018;34:2087–95.10.1093/bioinformatics/bty028Search in Google Scholar PubMed PubMed Central
 Koscielny G, An P, Carvalho-Silva D, Cham JA, Fumis L, Gasparyan R, et al. Open Targets: a platform for therapeutic target identification and validation. Nucleic Acids Res 2016;45:D985–94.10.1093/nar/gkw1055Search in Google Scholar PubMed PubMed Central
 Yang H, Robinson PN, Wang K. Phenolyzer: phenotype-based prioritization of candidate genes for human diseases. Nat Methods 2015;12:841–3.10.1038/nmeth.3484Search in Google Scholar PubMed PubMed Central
 Pinero J, Bravo À, Queralt-Rosinach N, Gutiérrez-Sacristán A, Deu-Pons J, Centeno E, et al. DisGeNET: a comprehensive platform integrating information on human disease-associated genes and variants. Nucleic Acids Res 2016;45:D833–9.10.1093/nar/gkw943Search in Google Scholar PubMed PubMed Central
 Fontaine JF, Priller F, Barbosa-Silva A, Andrade-Navarro MA. Génie: literature-based gene prioritization at multi genomic scale. Nucleic Acids Res 2011;39(suppl 2):W455–61.10.1093/nar/gkr246Search in Google Scholar PubMed PubMed Central
 Jourquin J, Duncan D, Shi Z, Zhang B. GLAD4U: deriving and prioritizing gene lists from PubMed literature. BMC Genomics 2012;13(Suppl 8):S20.10.1186/1471-2164-13-S8-S20Search in Google Scholar PubMed PubMed Central
 Warde-Farley D, Donaldson SL, Comes O, Zuberi K, Badrawi R, Chao P, et al. The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res 2010;38(suppl 2):W214–20.10.1093/nar/gkq537Search in Google Scholar PubMed PubMed Central
 Ghiassian SD, Menche J, Barabási AL. A DIseAse MOdule Detection (DIAMOnD) algorithm derived from a systematic analysis of connectivity patterns of disease proteins in the human Interactome. PLoS Comput Biol 2015;11:e1004120.10.1371/journal.pcbi.1004120Search in Google Scholar PubMed PubMed Central
 Tranchevent LC, Ardeshirdavani A, ElShal S, Alcaide D, Aerts J, Auboeuf D, et al. Candidate gene prioritization with Endeavour. Nucleic Acids Res 2016;44:W117–21.10.1093/nar/gkw365Search in Google Scholar PubMed PubMed Central
 Kacprowski T, Doncheva NT, Albrecht M. NetworkPrioritizer: a versatile tool for network-based prioritization of candidate disease genes or other molecules. Bioinformatics 2013;29:1471–3.10.1093/bioinformatics/btt164Search in Google Scholar PubMed PubMed Central
 Meshkin A, Shakery A, Masoudi-Nejad A. GPS: Identification of disease genes by rank aggregation of multi-genomic scoring schemes. Genomics 2018;111:612–8.10.1016/j.ygeno.2018.03.017Search in Google Scholar PubMed
 Menche J, Sharma A, Kitsak M, Ghiassian SD, Vidal M, Loscalzo J, et al. Uncovering disease-disease relationships through the incomplete interactome. Science 2015;347:1257601.10.1126/science.1257601Search in Google Scholar PubMed PubMed Central
 Goh KI, Cusick ME, Valle D, Childs B, Vidal M, Barabasi AL. The human disease network. Proc Natl Acad Sci USA 2007;104:8685–90.10.1073/pnas.0701361104Search in Google Scholar PubMed PubMed Central
 Ozgur A, Vu T, Erkan G, Radev DR. Identifying gene-disease associations using centrality on a literature mined gene-interaction network. Bioinformatics 2008;24:i277–85.10.1093/bioinformatics/btn182Search in Google Scholar PubMed PubMed Central
 Lee Y, Li H, Li J, Rebman E, Achour I, Regan KE, et al. Network models of genome-wide association studies uncover the topological centrality of protein interactions in complex diseases. J Am Med Inform Assoc 2013;20:619–29.10.1136/amiajnl-2012-001519Search in Google Scholar PubMed PubMed Central
 Bragina EY, Tiys ES, Rudko AA, Ivanisenko VA, Freidin MB. Novel tuberculosis susceptibility candidate genes revealed by the reconstruction and analysis of associative networks. Infect Genet Evol 2016;46:118–23.10.1016/j.meegid.2016.10.030Search in Google Scholar PubMed
 Saik OV, Demenkov PS, Ivanisenko TV, Bragina EY, Freidin MB, Dosenko VE, et al. Search for new candidate genes involved in the comorbidity of asthma and hypertension based on automatic analysis of scientific literature. J Integr Bioinform 2018;15. Available from: https://doi.org/10.1515/jib-2018-0054.10.1515/jib-2018-0054Search in Google Scholar PubMed PubMed Central
 Guala D, Sjölund E, Sonnhammer ELL. MaxLink: network-based prioritization of genes tightly linked to a disease seed set. Bioinformatics 2014;30:2689–90.10.1093/bioinformatics/btu344Search in Google Scholar PubMed
 George RA, Liu JY, Feng LL, Bryson-Richardson RJ, Fatkin D, Wouters MA. Analysis of protein sequence and interaction data for candidate disease gene prediction. Nucleic Acids Res 2006;34:e130.10.1093/nar/gkl707Search in Google Scholar PubMed PubMed Central
 Guney E, Oliva B. Exploiting protein-protein interaction networks for genome-wide disease-gene prioritization. PLoS One 2012;7:e43557.10.1371/journal.pone.0043557Search in Google Scholar PubMed PubMed Central
 Chen J, Bardes EE, Aronow BJ, Jegga AG. ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res 2009;37:W305–11.10.1093/nar/gkp427Search in Google Scholar PubMed PubMed Central
 Köhler S, Bauer S, Horn D, Robinson PN. Walking the interactome for prioritization of candidate disease genes. Am J Hum Genet 2008;82:949–58.10.1016/j.ajhg.2008.02.013Search in Google Scholar PubMed PubMed Central
 Luo J, Liang S. Prioritization of potential candidate disease genes by topological similarity of protein–protein interaction network and phenotype data. J Biomed Inform 2015;53:229–36.10.1016/j.jbi.2014.11.004Search in Google Scholar PubMed
 Le DH, Kwon YK. GPEC: a Cytoscape plug-in for random walk-based gene prioritization and biomedical evidence collection. Comput Biol Chem 2012;37:17–23.10.1016/j.compbiolchem.2012.02.004Search in Google Scholar PubMed
 Lysenko A, Boroevich KA, Tsunoda T. Arete – candidate gene prioritization using biological network topology with additional evidence types. BioData Mining 2017;10:22.10.1186/s13040-017-0141-9Search in Google Scholar PubMed PubMed Central
 Le DH, Kwon YK. Neighbor-favoring weight reinforcement to improve random walk-based disease gene prioritization. Comput Biol Chem 2013;44:1–8.10.1016/j.compbiolchem.2013.01.001Search in Google Scholar PubMed
 Erten S, Bebek G, Koyutürk M. Vavien: an algorithm for prioritizing candidate disease genes based on topological similarity of proteins in interaction networks. J Comput Biol 2011;18:1561–74.10.1089/cmb.2011.0154Search in Google Scholar PubMed PubMed Central
 Li Y, Patra JC. Genome-wide inferring gene–phenotype relationship by walking on the heterogeneous network. Bioinformatics 2010;26:1219–24.10.1093/bioinformatics/btq108Search in Google Scholar PubMed
 White S, Smyth P. Algorithms for estimating relative importance in networks. In: Proceedings of the ninth ACM SIGKDD international conference on Knowledge discovery and data mining – KDD ’03. ACM Press, 2003. Available from: https://doi.org/10.1145/956750.956782.10.1145/956750.956782Search in Google Scholar
 Vanunu O, Magger O, Ruppin E, Shlomi T, Sharan R. Associating genes and protein complexes with disease via network propagation. PLoS Comput Biol 2010;6:e1000641.10.1371/journal.pcbi.1000641Search in Google Scholar PubMed PubMed Central
 Fang M, Hu X, He T, et al. Prioritizing disease-causing genes based on network diffusion and rank concordance. In: 2014 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2014. Available from: https://doi.org/10.1109/bibm.2014.6999162.10.1109/BIBM.2014.6999162Search in Google Scholar
 Mosca E, Bersanelli M, Gnocchi M, Moscatelli M, Castellani G, Milanesi L, et al. Network diffusion-based prioritization of Autism risk genes identifies significantly connected gene modules. Front Genet 2017;8. Available from: https://doi.org/10.3389/fgene.2017.00129.10.3389/fgene.2017.00129Search in Google Scholar PubMed PubMed Central
 Suthram S, Beyer A, Karp RM, Eldar Y, Ideker T. eQED: an efficient method for interpreting eQTL associations using protein networks. Mol Syst Biol 2008;4:162.10.1038/msb.2008.4Search in Google Scholar PubMed PubMed Central
 Mostafavi S, Ray D, Warde-Farley D, Grouios C, Morris Q. GeneMANIA: a real-time multiple association network integration algorithm for predicting gene function. Genome Biol 2008;9(Suppl 1):S4.10.1186/gb-2008-9-s1-s4Search in Google Scholar PubMed PubMed Central
 Zhu X, Ghahramani Z, Lafferty J. Semi-supervised learning using gaussian fields and harmonic functions. In: IN ICML, 2003:912–9.Search in Google Scholar
 Navlakha S, Kingsford C. The power of protein interaction networks for associating genes with diseases. Bioinformatics 2010;26:1057–63.10.1093/bioinformatics/btq076Search in Google Scholar PubMed PubMed Central
 Cheng D, Knox C, Young N, Stothard P, Damaraju S, Wishart DS. PolySearch: a web-based text mining system for extracting relationships between human diseases, genes, mutations, drugs and metabolites. Nucleic Acids Res 2008;36:W399–405.10.1093/nar/gkn296Search in Google Scholar PubMed PubMed Central
 Li J, Lin X, Teng Y, Qi S, Xiao D, Zhang J, et al. A comprehensive evaluation of disease phenotype networks for gene Prioritization. PLoS One 2016;11:e0159457.10.1371/journal.pone.0159457Search in Google Scholar PubMed PubMed Central
 Huang JK, Carlin DE, Yu MK, Zhang W, Kreisberg JF, Tamayo P, et al. Systematic evaluation of molecular networks for discovery of disease genes. Cell Syst 2018;6:484–95.e5.10.1016/j.cels.2018.03.001Search in Google Scholar PubMed PubMed Central
 Schmitt T, Ogris C, Sonnhammer ELL. FunCoup 3.0: database of genome-wide functional coupling networks. Nucleic Acids Res 2013;42:D380–8.10.1093/nar/gkt984Search in Google Scholar PubMed PubMed Central
 Jensen LJ, Kuhn M, Stark M, Chaffron S, Creevey C, Muller J, et al. STRING 8–a global view on proteins and their functional interactions in 630 organisms. Nucleic Acids Res 2009;37:D412–6.10.1093/nar/gkn760Search in Google Scholar PubMed PubMed Central
 Himmelstein DS, Baranzini SE. Heterogeneous network edge prediction: a data integration approach to prioritize disease-associated genes. PLOS Comput Biol 2015;11:e1004259.10.1371/journal.pcbi.1004259Search in Google Scholar PubMed PubMed Central
 Zhang Y, Liu J, Liu X, Fan X, Hong Y, Wang Y, et al. Prioritizing disease genes with an improved dual label propagation framework. BMC Bioinformatics 2018;19:47.10.1186/s12859-018-2040-6Search in Google Scholar PubMed PubMed Central
 Pinero J, Queralt-Rosinach N, Bravo A, Deu-Pons J, Bauer-Mehren A, Baron M, et al. DisGeNET: a discovery platform for the dynamical exploration of human diseases and their genes. Database 2015;2015:bav028.10.1093/database/bav028Search in Google Scholar PubMed PubMed Central
 Bauer-Mehren A, Rautschka M, Sanz F, Furlong LI. DisGeNET: a Cytoscape plugin to visualize, integrate, search and analyze gene–disease networks. Bioinformatics 2010;26:2924–6.10.1093/bioinformatics/btq538Search in Google Scholar PubMed
 Chen J, Xu H, Aronow BJ, Jegga AG. Improved human disease candidate gene prioritization using mouse phenotype. BMC Bioinformatics 2007;8:392.10.1186/1471-2105-8-392Search in Google Scholar PubMed PubMed Central
 Tranchevent LC, Barriot R, Yu S, Vooren SV, Loo PV, Coessens B, et al. ENDEAVOUR update: a web resource for gene prioritization in multiple species. Nucleic Acids Res 2008;36:W377–84.10.1093/nar/gkn325Search in Google Scholar PubMed PubMed Central
 Kumar AA, Laer LV, Alaerts M, Ardeshirdavani A, Moreau Y, Laukens K, et al. pBRIT: gene prioritization by correlating functional and phenotypic annotations through integrative data fusion. Bioinformatics 2018;34:2254–62.10.1093/bioinformatics/bty079Search in Google Scholar PubMed PubMed Central
 Chen Y, Wu X, Jiang R. Integrating human omics data to prioritize candidate genes. BMC Med Genomics 2013;6. Available from: https://doi.org/10.1186/1755-8794-6-57.10.1186/1755-8794-6-57Search in Google Scholar PubMed PubMed Central
 Wagner AH, Taylor KR, DeLuca AP, Casavant TL, Mullins RF, Stone EM, et al. Prioritization of retinal disease genes: an integrative approach. Hum Mutat 2013;34:853–9.10.1002/humu.22317Search in Google Scholar PubMed PubMed Central
 Nitsch D, Gonçalves JP, Ojeda F, de Moor B, Moreau Y. Candidate gene prioritization by network analysis of differential expression using machine learning approaches. BMC Bioinformatics 2010;11:460.10.1186/1471-2105-11-460Search in Google Scholar PubMed PubMed Central
 Cogill S, Wang L. Support vector machine model of developmental brain gene expression data for prioritization of Autism risk gene candidates. Bioinformatics 2016;32:3611–8.10.1093/bioinformatics/btw498Search in Google Scholar PubMed
 Zampieri G, Tran DV, Donini M, Navarin N, Aiolli F, Sperduti A, et al. Scuba: scalable kernel-based gene prioritization. BMC Bioinformatics 2018;19:23.10.1186/s12859-018-2025-5Search in Google Scholar PubMed PubMed Central
 Gan M, Li W, Zeng W, Wang X, Jiang R. Mimvec: a deep learning approach for analyzing the human phenome. BMC Systems Biology 2017;11. Available from: https://doi.org/10.1186/s12918-017-0451-z.10.1186/s12918-017-0451-zSearch in Google Scholar PubMed PubMed Central
 Isakov O, Dotan I, Ben-Shachar S. Machine learning–based gene prioritization identifies novel candidate risk genes for inflammatory bowel disease. Inflamm Bowel Dis 2017;23:1516–23.10.1097/MIB.0000000000001222Search in Google Scholar PubMed
 Yu S, Falck T, Daemen A, Tranchevent LC, Suykens JA, Moor BD, et al. L2-norm multiple kernel learning and its application to biomedical data fusion. BMC Bioinformatics 2010;11:309.10.1186/1471-2105-11-309Search in Google Scholar PubMed PubMed Central
 Mordelet F, Vert JP. ProDiGe: Prioritization Of Disease Genes with multitask machine learning from positive and unlabeled examples. BMC Bioinformatics 2011;12:389.10.1186/1471-2105-12-389Search in Google Scholar PubMed PubMed Central
 Pers TH, Dworzyński P, Thomas CE, Lage K, Brunak S. MetaRanker 2.0: a web server for prioritization of genetic variation data. Nucleic Acids Res 2013;41:W104–8.10.1093/nar/gkt387Search in Google Scholar PubMed PubMed Central
 Kolde R, Laur S, Adler P, Vilo J. Robust rank aggregation for gene list integration and meta-analysis. Bioinformatics 2012;28:573–80.10.1093/bioinformatics/btr709Search in Google Scholar PubMed PubMed Central
 Franke L, van Bakel H, Fokkens L, de Jong ED, Egmont-Petersen M, Wijmenga C. Reconstruction of a functional human gene network, with an application for prioritizing positional candidate genes. Am J Hum Genet 2006;78:1011–25.10.1086/504300Search in Google Scholar PubMed PubMed Central
 Maass PG, Aydin A, Luft FC, Schächterle C, Weise A, Stricker S, et al. PDE3A mutations cause autosomal dominant hypertension with brachydactyly. Nature Genet 2015;47:647–53.10.1038/ng.3302Search in Google Scholar PubMed
 Amberger J, Bocchini CA, Scott AF, Hamosh A. McKusick’s online mendelian inheritance in man (OMIM(R)). Nucleic Acids Res 2009;37:D793–6.10.1093/nar/gkn665Search in Google Scholar PubMed PubMed Central
 Pavan S, Rommel K, Marquina MEM, Höhn S, Lanneau V, Rath A. Clinical practice guidelines for rare diseases: the orphanet database. PLoS One 2017;12:e0170365.10.1371/journal.pone.0170365Search in Google Scholar PubMed PubMed Central
 Firth HV, Richards SM, Bevan AP, Clayton S, Corpas M, Rajan D, et al. DECIPHER: database of chromosomal imbalance and phenotype in humans using ensembl resources. Am J Hum Genet 2009;84:524–33.10.1016/j.ajhg.2009.03.010Search in Google Scholar PubMed PubMed Central
 Landrum MJ, Lee JM, Benson M, Brown G, Chao C, Chitipiralla S, et al. ClinVar: public archive of interpretations of clinically relevant variants. Nucleic Acids Res 2015;44:D862–8.10.1093/nar/gkv1222Search in Google Scholar PubMed PubMed Central
 Smith CL, Blake JA, Kadin JA, Richardson JE, Bult CJ, Mouse Genome Database Group. Mouse Genome Database (MGD)-2018: knowledgebase for the laboratory mouse. Nucleic Acids Res 2017;46:D836–42.10.1093/nar/gkx1006Search in Google Scholar PubMed PubMed Central
 Shimoyama M, Pons JD, Hayman GT, Laulederkind SJF, Liu W, Nigam R, et al. The Rat Genome Database 2015: genomic, phenotypic and environmental variations and disease. Nucleic Acids Res 2014;43:D743–50.10.1093/nar/gku1026Search in Google Scholar PubMed PubMed Central
 Hindorff LA, Sethupathy P, Junkins HA, Ramos EM, Mehta JP, Collins FS, et al. Potential etiologic and functional implications of genome-wide association loci for human diseases and traits. Proc Natl Acad Sci USA 2009;106:9362–7.10.1073/pnas.0903103106Search in Google Scholar PubMed PubMed Central
 Edwards SL, Beesley J, French JD, Dunning AM. Beyond GWASs: illuminating the dark road from association to function. Am J Hum Genet 2013;93:779–97.10.1016/j.ajhg.2013.10.012Search in Google Scholar PubMed PubMed Central
 Joehanes R, Zhang X, Huan T, Yao C, xia Ying S, Nguyen QT, et al. Integrated genome-wide analysis of expression quantitative trait loci aids interpretation of genomic association studies. Genome Biol 2017;18:16.10.1186/s13059-016-1142-6Search in Google Scholar PubMed PubMed Central
 Cookson W, Liang L, Abecasis G, Moffatt M, Lathrop M. Mapping complex disease traits with global gene expression. Nat Rev Genet 2009;10:184–94.10.1038/nrg2537Search in Google Scholar PubMed PubMed Central
 Chumbalkar VC, Subhashini C, Dhople VM, Sundaram CS, Jagannadham MV, Kumar KN, et al. Differential protein expression in human gliomas and molecular insights. Proteomics 2005;5:1167–77.10.1002/pmic.200401202Search in Google Scholar PubMed
 Gebregiworgis T, Powers R. Application of NMR metabolomics to search for human disease biomarkers. Comb Chem High Throughput Screen 2012;15:595–610.10.2174/138620712802650522Search in Google Scholar PubMed PubMed Central
 Kolesnikov N, Hastings E, Keays M, Melnichuk O, Tang YA, Williams E, et al. ArrayExpress update—simplifying data submissions. Nucleic Acids Res 2014;43:D1113–6.10.1093/nar/gku1057Search in Google Scholar PubMed PubMed Central
 Barrett T, Wilhite SE, Ledoux P, Evangelista C, Kim IF, Tomashevsky M, et al. NCBI GEO: archive for functional genomics data sets—update. Nucleic Acids Res 2012;41:D991–5.10.1093/nar/gks1193Search in Google Scholar PubMed PubMed Central
 Ardlie KG, Deluca DS, Segre AV, Sullivan TJ, Young TR, Gelfand ET, et al. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science 2015;348:648–60.10.1126/science.1262110Search in Google Scholar PubMed PubMed Central
 Berglund L, Björling E, Oksvold P, Fagerberg L, Asplund A, Szigyarto CAK, et al. A genecentric human protein atlas for expression profiles based on antibodies. Mol Cell Proteomics 2008;7:2019–27.10.1074/mcp.R800013-MCP200Search in Google Scholar PubMed
 Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A, et al. Tissue-based map of the human proteome. Science 2015;347:1260419.10.1126/science.1260419Search in Google Scholar PubMed
 Antanaviciute A, Daly C, Crinnion LA, Markham AF, Watson CM, Bonthron DT, et al. GeneTIER: prioritization of candidate disease genes using tissue-specific gene expression profiles. Bioinformatics 2015;31:2728–35.10.1093/bioinformatics/btv196Search in Google Scholar PubMed PubMed Central
 Davis AP, Grondin CJ, Johnson RJ, Sciaky D, King BL, McMorran R, et al. The comparative toxicogenomics database: update 2017. Nucleic Acids Res 2016;45:D972–8.10.1093/nar/gkw838Search in Google Scholar PubMed PubMed Central
 Rzhetsky A, Wajngurt D, Park N, Zheng T. Probing genetic overlap among complex human phenotypes. Proc Natl Acad Sci USA 2007;104:11694–9.10.1073/pnas.0704820104Search in Google Scholar PubMed PubMed Central
 Kibbe WA, Arze C, Felix V, Mitraka E, Bolton E, Fu G, et al. Disease Ontology 2015 update: an expanded and updated database of human diseases for linking biomedical knowledge through disease data. Nucleic Acids Res 2014;43:D1071–8.10.1093/nar/gku1011Search in Google Scholar PubMed PubMed Central
 Köhler S, Doelken SC, Mungall CJ, Bauer S, Firth HV, Bailleul-Forestier I, et al. The Human Phenotype Ontology project: linking molecular biology and disease through phenotype data. Nucleic Acids Res 2013;42:D966–74.10.1093/nar/gkt1026Search in Google Scholar PubMed PubMed Central
 Groza T, Köhler S, Moldenhauer D, Vasilevsky N, Baynam G, Zemojtel T, et al. The human phenotype ontology: semantic unification of common and rare disease. Am J Hum Genet 2015;97:111–24.10.1016/j.ajhg.2015.05.020Search in Google Scholar PubMed PubMed Central
 Köhler S, Vasilevsky NA, Engelstad M, Foster E, McMurry J, Aymé S, et al. The human phenotype ontology in 2017. Nucleic Acids Res 2016;45:D865–76.10.1093/nar/gkw1039Search in Google Scholar PubMed PubMed Central
 Mungall CJ, McMurry JA, Köhler S, Balhoff JP, Borromeo C, Brush M, et al. The Monarch Initiative: an integrative data and analytic platform connecting phenotypes to genotypes across species. Nucleic Acids Res 2016;45:D712–22.10.1093/nar/gkw1128Search in Google Scholar PubMed PubMed Central
 Smith CL, Eppig JT. The mammalian phenotype ontology: enabling robust annotation and comparative analysis. Wiley Interdiscip Rev Syst Biol Med 2009;1:390–9.10.1002/wsbm.44Search in Google Scholar PubMed PubMed Central
 Prasad TSK, Goel R, Kandasamy K, Keerthikumar S, Kumar S, Mathivanan S, et al. Human protein reference database–2009 update. Nucleic Acids Res 2009;37:D767–72.10.1093/nar/gkn892Search in Google Scholar PubMed PubMed Central
 Chatr-aryamontri A, Oughtred R, Boucher L, Rust J, Chang C, Kolas NK, et al. The BioGRID interaction database: 2017 update. Nucleic Acids Res 2016;45:D369–79.10.1093/nar/gkw1102Search in Google Scholar PubMed PubMed Central
 Breuer K, Foroushani AK, Laird MR, Chen C, Sribnaia A, Lo R, et al. InnateDB: systems biology of innate immunity and beyond—recent updates and continuing curation. Nucleic Acids Res 2012;41:D1228–33.10.1093/nar/gks1147Search in Google Scholar PubMed PubMed Central
 Launay G, Salza R, Multedo D, Thierry-Mieg N, Ricard-Blum S. MatrixDB, the extracellular matrix interaction database: updated content, a new navigator and expanded functionalities. Nucleic Acids Res 2014;43:D321–7.10.1093/nar/gku1091Search in Google Scholar PubMed PubMed Central
 Huttlin EL, Ting L, Bruckner RJ, Gebreab F, Gygi MP, Szpyt J, et al. The BioPlex network: a systematic exploration of the human interactome. Cell 2015;162:425–40.10.1016/j.cell.2015.06.043Search in Google Scholar PubMed PubMed Central
 Huttlin EL, Bruckner RJ, Paulo JA, Cannon JR, Ting L, Baltier K, et al. Architecture of the human interactome defines protein communities and disease networks. Nature 2017;545:505–9.10.1038/nature22366Search in Google Scholar PubMed PubMed Central
 Orchard S, Ammari M, Aranda B, Breuza L, Briganti L, Broackes-Carter F, et al. The MIntAct project—IntAct as a common curation platform for 11 molecular interaction databases. Nucleic Acids Res 2013;42:D358–63.10.1093/nar/gkt1115Search in Google Scholar PubMed PubMed Central
 Licata L, Briganti L, Peluso D, Perfetto L, Iannuccelli M, Galeota E, et al. MINT, the molecular interaction database: 2012 update. Nucleic Acids Res 2011;40:D857–61.10.1093/nar/gkr930Search in Google Scholar PubMed PubMed Central
 López Y, Nakai K, Patil A. HitPredict version 4: comprehensive reliability scoring of physical protein–protein interactions from more than 100 species. Database 2015;2015:bav117.10.1093/database/bav117Search in Google Scholar PubMed PubMed Central
 Ruepp A, Waegele B, Lechner M, Brauner B, Dunger-Kaltenbach I, Fobo G, et al. CORUM: the comprehensive resource of mammalian protein complexes—2009. Nucleic Acids Res 2009;38(suppl 1):D497–501.10.1093/nar/gkp914Search in Google Scholar PubMed PubMed Central
 Meldal BHM, Forner-Martinez O, Costanzo MC, Dana J, Demeter J, Dumousseau M, et al. The complex portal – an encyclopaedia of macromolecular complexes. Nucleic Acids Res 2014;43:D479–84.10.1093/nar/gku975Search in Google Scholar PubMed PubMed Central
 Caspi R, Billington R, Ferrer L, Foerster H, Fulcher CA, Keseler IM, et al. The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res 2015;44:D471–80.10.1093/nar/gkv1164Search in Google Scholar PubMed PubMed Central
 Türei D, Korcsmáros T, Saez-Rodriguez J. OmniPath: guidelines and gateway for literature-curated signaling pathway resources. Nat Methods 2016;13:966–7.10.1038/nmeth.4077Search in Google Scholar PubMed
 Perfetto L, Briganti L, Calderone A, Perpetuini AC, Iannuccelli M, Langone F, et al. SIGNOR: a database of causal relationships between biological entities. Nucleic Acids Res 2015;44:D548–54.10.1093/nar/gkv1048Search in Google Scholar PubMed PubMed Central
 Fazekas D, Koltai M, Türei D, Módos D, Pálfy M, Dúl Z, et al. SignaLink 2 – a signaling pathway resource with multi-layered regulatory networks. BMC Systems Biology 2013;7:7.10.1186/1752-0509-7-7Search in Google Scholar PubMed PubMed Central
 Hornbeck PV, Zhang B, Murray B, Kornhauser JM, Latham V, Skrzypek E. PhosphoSitePlus, 2014: mutations, PTMs and recalibrations. Nucleic Acids Res 2014;43:D512–20.10.1093/nar/gku1267Search in Google Scholar PubMed PubMed Central
 Khan A, Fornes O, Stigliani A, Gheorghe M, Castro-Mondragon JA, vand̃er Lee R, et al. JASPAR 2018: update of the open-access database of transcription factor binding profiles and its web framework. Nucleic Acids Res 2017;46:D260–6.10.1093/nar/gkx1126Search in Google Scholar PubMed PubMed Central
 Wingender E. The TRANSFAC project as an example of framework technology that supports the analysis of genomic regulation. Brief Bioinform 2008;9:326–32.10.1093/bib/bbn016Search in Google Scholar PubMed
 Chou CH, Shrestha S, Yang CD, Chang NW, Lin YL, Liao KW, et al. miRTarBase update 2018: a resource for experimentally validated microRNA-target interactions. Nucleic Acids Res 2017;46:D296–302.10.1093/nar/gkx1067Search in Google Scholar PubMed PubMed Central
 Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res 2015;44:D457–62.10.1093/nar/gkv1070Search in Google Scholar PubMed PubMed Central
 Fabregat A, Jupe S, Matthews L, Sidiropoulos K, Gillespie M, Garapati P, et al. The reactome pathway knowledgebase. Nucleic Acids Res 2017;46:D649–55.10.1093/nar/gkx1132Search in Google Scholar PubMed PubMed Central
 Kamburov A, Stelzl U, Lehrach H, Herwig R. The ConsensusPathDB interaction database: 2013 update. Nucleic Acids Res 2012;41:D793–800.10.1093/nar/gks1055Search in Google Scholar PubMed PubMed Central
 Kalathur RKR, Pinto JP, Sahoo B, Chaurasia G, Futschik ME. HDNetDB: a molecular interaction database for network-oriented investigations into Huntington’s disease. Sci Rep 2017;7. Available from: https://doi.org/10.1038/s41598-017-05224-0.10.1038/s41598-017-05224-0Search in Google Scholar PubMed PubMed Central
 Mizuno S, Iijima R, Ogishima S, Kikuchi M, Matsuoka Y, Ghosh S, et al. AlzPathway: a comprehensive map of signaling pathways of Alzheimer’s disease. BMC Syst Biol 2012;6:52.10.1186/1752-0509-6-52Search in Google Scholar PubMed PubMed Central
 van Dam S, Võsa U, van der Graaf A, Franke L, de Magalhães JP. Gene co-expression analysis for functional classification and gene–disease predictions. Brief Bioinform 2017;19:575–92.10.1093/bib/bbw139Search in Google Scholar PubMed PubMed Central
 Conesa A, Gotz S, Garcia-Gomez JM, Terol J, Talon M, Robles M. Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 2005;21:3674–6.10.1093/bioinformatics/bti610Search in Google Scholar PubMed
 Vazquez A, Flammini A, Maritan A, Vespignani A. Global protein function prediction from protein-protein interaction networks. Nat Genet 2003;21:697–700.10.1038/nbt825Search in Google Scholar PubMed
 Letovsky S, Kasif S. Predicting protein function from protein/protein interaction data: a probabilistic approach. Bioinformatics 2003;19(Suppl 1):i197–204.10.1093/bioinformatics/btg1026Search in Google Scholar PubMed
 Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool for the unification of biology. Nat Genet 2000;25:25–9.10.1038/75556Search in Google Scholar PubMed PubMed Central
 Ivanisenko VA, Saik OV, Ivanisenko NV, Tiys ES, Ivanisenko TV, Demenkov PS, et al. ANDSystem: an Associative Network Discovery System for automated literature mining in the field of biology. BMC Syst Biol 2015;9(Suppl 2):S2.10.1186/1752-0509-9-S2-S2Search in Google Scholar PubMed PubMed Central
 Zuberi K, Franz M, Rodriguez H, Montojo J, Lopes CT, Bader GD, et al. GeneMANIA prediction server 2013 update. Nucleic Acids Res 2013;41:W115–22.10.1093/nar/gkt533Search in Google Scholar PubMed PubMed Central
 Chen YA, Tripathi LP, Mizuguchi K. TargetMine, an integrated data warehouse for candidate gene prioritisation and target discovery. PLoS One 2011;6:e17844.10.1371/journal.pone.0017844Search in Google Scholar PubMed PubMed Central
 Anderson D, Lassmann T. A phenotype centric benchmark of variant prioritisation tools. NPJ Genom Med 2018;3. Available from: https://doi.org/10.1038/s41525-018-0044-9.10.1038/s41525-018-0044-9Search in Google Scholar PubMed PubMed Central
 Eilbeck K, Quinlan A, Yandell M. Settling the score: variant prioritization and Mendelian disease. Nat Rev Genet 2017;18:599–612.10.1038/nrg.2017.52Search in Google Scholar PubMed PubMed Central
 Radivojac P, Clark WT, Oron TR, Schnoes AM, Wittkop T, Sokolov A, et al. A large-scale evaluation of computational protein function prediction. Nat Methods 2013;10:221–7.10.1038/nmeth.2340Search in Google Scholar PubMed PubMed Central
 Porta-Pardo E, Kamburov A, Tamborero D, Pons T, Grases D, Valencia A, et al. Comparison of algorithms for the detection of cancer drivers at subgene resolution. Nat Methods 2017;14:782–8.10.1038/nmeth.4364Search in Google Scholar PubMed PubMed Central
 Shin J, Yang S, Kim E, Kim CY, Shim H, Cho A, et al. FlyNet: a versatile network prioritization server for the Drosophila community. Nucleic Acids Res 2015;43:W91–7.10.1093/nar/gkv453Search in Google Scholar PubMed PubMed Central
 Zwaenepoel A, Diels T, Amar D, Parys TV, Shamir R, de Peer YV, et al. MorphDB: prioritizing genes for specialized metabolism pathways and gene ontology categories in plants. Front Plant Sci 2018;9. Available from: https://doi.org/10.3389/fpls.2018.00352.10.3389/fpls.2018.00352Search in Google Scholar PubMed PubMed Central
 Hassani-Pak K, Castellote M, Esch M, Hindle M, Lysenko A, Taubert J, et al. Developing integrated crop knowledge networks to advance candidate gene discovery. Appl Transl Genom 2016;11:18–26.10.1016/j.atg.2016.10.003Search in Google Scholar PubMed PubMed Central
 Su L, Liu G, Bai T, Meng X, Ma Q. MGOGP: a gene module-based heuristic algorithm for cancer-related gene prioritization. BMC Bioinformatics 2018;19. Available from: https://doi.org/10.1186/s12859-018-2216-0.10.1186/s12859-018-2216-0Search in Google Scholar PubMed PubMed Central
 Oti M, Ballouz S, Wouters MA. Web tools for the prioritization of candidate disease genes. In: Methods in molecular biology. Humana Press, 2011:189–206. Available from: https://doi.org/10.1007/978-1-61779-176-5_12.10.1007/978-1-61779-176-5_12Search in Google Scholar PubMed
 Yu L, Wynn J, Cheung YH, Shen Y, Mychaliska GB, Crombleholme TM, et al. Variants in GATA4 are a rare cause of familial and sporadic congenital diaphragmatic hernia. Hum Genet 2012;132:285–92.10.1007/s00439-012-1249-0Search in Google Scholar PubMed PubMed Central
 Perez-Iratxeta C, Bork P, Andrade-Navarro MA. Update of the G2D tool for prioritization of gene candidates to inherited diseases. Nucleic Acids Res 2007;35:W212–6.10.1093/nar/gkm223Search in Google Scholar PubMed PubMed Central
 Hériché JK, Lees JG, Morilla I, Walter T, Petrova B, Roberti MJ, et al. Integration of biological data by kernels on graph nodes allows prediction of new genes involved in mitotic chromosome condensation. Mol Biol Cell 2014;25:2522–36.10.1091/mbc.e13-04-0221Search in Google Scholar PubMed PubMed Central
 Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29–36.10.1148/radiology.143.1.7063747Search in Google Scholar PubMed
 Lasko TA, Bhagwat JG, Zou KH, Ohno-Machado L. The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform 2005;38:404–15.10.1016/j.jbi.2005.02.008Search in Google Scholar PubMed
 Sharma A, Menche J, Huang CC, Ort T, Zhou X, Kitsak M, et al. A disease module in the interactome explains disease heterogeneity, drug response and captures novel pathways and genes in asthma. Hum Mol Genet 2015;24:3005–20.10.1093/hmg/ddv001Search in Google Scholar PubMed PubMed Central
 Liu B, Jin M, Zeng P. Prioritization of candidate disease genes by combining topological similarity and semantic similarity. J Biomed Inform 2015;57:1–5.10.1016/j.jbi.2015.07.005Search in Google Scholar PubMed
 Zitnik M, Sosič R, Leskovec J. Prioritizing network communities. Nat Commun 2018;9. Available from: https://doi.org/10.1038/s41467-018-04948-5.10.1038/s41467-018-04948-5Search in Google Scholar PubMed PubMed Central
 Peltonen L, Perola M, Naukkarinen J, Palotie A. Lessons from studying monogenic disease for common disease. Hum Mol Gene 2006;15(suppl 1):R67–74.10.1093/hmg/ddl060Search in Google Scholar PubMed
 Tian R, Basu MK, Capriotti E. ContrastRank: a new method for ranking putative cancer driver genes and classification of tumor samples. Bioinformatics 2014;30:i572–8.10.1093/bioinformatics/btu466Search in Google Scholar PubMed PubMed Central
 Zhang C, Shen Y. A cell type-specific expression signature predicts haploinsufficient autism-susceptibility genes. Hum Mutat 2016;38:204–15.10.1002/humu.23147Search in Google Scholar PubMed PubMed Central
The online version of this article offers supplementary material (DOI: https://doi.org/10.1515/jib-2018-0069).
© 2019, Olga Zolotareva and Maren Kleine published by Walter de Gruyter GmbH, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 Public License.