Predicting disease genes using protein-protein interactions

Integrated analysis supports ATXN1 as a schizophrenia risk gene

Protein-protein interaction (PPI) is informative in identifying hidden disease risk genes that tend to interact with known risk genes usually working together in the same disease module. With the use of an integrated approach combining PPI information with pathway and expression analysis as well as genome-wide association study (GWAS), we intended to find new risk genes for schizophrenia (SCZ). We showed that ATXN1 was the only direct PPI partner of the know SCZ risk gene ZNF804A, and it also had direct PPIs with other 18 known SCZ risk genes. ATXN1 serves as one of the hub genes in the PPI network containing many known SCZ risk genes, and this network is significantly enriched for the MAPK signaling pathway. Further gene expression analysis indicated that ATXN1 is highly expressed in prefrontal cortex, and SCZ patients had significantly decreased expression compared with healthy controls. Finally, the published GWAS data supports an association of ATXN1 with SCZ as well as other psychiatric disorders though not reaching genome-wide significance. These convergent evidences support ATXN1 as a promising risk gene for SCZ, and the integrated approach serves as a useful tool for dissecting the genetic basis of psychiatric disorders.

The integrated landscape of causal genes and pathways in schizophrenia

Genome-wide association studies (GWAS) have identified more than 100 loci that show robust association with schizophrenia risk. However, due to the complexity of linkage disequilibrium and gene regulatory, it is challenging to pinpoint the causal genes at the risk loci and translate the genetic findings from GWAS into disease mechanism and clinical treatment. Here we systematically predicted the plausible candidate causal genes for schizophrenia at genome-wide level. We utilized different approaches and strategies to predict causal genes for schizophrenia, including Sherlock, SMR, DAPPLE, Prix Fixe, NetWAS, and DEPICT. By integrating the results from different prediction approaches, we identified six top candidates that represent promising causal genes for schizophrenia, including CNTN4, GATAD2A, GPM6A, MMP16, PSMA4, and TCF4. Besides, we also identified 35 additional high-confidence causal genes for schizophrenia. The identified causal genes showed distinct spatio-temporal expression patterns in developing and adult human brain. Cell-type-specific expression analysis indicated that the expression level of the predicted causal genes was significantly higher in neurons compared with oligodendrocytes and microglia (P < 0.05). We found that synaptic transmission-related genes were significantly enriched among the identified causal genes (P < 0.05), providing further support for the dysregulation of synaptic transmission in schizophrenia. Finally, we showed that the top six causal genes are dysregulated in schizophrenia cases compared with controls and knockdown of these genes impaired the proliferation of neuronal cells. Our study depicts the landscape of plausible schizophrenia causal genes for the first time. Further genetic and functional validation of these genes will provide mechanistic insights into schizophrenia pathogenesis and may facilitate to provide potential targets for future therapeutics and diagnostics.

Special Protein Molecules Computational Identification

Computational identification of special protein molecules is a key issue in understanding protein function. It can guide molecular experiments and help to save costs. I assessed 18 papers published in the special issue of Int. J. Mol. Sci., and also discussed the related works. The computational methods employed in this special issue focused on machine learning, network analysis, and molecular docking. New methods and new topics were also proposed. There were in addition several wet experiments, with proven results showing promise. I hope our special issue will help in protein molecules identification researches.

Special Protein Molecules Computational Identification

Computational identification of special protein molecules is a key issue in understanding protein function. It can guide molecular experiments and help to save costs. I assessed 18 papers published in the special issue of Int. J. Mol. Sci., and also discussed the related works. The computational methods employed in this special issue focused on machine learning, network analysis, and molecular docking. New methods and new topics were also proposed. There were in addition several wet experiments, with proven results showing promise. I hope our special issue will help in protein molecules identification researches.

The research on gene-disease association based on text-mining of PubMed

Background

The associations between genes and diseases are of critical significance in aspects of prevention, diagnosis and treatment. Although gene-disease relationships have been investigated extensively, much of the underpinnings of these associations are yet to be elucidated.

Methods

A novel method integrates MeSH database, term weight (TW), and co-occurrence methods to predict gene-disease associations based on the cosine similarity between gene vectors and disease vectors. Vectors are transformed from the texts of documents in the PubMed database according to the appearance and location of the gene or disease terms. The disease related text data has been optimized during the process of constructing vectors.

Results

The overall distribution of cosine similarity value was investigated. By using the gene-disease association data in OMIM database as golden standard, the performance of cosine similarity in predicting gene-disease linkage was evaluated. The effects of applying weight matrix, penalty weights for keywords (PWK), and normalization were also investigated. Finally, we demonstrated that our method outperforms heterogeneous network edge prediction (HNEP) in aspects of precision rate and recall rate.

Conclusions

Our method proposed in this paper is easy to be conducted and the results can be integrated with other models to improve the overall performance of gene-disease association predictions.

Population-dependent Intron Retention and DNA Methylation in Breast Cancer

Regulation of gene expression by DNA methylation in gene promoter regions is well studied; however, the effects of methylation in the gene body (exons and introns) on gene expression are comparatively understudied. Recently, hypermethylation has been implicated in the inclusion of alternatively spliced exons; moreover, exon recognition can be enhanced by recruiting the methyl-CpG-binding protein (MeCP2) to hypermethylated sites. This study examines whether the methylation status of an intron is correlated with how frequently the intron is retained during splicing using DNA methylation and RNA sequencing data from breast cancer tissue specimens in The Cancer Genome Atlas. Interestingly, hypomethylation of introns is correlated with higher levels of intron expression in mRNA and the methylation level of an intron is inversely correlated with its retention in mRNA from the gene in which it is located. Furthermore, significant population differences were observed in the methylation level of retained introns. In African-American donors, retained introns were not only less methylated compared to European-American donors, but also were more highly expressed. This underscores the need for understanding epigenetic differences in populations and their correlation with breast cancer is an important step toward achieving personalized cancer care.Implications: This research contributes to the understanding of how epigenetic markers in the gene body communicate with the transcriptional machinery to control transcript diversity and differential biological response to changes in methylation status could underlie some of the known, yet unexplained, disparities in certain breast cancer patient populations. Mol Cancer Res; 16(3); 461-9. ©2018 AACR.

ModuleDiscoverer: Identification of regulatory modules in protein-protein interaction networks

The identification of disease-associated modules based on protein-protein interaction networks (PPINs) and gene expression data has provided new insights into the mechanistic nature of diverse diseases. However, their identification is hampered by the detection of protein communities within large-scale, whole-genome PPINs. A presented successful strategy detects a PPIN's community structure based on the maximal clique enumeration problem (MCE), which is a non-deterministic polynomial time-hard problem. This renders the approach computationally challenging for large PPINs implying the need for new strategies. We present ModuleDiscoverer, a novel approach for the identification of regulatory modules from PPINs and gene expression data. Following the MCE-based approach, ModuleDiscoverer uses a randomization heuristic-based approximation of the community structure. Given a PPIN of Rattus norvegicus and public gene expression data, we identify the regulatory module underlying a rodent model of non-alcoholic steatohepatitis (NASH), a severe form of non-alcoholic fatty liver disease (NAFLD). The module is validated using single-nucleotide polymorphism (SNP) data from independent genome-wide association studies and gene enrichment tests. Based on gene enrichment tests, we find that ModuleDiscoverer performs comparably to three existing module-detecting algorithms. However, only our NASH-module is significantly enriched with genes linked to NAFLD-associated SNPs. ModuleDiscoverer is available at http://www.hki-jena.de/index.php/0/2/490 (Others/ModuleDiscoverer).

Revealing Alzheimer's disease genes spectrum in the whole-genome by machine learning

Background

Alzheimer’s disease (AD) is an important, progressive neurodegenerative disease, with a complex genetic architecture. A key goal of biomedical research is to seek out disease risk genes, and to elucidate the function of these risk genes in the development of disease. For this purpose, expanding the AD-associated gene set is necessary. In past research, the prediction methods for AD related genes has been limited in their exploration of the target genome regions. We here present a genome-wide method for AD candidate genes predictions.

Methods

We present a machine learning approach (SVM), based upon integrating gene expression data with human brain-specific gene network data, to discover the full spectrum of AD genes across the whole genome.

Results

We classified AD candidate genes with an accuracy and the area under the receiver operating characteristic (ROC) curve of 84.56% and 94%. Our approach provides a supplement for the spectrum of AD-associated genes extracted from more than 20,000 genes in a genome wide scale.

Conclusions

In this study, we have elucidated the whole-genome spectrum of AD, using a machine learning approach. Through this method, we expect for the candidate gene catalogue to provide a more comprehensive annotation of AD for researchers.

Electronic supplementary material

The online version of this article (10.1186/s12883-017-1010-3) contains supplementary material, which is available to authorized users.

Integrated rare variant-based risk gene prioritization in disease case-control sequencing studies

Rare variants of major effect play an important role in human complex diseases and can be discovered by sequencing-based genome-wide association studies. Here, we introduce an integrated approach that combines the rare variant association test with gene network and phenotype information to identify risk genes implicated by rare variants for human complex diseases. Our data integration method follows a 'discovery-driven' strategy without relying on prior knowledge about the disease and thus maintains the unbiased character of genome-wide association studies. Simulations reveal that our method can outperform a widely-used rare variant association test method by 2 to 3 times. In a case study of a small disease cohort, we uncovered putative risk genes and the corresponding rare variants that may act as genetic modifiers of congenital heart disease in 22q11.2 deletion syndrome patients. These variants were missed by a conventional approach that relied on the rare variant association test alone.

Case-control sequencing studies are a promising design to uncover risk genes of human complex diseases implicated by rare variants. The recent development of different types of rare variant association tests has improved the statistical power to identify disease genes that harbor risk rare variants. However, none of the recent sequencing-based genome-wide association studies identified robust disease association of rare variants or genes based on them. Due to limited sample sizes that can be feasibly achieved in real applications, current rare variant association tests can only generate marginal association signals for most risk genes. Here we proposed an integrated method that combined association signals with orthogonal biological evidence to uncover risk genes in sequencing studies. Designed to address the lack-of-power issue, our method was shown to effectively uncover risk genes with marginal association signals in data simulation. Indeed, in a real application demonstrated in our case study our method disclosed important risk genes of congenital heart disease in 22q11.2 deletion syndrome that were missed by the previous study.

Network-based method for mining novel HPV infection related genes using random walk with restart algorithm

The human papillomavirus (HPV), a common virus that infects the reproductive tract, may lead to malignant changes within the infection area in certain cases and is directly associated with such cancers as cervical cancer, anal cancer, and vaginal cancer. Identification of novel HPV infection related genes can lead to a better understanding of the specific signal pathways and cellular processes related to HPV infection, providing information for the development of more efficient therapies. In this study, several novel HPV infection related genes were predicted by a computation method based on the known genes involved in HPV infection from HPVbase. This method applied the algorithm of random walk with restart (RWR) to a protein-protein interaction (PPI) network. The candidate genes were further filtered by the permutation and association tests. These steps eliminated genes occupying special positions in the PPI network and selected key genes with strong associations to known HPV infection related genes based on the interaction confidence and functional similarity obtained from published databases, such as STRING, gene ontology (GO) terms and KEGG pathways. Our study identified 104 novel HPV infection related genes, a number of which were confirmed to relate to the infection processes and complications of HPV infection, as reported in the literature. These results demonstrate the reliability of our method in identifying HPV infection related genes. This article is part of a Special Issue entitled: Accelerating Precision Medicine through Genetic and Genomic Big Data Analysis edited by Yudong Cai & Tao Huang.

Functional diversity of topological modules in human protein-protein interaction networks

A large-scale molecular interaction network of protein-protein interactions (PPIs) enables the automatic detection of molecular functional modules through a computational approach. However, the functional modules that are typically detected by topological community detection algorithms may be diverse in functional homogeneity and are empirically considered to be default functional modules. Thus, a significant challenge that has been described but not elucidated is investigating the relationship between topological modules and functional modules. We systematically investigated this issue by initially using seven widely used community detection algorithms to partition the PPI network into communities. Four homogeneity measures were subsequently implemented to evaluate the functional homogeneity of protein community. We determined that a significant portion of topological modules with heterogeneous functionality exists and should be further investigated; moreover, these findings indicated that topologically based functional module detection approaches must be reconsidered. Furthermore, we found that the functional homogeneity of topological modules is positively correlated with their edge densities, degree of association with diseases and general Gene Ontology (GO) terms. Thus, topologically based module detection approaches should be used with caution in the identification of functional modules with high homogeneity

Discovery of Bladder Cancer-related Genes Using Integrative Heterogeneous Network Modeling of Multi-omics Data

In human health, a fundamental challenge is the identification of disease-related genes. Bladder cancer (BC) is a worldwide malignant tumor, which has resulted in 170,000 deaths in 2010 up from 114,000 in 1990. Moreover, with the emergence of multi-omics data, more comprehensive analysis of human diseases become possible. In this study, we propose a multi-step approach for the identification of BC-related genes by using integrative Heterogeneous Network Modeling of Multi-Omics data (iHNMMO). The heterogeneous network model properly and comprehensively reflects the multiple kinds of relationships between genes in the multi-omics data of BC, including general relationships, unique relationships under BC condition, correlational relationships within each omics and regulatory relationships between different omics. Besides, a network-based propagation algorithm with resistance is utilized to quantize the relationships between genes and BC precisely. The results of comprehensive performance evaluation suggest that iHNMMO significantly outperforms other approaches. Moreover, further analysis suggests that the top ranked genes may be functionally implicated in BC, which also confirms the superiority of iHNMMO. In summary, this study shows that disease-related genes can be better identified through reasonable integration of multi-omics data.

Impaired AMPA receptor trafficking by a double knockout of zebrafish olfactomedin1a/b

The olfm1a and olfm1b genes in zebrafish encode conserved secreted glycoproteins. These genes are preferentially expressed in the brain and retina starting from 16 h post-fertilization until adulthood. Functions of the Olfm1 gene is still unclear. Here, we produced and analyzed a null zebrafish mutant of both olfm1a and olfm1b genes (olfm1 null). olfm1 null fish were born at a normal Mendelian ratio and showed normal body shape and fertility as well as no visible defects from larval stages to adult. Olfm1 proteins were preferentially localized in the synaptosomes of the adult brain. Olfm1 co-immunoprecipitated with GluR2 and soluble NSF attachment protein receptor complexes indicating participation of Olfm1 in both pre- and post-synaptic events. Phosphorylation of GluR2 was not changed while palmitoylation of GluR2 was decreased in the brain synaptosomal membrane fraction of olfm1 null compared with wt fish. The levels of GluR2, SNAP25, flotillin1, and VAMP2 were markedly reduced in the synaptic microdomain of olfm1 null brain compared with wt. The internalization of GluR2 in retinal cells and the localization of VAMP2 in brain synaptosome were modified by olfm1 null mutation. This indicates that Olfm1 may regulate receptor trafficking from the intracellular compartments to the synaptic membrane microdomain, partly through the alteration of post-translational GluR2 modifications such as palmitoylation. Olfm1 may be considered a novel regulator of the composition and function of the α-amino-3-hydroxy-5-methylisoxazole-4-propionate receptor complex.

Effect of Aggregation Operators on Network-Based Disease Gene Prioritization: A Case Study on Blood Disorders

Owing to the innate noise in the biological data sources, a single source or a single measure do not suffice for an effective disease gene prioritization. So, the integration of multiple data sources or aggregation of multiple measures is the need of the hour. The aggregation operators combine multiple related data values to a single value such that the combined value has the effect of all the individual values. In this paper, an attempt has been made for applying the fuzzy aggregation on the network-based disease gene prioritization and investigate its effect under noise conditions. This study has been conducted for a set of 15 blood disorders by fusing four different network measures, computed from the protein interaction network, using a selected set of aggregation operators and ranking the genes on the basis of the aggregated value. The aggregation operator-based rankings have been compared with the "Random walk with restart" gene prioritization method. The impact of noise has also been investigated by adding varying proportions of noise to the seed set. The results reveal that for all the selected blood disorders, the Mean of Maximal operator has relatively outperformed the other aggregation operators for noisy as well as non-noisy data.

Significance and Functional Similarity for Identification of Disease Genes

One of the most significant research issues in functional genomics is insilico identification of disease related genes. In this regard, the paper presents a new gene selection algorithm, termed as SiFS, for identification of disease genes. It integrates the information obtained from interaction network of proteins and gene expression profiles. The proposed SiFS algorithm culls out a subset of genes from microarray data as disease genes by maximizing both significance and functional similarity of the selected gene subset. Based on the gene expression profiles, the significance of a gene with respect to another gene is computed using mutual information. On the other hand, a new measure of similarity is introduced to compute the functional similarity between two genes. Information derived from the protein-protein interaction network forms the basis of the proposed SiFS algorithm. The performance of the proposed gene selection algorithm and new similarity measure, is compared with that of other related methods and similarity measures, using several cancer microarray data sets.

Myocilin Regulates Metalloprotease 2 Activity Through Interaction With TIMP3

Purpose

To elucidate functions of wild-type myocilin, a secreted glycoprotein associated with glaucoma.

Methods

Lysates of mouse eyes were used for immunoprecipitation with affinity-purified antibodies against mouse myocilin. Shotgun proteomic analysis was used for the identification of proteins interacting with myocilin. Colocalization of myocilin and tissue inhibitor of metalloproteinases 3 (TIMP3) in different eye structures was investigated by a multiplex fluorescent in situ hybridization and immunofluorescent labeling with subsequent confocal microscopy. Matrix metalloproteinase 2 (MMP2) activity assay was used to test effects of myocilin on TIMP3 inhibitory action.

Results

TIMP3 was identified by a shotgun proteomic analysis as a protein that was coimmunoprecipitated with myocilin from eye lysates of wild-type and transgenic mice expressing elevated levels of mouse myocilin but not from lysates of transgenic mice expressing mutated mouse myocilin. Interaction of myocilin and TIMP3 was confirmed by coimmunoprecipitation of myocilin and TIMP3 from HEK293 cells transiently transfected with cDNAs encoding these proteins. The olfactomedin domain of myocilin is essential for interaction with TIMP3. In the eye, the main sites of myocilin and TIMP3 colocalization are the trabecular meshwork, sclera, and choroid. Using purified proteins, it has been shown that myocilin markedly enhanced the inhibitory activity of TIMP3 toward MMP2.

Conclusions

Myocilin may serve as a modulator of TIMP3 activity via interactions with the myocilin olfactomedin domain. Our data imply that in the case of MYOCILIN null or some glaucoma-causing mutations, inhibitory activity of TIMP3 toward MMP2 might be reduced, mimicking deleterious mutations in the TIMP3 gene.

Constructing an integrated gene similarity network for the identification of disease genes

BACKGROUND

Discovering novel genes that are involved human diseases is a challenging task in biomedical research. In recent years, several computational approaches have been proposed to prioritize candidate disease genes. Most of these methods are mainly based on protein-protein interaction (PPI) networks. However, since these PPI networks contain false positives and only cover less half of known human genes, their reliability and coverage are very low. Therefore, it is highly necessary to fuse multiple genomic data to construct a credible gene similarity network and then infer disease genes on the whole genomic scale.

RESULTS

We proposed a novel method, named RWRB, to infer causal genes of interested diseases. First, we construct five individual gene (protein) similarity networks based on multiple genomic data of human genes. Then, an integrated gene similarity network (IGSN) is reconstructed based on similarity network fusion (SNF) method. Finally, we employee the random walk with restart algorithm on the phenotype-gene bilayer network, which combines phenotype similarity network, IGSN as well as phenotype-gene association network, to prioritize candidate disease genes. We investigate the effectiveness of RWRB through leave-one-out cross-validation methods in inferring phenotype-gene relationships. Results show that RWRB is more accurate than state-of-the-art methods on most evaluation metrics. Further analysis shows that the success of RWRB is benefited from IGSN which has a wider coverage and higher reliability comparing with current PPI networks. Moreover, we conduct a comprehensive case study for Alzheimer's disease and predict some novel disease genes that supported by literature.

CONCLUSIONS

RWRB is an effective and reliable algorithm in prioritizing candidate disease genes on the genomic scale. Software and supplementary information are available at http://nclab.hit.edu.cn/~tianzhen/RWRB/ .

A network-based method using a random walk with restart algorithm and screening tests to identify novel genes associated with Menière's disease

As a chronic illness derived from hair cells of the inner ear, Menière’s disease (MD) negatively influences the quality of life of individuals and leads to a number of symptoms, such as dizziness, temporary hearing loss, and tinnitus. The complete identification of novel genes related to MD would help elucidate its underlying pathological mechanisms and improve its diagnosis and treatment. In this study, a network-based method was developed to identify novel MD-related genes based on known MD-related genes. A human protein-protein interaction (PPI) network was constructed using the PPI information reported in the STRING database. A classic ranking algorithm, the random walk with restart (RWR) algorithm, was employed to search for novel genes using known genes as seed nodes. To make the identified genes more reliable, a series of screening tests, including a permutation test, an interaction test and an enrichment test, were designed to select essential genes from those obtained by the RWR algorithm. As a result, several inferred genes, such as CD4, NOTCH2 and IL6, were discovered. Finally, a detailed biological analysis was performed on fifteen of the important inferred genes, which indicated their strong associations with MD.

Network propagation: a universal amplifier of genetic associations

Biological networks are powerful resources for the discovery of genes and genetic modules that drive disease. Fundamental to network analysis is the concept that genes underlying the same phenotype tend to interact; this principle can be used to combine and to amplify signals from individual genes. Recently, numerous bioinformatic techniques have been proposed for genetic analysis using networks, based on random walks, information diffusion and electrical resistance. These approaches have been applied successfully to identify disease genes, genetic modules and drug targets. In fact, all these approaches are variations of a unifying mathematical machinery - network propagation - suggesting that it is a powerful data transformation method of broad utility in genetic research.

Identification of oral cancer related candidate genes by integrating protein-protein interactions, gene ontology, pathway analysis and immunohistochemistry

In the recent years, bioinformatics methods have been reported with a high degree of success for candidate gene identification. In this milieu, we have used an integrated bioinformatics approach assimilating information from gene ontologies (GO), protein–protein interaction (PPI) and network analysis to predict candidate genes related to oral squamous cell carcinoma (OSCC). A total of 40973 PPIs were considered for 4704 cancer-related genes to construct human cancer gene network (HCGN). The importance of each node was measured in HCGN by ten different centrality measures. We have shown that the top ranking genes are related to a significantly higher number of diseases as compared to other genes in HCGN. A total of 39 candidate oral cancer target genes were predicted by combining top ranked genes and the genes corresponding to significantly enriched oral cancer related GO terms. Initial verification using literature and available experimental data indicated that 29 genes were related with OSCC. A detailed pathway analysis led us to propose a role for the selected candidate genes in the invasion and metastasis in OSCC. We further validated our predictions using immunohistochemistry (IHC) and found that the gene FLNA was upregulated while the genes ARRB1 and HTT were downregulated in the OSCC tissue samples.

Common variants near IKZF1 are associated with primary Sjögren's syndrome in Han Chinese

Primary Sjögren's syndrome (pSS) is a systematic autoimmune disease with evidence of genetic predisposition. The IKZF1 (IKAROS family zinc finger 1 (Ikaros)) gene is located at 7p12.2, encodes a transcription factor related to chromatin remodeling, regulates lymphocyte differentiation, and has been reported to be associated with some autoimmune diseases. However, there have been no reports of an association between IKZF1 and pSS. To investigate the possibility of an association between the IKZF1 locus and pSS, we selected two single nucleotide polymorphisms (SNPs) in the IKZF1 locus, rs4917129 and rs4917014, based on a detailed analysis of genome-wide association study (GWAS) data and performed genotyping in 665 Han Chinese pSS patients and 863 healthy controls. The results of an association test showed significant association signals (rs4917129: P-value = 5.5e-4, OR (odds ratio) = 0.72, 95% CI (confidence interval) = 0.60-0.87; rs4917014: P-value = 1.2e-3, OR = 0.76, 95% CI = 0.64-0.89). A meta-analysis that combined the above results with data from previous GWAS, further confirmed these associations (rs4917129: Pmeta = 4.24e-8, ORmeta = 0.70, 95% CI = 0.61-0.79; rs4917014: Pmeta = 6.0e-8, ORmeta = 0.72, 95% CI = 0.64-0.81). A bioinformatics analysis indicated that both SNPs were located in a putative enhancer area in immune-related cell lines and tissues. A protein-protein interaction analysis found that IKZF1, together with GTF2I (an SS susceptibility gene newly identified through GWAS), could interact with histone deacetylase family proteins. In summary, this is the first study to report an association between IKZF1 and SS in Han Chinese.

Determination of Genes Related to Uveitis by Utilization of the Random Walk with Restart Algorithm on a Protein-Protein Interaction Network

Uveitis, defined as inflammation of the uveal tract, may cause blindness in both young and middle-aged people. Approximately 10–15% of blindness in the West is caused by uveitis. Therefore, a comprehensive investigation to determine the disease pathogenesis is urgent, as it will thus be possible to design effective treatments. Identification of the disease genes that cause uveitis is an important requirement to achieve this goal. To begin to answer this question, in this study, a computational method was proposed to identify novel uveitis-related genes. This method was executed on a large protein–protein interaction network and employed a popular ranking algorithm, the Random Walk with Restart (RWR) algorithm. To improve the utility of the method, a permutation test and a procedure for selecting core genes were added, which helped to exclude false discoveries and select the most important candidate genes. The five-fold cross-validation was adopted to evaluate the method, yielding the average F1-measure of 0.189. In addition, we compared our method with a classic GBA-based method to further indicate its utility. Based on our method, 56 putative genes were chosen for further assessment. We have determined that several of these genes (e.g., CCL4, Jun, and MMP9) are likely to be important for the pathogenesis of uveitis.

A Heterogeneous Network Based Method for Identifying GBM-Related Genes by Integrating Multi-Dimensional Data

The emergence of multi-dimensional data offers opportunities for more comprehensive analysis of the molecular characteristics of human diseases and therefore improving diagnosis, treatment, and prevention. In this study, we proposed a heterogeneous network based method by integrating multi-dimensional data (HNMD) to identify GBM-related genes. The novelty of the method lies in that the multi-dimensional data of GBM from TCGA dataset that provide comprehensive information of genes, are combined with protein-protein interactions to construct a weighted heterogeneous network, which reflects both the general and disease-specific relationships between genes. In addition, a propagation algorithm with resistance is introduced to precisely score and rank GBM-related genes. The results of comprehensive performance evaluation show that the proposed method significantly outperforms the network based methods with single-dimensional data and other existing approaches. Subsequent analysis of the top ranked genes suggests they may be functionally implicated in GBM, which further corroborates the superiority of the proposed method. The source code and the results of HNMD can be downloaded from the following URL: http://bioinformatics.ustc.edu.cn/hnmd/ .

Integrated network analysis reveals potentially novel molecular mechanisms and therapeutic targets of refractory epilepsies

Epilepsy is a complex neurological disorder and a significant health problem. The pathogenesis of epilepsy remains obscure in a significant number of patients and the current treatment options are not adequate in about a third of individuals which were known as refractory epilepsies (RE). Network medicine provides an effective approach for studying the molecular mechanisms underlying complex diseases. Here we integrated 1876 disease-gene associations of RE and located those genes to human protein-protein interaction (PPI) network to obtain 42 significant RE-associated disease modules. The functional analysis of these disease modules showed novel molecular pathological mechanisms of RE, such as the novel enriched pathways (e.g., "presynaptic nicotinic acetylcholine receptors", "signaling by insulin receptor"). Further analysis on the relationships between current drug targets and the RE-related disease genes showed the rational mechanisms of most antiepileptic drugs. In addition, we detected ten potential novel drug targets (e.g., KCNA1, KCNA4-6, KCNC3, KCND2, KCNMA1, CAMK2G, CACNB4 and GRM1) located in three RE related disease modules, which might provide novel insights into the new drug discovery for RE therapy.

Sex-Specific Associations between Particulate Matter Exposure and Gene Expression in Independent Discovery and Validation Cohorts of Middle-Aged Men and Women

BACKGROUND

Particulate matter (PM) exposure leads to premature death, mainly due to respiratory and cardiovascular diseases.

OBJECTIVES

Identification of transcriptomic biomarkers of air pollution exposure and effect in a healthy adult population.

METHODS

Microarray analyses were performed in 98 healthy volunteers (48 men, 50 women). The expression of eight sex-specific candidate biomarker genes (significantly associated with PM₁₀ in the discovery cohort and with a reported link to air pollution-related disease) was measured with qPCR in an independent validation cohort (75 men, 94 women). Pathway analysis was performed using Gene Set Enrichment Analysis. Average daily PM_2.5 and PM₁₀ exposures over 2-years were estimated for each participant's residential address using spatiotemporal interpolation in combination with a dispersion model.

RESULTS

Average long-term PM₁₀ was 25.9 (± 5.4) and 23.7 (± 2.3) μg/m³ in the discovery and validation cohorts, respectively. In discovery analysis, associations between PM₁₀ and the expression of individual genes differed by sex. In the validation cohort, long-term PM₁₀ was associated with the expression of DNAJB5 and EAPP in men and ARHGAP4 (p = 0.053) in women. AKAP6 and LIMK1 were significantly associated with PM₁₀ in women, although associations differed in direction between the discovery and validation cohorts. Expression of the eight candidate genes in the discovery cohort differentiated between validation cohort participants with high versus low PM₁₀ exposure (area under the receiver operating curve = 0.92; 95% CI: 0.85, 1.00; p = 0.0002 in men, 0.86; 95% CI: 0.76, 0.96; p = 0.004 in women).

CONCLUSIONS

Expression of the sex-specific candidate genes identified in the discovery population predicted PM₁₀ exposure in an independent cohort of adults from the same area. Confirmation in other populations may further support this as a new approach for exposure assessment, and may contribute to the discovery of molecular mechanisms for PM-induced health effects.

Identification of Candidate Genes Related to Inflammatory Bowel Disease Using Minimum Redundancy Maximum Relevance, Incremental Feature Selection, and the Shortest-Path Approach

Identification of disease genes is a hot topic in biomedicine and genomics. However, it is a challenging problem because of the complexity of diseases. Inflammatory bowel disease (IBD) is an idiopathic disease caused by a dysregulated immune response to host intestinal microflora. It has been proven to be associated with the development of intestinal malignancies. Although the specific pathological characteristics and genetic background of IBD have been partially revealed, it is still an overdetermined disease and the blueprint of all genetic variants still needs to be improved. In this study, a novel computational method was built to identify genes related to IBD. Samples from two subtypes of IBD (ulcerative colitis and Crohn's disease) and normal samples were employed. By analyzing the gene expression profiles of these samples using minimum redundancy maximum relevance and incremental feature selection, 21 genes were obtained that could effectively distinguish samples from the two subtypes of IBD and the normal samples. Then, the shortest-path approach was used to search for an additional 20 genes in a large network constructed using protein-protein interactions based on the above-mentioned 21 genes. Analyses of the 41 genes obtained indicate that they are closely associated with this disease.

Identifying and Analyzing Novel Epilepsy-Related Genes Using Random Walk with Restart Algorithm

As a pathological condition, epilepsy is caused by abnormal neuronal discharge in brain which will temporarily disrupt the cerebral functions. Epilepsy is a chronic disease which occurs in all ages and would seriously affect patients' personal lives. Thus, it is highly required to develop effective medicines or instruments to treat the disease. Identifying epilepsy-related genes is essential in order to understand and treat the disease because the corresponding proteins encoded by the epilepsy-related genes are candidates of the potential drug targets. In this study, a pioneering computational workflow was proposed to predict novel epilepsy-related genes using the random walk with restart (RWR) algorithm. As reported in the literature RWR algorithm often produces a number of false positive genes, and in this study a permutation test and functional association tests were implemented to filter the genes identified by RWR algorithm, which greatly reduce the number of suspected genes and result in only thirty-three novel epilepsy genes. Finally, these novel genes were analyzed based upon some recently published literatures. Our findings implicate that all novel genes were closely related to epilepsy. It is believed that the proposed workflow can also be applied to identify genes related to other diseases and deepen our understanding of the mechanisms of these diseases.

Predicting disease-related genes using integrated biomedical networks

Background

Identifying the genes associated to human diseases is crucial for disease diagnosis and drug design. Computational approaches, esp. the network-based approaches, have been recently developed to identify disease-related genes effectively from the existing biomedical networks. Meanwhile, the advance in biotechnology enables researchers to produce multi-omics data, enriching our understanding on human diseases, and revealing the complex relationships between genes and diseases. However, none of the existing computational approaches is able to integrate the huge amount of omics data into a weighted integrated network and utilize it to enhance disease related gene discovery.

Results

We propose a new network-based disease gene prediction method called SLN-SRW (Simplified Laplacian Normalization-Supervised Random Walk) to generate and model the edge weights of a new biomedical network that integrates biomedical data from heterogeneous sources, thus far enhancing the disease related gene discovery.

Conclusions

The experiment results show that SLN-SRW significantly improves the performance of disease gene prediction on both the real and the synthetic data sets.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-016-3263-4) contains supplementary material, which is available to authorized users.

Biomarker identification of hepatocellular carcinoma using a methodical literature mining strategy

Hepatocellular carcinoma (HCC), one of the most common causes of cancer-related deaths, carries a 5-year survival rate of 18%, underscoring the need for robust biomarkers. In spite of the increased availability of HCC related literatures, many of the promising biomarkers reported have not been validated for clinical use. To narrow down the wide range of possible biomarkers for further clinical validation, bioinformaticians need to sort them out using information provided in published works. Biomedical text mining is an automated way to obtain information of interest within the massive collection of biomedical knowledge, thus enabling extraction of data for biomarkers associated with certain diseases. This method can significantly reduce both the time and effort spent on studying important maladies such as liver diseases. Herein, we report a text mining-aided curation pipeline to identify potential biomarkers for liver cancer. The curation pipeline integrates PubMed E-Utilities to collect abstracts from PubMed and recognize several types of named entities by machine learning-based and pattern-based methods. Genes/proteins from evidential sentences were classified as candidate biomarkers using a convolutional neural network. Lastly, extracted biomarkers were ranked depending on several criteria, such as the frequency of keywords and articles and the journal impact factor, and then integrated into a meaningful list for bioinformaticians. Based on the developed pipeline, we constructed MarkerHub, which contains 2128 candidate biomarkers extracted from PubMed publications from 2008 to 2017. Database URL: http://markerhub.iis.sinica.edu.tw.

Exploring the molecular basis of age-related disease comorbidities using a multi-omics graphical model

Although association studies have unveiled numerous correlations of biochemical markers with age and age-related diseases, we still lack an understanding of their mutual dependencies. To find molecular pathways that underlie age-related diseases as well as their comorbidities, we integrated aging markers from four different high-throughput omics datasets, namely epigenomics, transcriptomics, glycomics and metabolomics, with a comprehensive set of disease phenotypes from 510 participants of the TwinsUK cohort. We used graphical random forests to assess conditional dependencies between omics markers and phenotypes while eliminating mediated associations. Applying this novel approach for multi-omics data integration yields a model consisting of seven modules that represent distinct aspects of aging. These modules are connected by hubs that potentially trigger comorbidities of age-related diseases. As an example, we identified urate as one of these key players mediating the comorbidity of renal disease with body composition and obesity. Body composition variables are in turn associated with inflammatory IgG markers, mediated by the expression of the hormone oxytocin. Thus, oxytocin potentially contributes to the development of chronic low-grade inflammation, which often accompanies obesity. Our multi-omics graphical model demonstrates the interconnectivity of age-related diseases and highlights molecular markers of the aging process that might drive disease comorbidities.

Network-Guided Key Gene Discovery for a Given Cellular Process

Identification of key genes for a given physiological or pathological process is an essential but still very challenging task for the entire biomedical research community. Statistics-based approaches, such as genome-wide association study (GWAS)- or quantitative trait locus (QTL)-related analysis have already made enormous contributions to identifying key genes associated with a given disease or phenotype, the success of which is however very much dependent on a huge number of samples. Recent advances in network biology, especially network inference directly from genome-scale data and the following-up network analysis, opens up new avenues to predict key genes driving a given biological process or cellular function. Here we review and compare the current approaches in predicting key genes, which have no chances to stand out by classic differential expression analysis, from gene-regulatory, protein-protein interaction, or gene expression correlation networks. We elaborate these network-based approaches mainly in the context of immunology and infection, and urge more usage of correlation network-based predictions. Such network-based key gene discovery approaches driven by information-enriched 'omics' data should be very useful for systematic key gene discoveries for any given biochemical process or cellular function, and also valuable for novel drug target discovery and novel diagnostic, prognostic and therapeutic-efficiency marker prediction for a specific disease or disorder.

Focus on the emerging new fields of Network Physiology and Network Medicine

Despite the vast progress and achievements in systems biology and integrative physiology in the last decades, there is still a significant gap in understanding the mechanisms through which (i) genomic, proteomic and metabolic factors and signaling pathways impact vertical processes across cells, tissues and organs leading to the expression of different disease phenotypes and influence the functional and clinical associations between diseases, and (ii) how diverse physiological systems and organs coordinate their functions over a broad range of space and time scales and horizontally integrate to generate distinct physiologic states at the organism level. Two emerging fields, network medicine and network physiology, aim to address these fundamental questions. Novel concepts and approaches derived from recent advances in network theory, coupled dynamical systems, statistical and computational physics show promise to provide new insights into the complexity of physiological structure and function in health and disease, bridging the genetic and sub-cellular level with inter-cellular interactions and communications among integrated organ systems and sub-systems. These advances form first building blocks in the methodological formalism and theoretical framework necessary to address fundamental problems and challenges in physiology and medicine. This 'focus on' issue contains 26 articles representing state-of-the-art contributions covering diverse systems from the sub-cellular to the organism level where physicists have key role in laying the foundations of these new fields.

Predicting diabetes mellitus genes via protein-protein interaction and protein subcellular localization information

BACKGROUND

Diabetes mellitus characterized by hyperglycemia as a result of insufficient production of or reduced sensitivity to insulin poses a growing threat to the health of people. It is a heterogeneous disorder with multiple etiologies consisting of type 1 diabetes, type 2 diabetes, gestational diabetes and so on. Diabetes-associated protein/gene prediction is a key step to understand the cellular mechanisms related to diabetes mellitus. Compared with experimental methods, computational predictions of candidate proteins/genes are cheaper and more effortless. Protein-protein interaction (PPI) data produced by the high-throughput technology have been used to prioritize candidate disease genes/proteins. However, the false interactions in the PPI data seriously hurt computational methods performance. In order to address that particular question, new methods are developed to identify candidate disease genes/proteins via integrating biological data from other sources.

RESULTS

In this study, a new framework called PDMG is proposed to predict candidate disease genes/proteins. First, the weighted networks are building in terms of the combination of the subcellular localization information and PPI data. To form the weighted networks, the importance of each compartment is evaluated based on the number of interacted proteins in this compartment. This is because the very different roles played by different compartments in cell activities. Besides, some compartments are more important than others. Based on the evaluated compartments, the interactions between proteins are scored and the weighted PPI networks are constructed. Second, the known disease genes are extracted from OMIM database as the seed genes to expand disease-specific networks based on the weighted networks. Third, the weighted values between a protein and its neighbors in the disease-related networks are added together and the sum is as the score of the protein. Last but not least, the proteins are ranked based on descending order of their scores. The candidate proteins in the top are considered to be associated with the diseases and are potential disease-related proteins. Various types of data, such as type 2 diabetes-associated genes, subcellular localizations and protein interactions, are used to test PDMG method.

CONCLUSIONS

The results show that the proteins/genes functionally exerting a direct influence over diabetes are consistently placed at the head of the queue. PDMG expands and ranks 445 candidate proteins from the seed set including original 27 type 2 diabetes proteins. Out of the top 27 proteins, 14 proteins are the real type 2 diabetes proteins. The literature extracted from the PubMed database has proved that, out of 13 novel proteins, 8 proteins are associated with diabetes.

Networks and games for precision medicine

Recent advances in omics technologies provide the leverage for the emergence of precision medicine that aims at personalizing therapy to patient. In this undertaking, computational methods play a central role for assisting physicians in their clinical decision-making by combining data analysis and systems biology modelling. Complex diseases such as cancer or diabetes arise from the intricate interplay of various biological molecules. Therefore, assessing drug efficiency requires to study the effects of elementary perturbations caused by diseases on relevant biological networks. In this paper, we propose a computational framework called Network-Action Game applied to best drug selection problem combining Game Theory and discrete models of dynamics (Boolean networks). Decision-making is modelled using Game Theory that defines the process of drug selection among alternative possibilities, while Boolean networks are used to model the effects of the interplay between disease and drugs actions on the patient's molecular system. The actions/strategies of disease and drugs are focused on arc alterations of the interactome. The efficiency of this framework has been evaluated for drug prediction on a model of breast cancer signalling.

A Comprehensive Evaluation of Disease Phenotype Networks for Gene Prioritization

Identification of disease-causing genes is a fundamental challenge for human health studies. The phenotypic similarity among diseases may reflect the interactions at the molecular level, and phenotype comparison can be used to predict disease candidate genes. Online Mendelian Inheritance in Man (OMIM) is a database of human genetic diseases and related genes that has become an authoritative source of disease phenotypes. However, disease phenotypes have been described by free text; thus, standardization of phenotypic descriptions is needed before diseases can be compared. Several disease phenotype networks have been established in OMIM using different standardization methods. Two of these networks are important for phenotypic similarity analysis: the first and most commonly used network (mimMiner) is standardized by medical subject heading, and the other network (resnikHPO) is the first to be standardized by human phenotype ontology. This paper comprehensively evaluates for the first time the accuracy of these two networks in gene prioritization based on protein–protein interactions using large-scale, leave-one-out cross-validation experiments. The results show that both networks can effectively prioritize disease-causing genes, and the approach that relates two diseases using a logistic function improves prioritization performance. Tanimoto, one of four methods for normalizing resnikHPO, generates a symmetric network and it performs similarly to mimMiner. Furthermore, an integration of these two networks outperforms either network alone in gene prioritization, indicating that these two disease networks are complementary.

Integrative analysis of cancer genes in a functional interactome

The post-genomic era has resulted in the accumulation of high-throughput cancer data from a vast array of genomic technologies including next-generation sequencing and microarray. As such, the large amounts of germline variant and somatic mutation data that have been generated from GWAS and sequencing projects, respectively, show great promise in providing a systems-level view of these genetic aberrations. In this study, we analyze publicly available GWAS, somatic mutation, and drug target data derived from large databanks using a network-based approach that incorporates directed edge information under a randomized network hypothesis testing procedure. We show that these three classes of disease-associated nodes exhibit non-random topological characteristics in the context of a functional interactome. Specifically, we show that drug targets tend to lie upstream of somatic mutations and disease susceptibility germline variants. In addition, we introduce a new approach to measuring hierarchy between drug targets, somatic mutants, and disease susceptibility genes by utilizing directionality and path length information. Overall, our results provide new insight into the intrinsic relationships between these node classes that broaden our understanding of cancer. In addition, our results align with current knowledge on the therapeutic actionability of GWAS and somatic mutant nodes, while demonstrating relationships between node classes from a global network perspective.

Network analysis of genes and their association with diseases

A plethora of network-based approaches within the Systems Biology universe have been applied, to date, to investigate the underlying molecular mechanisms of various human diseases. In the present study, we perform a bipartite, topological and clustering graph analysis in order to gain a better understanding of the relationships between human genetic diseases and the relationships between the genes that are implicated in them. For this purpose, disease-disease and gene-gene networks were constructed from combined gene-disease association networks. The latter, were created by collecting and integrating data from three diverse resources, each one with different content covering from rare monogenic disorders to common complex diseases. This data pluralism enabled us to uncover important associations between diseases with unrelated phenotypic manifestations but with common genetic origin. For our analysis, the topological attributes and the functional implications of the individual networks were taken into account and are shortly discussed. We believe that some observations of this study could advance our understanding regarding the etiology of a disease with distinct pathological manifestations, and simultaneously provide the springboard for the development of preventive and therapeutic strategies and its underlying genetic mechanisms.

IDENTIFICATION OF NOVEL GENES RELATED TO DIABETIC RETINOPATHY USING PROTEIN–PROTEIN INTERACTION NETWORK AND GENE ONTOLOGIES

Diabetic retinopathy is the most common cause of blindness, associated with many biochemical pathways mediated by several genes and proteins. Disease gene identification can be achieved through several approaches but still it is a challenging task. This study, aimed to find out the novel genes associated with diabetic retinopathy. In this study, all the well-known genes associated with diabetic retinopathy were collected from databases and the protein interaction partners were identified. The interacting candidate genes were chosen by chromosomal locations, sharing with disease genes. The protein–protein interaction network was constructed and the key nodes (genes) were identified by degree, betweenness centrality, closeness centrality and eccentricity centrality. Further, the ontological terms, molecular function, biological process and cellular components were related with that of the disease genes with p-value [Formula: see text]. The genes UBC, FOS, ITGB1, FOXA2, CCND1, FOSL1, RXRA and NCAM1 were identified as potential genes associated with diabetic retinopathy. The molecular functions of these genes include protein binding, receptor activity, receptor binding, oxidoreductase activity, protein kinase activity, serine-type peptidase activity and growth factor. Many of the identified genes were clinically related as evidence by the literature.

Integrated Genomic and Network-Based Analyses of Complex Diseases and Human Disease Network

A disease phenotype generally reflects various pathobiological processes that interact in a complex network. The highly interconnected nature of the human protein interaction network (interactome) indicates that, at the molecular level, it is difficult to consider diseases as being independent of one another. Recently, genome-wide molecular measurements, data mining and bioinformatics approaches have provided the means to explore human diseases from a molecular basis. The exploration of diseases and a system of disease relationships based on the integration of genome-wide molecular data with the human interactome could offer a powerful perspective for understanding the molecular architecture of diseases. Recently, subnetwork markers have proven to be more robust and reliable than individual biomarker genes selected based on gene expression profiles alone, and achieve higher accuracy in disease classification. We have applied one of these methodologies to idiopathic dilated cardiomyopathy (IDCM) data that we have generated using a microarray and identified significant subnetworks associated with the disease. In this paper, we review the recent endeavours in this direction, and summarize the existing methodologies and computational tools for network-based analysis of complex diseases and molecular relationships among apparently different disorders and human disease network. We also discuss the future research trends and topics of this promising field.

Convergent Lines of Evidence Support LRP8 as a Susceptibility Gene for Psychosis

Reelin (RELN) is identified as a risk gene for major psychiatric disorders such as schizophrenia (SCZ) and bipolar disorder (BPD). However, the role of its downstream signaling molecule, the low-density lipoprotein receptor-related protein 8 (LRP8) in these illnesses is still unclear. To detect whether LRP8 is a susceptibility gene for SCZ and BPD, we analyzed the associations of single nucleotide polymorphisms (SNPs) in LRP8 in a total of 47,187 subjects (including 9379 SCZ patients; 6990 BPD patients; and 12,556 controls in a screening sample, and 1397 SCZ families, 3947 BPD patients, and 8387 controls in independent replications), and identified a non-synonymous SNP rs5174 in LRP8 significantly associated with SCZ and BPD as well as the combined psychosis phenotype (P _meta = 1.99 × 10^-5, odds ratio (OR) = 1.066, 95 % confidence interval (CI) = 1.035-1.098). The risk SNP rs5174 was also associated with LRP8 messenger RNA (mRNA) expression in multiple brain tissues across independent samples (lowest P = 0.00005). Further exploratory analysis revealed that LRP8 was preferentially expressed in fetal brain tissues. Protein-protein interaction (PPI) analysis demonstrated that LRP8 significantly participated in a highly interconnected PPI network build by top risk genes for SCZ and BPD (P = 7.0 × 10^-4). Collectively, we confirmed that LRP8 is a risk gene for psychosis, and our results provide useful information toward a better understanding of genetic mechanism involving LRP8 underlying risk of complex psychiatric disorders.

STITCH 5: augmenting protein-chemical interaction networks with tissue and affinity data

Interactions between proteins and small molecules are an integral part of biological processes in living organisms. Information on these interactions is dispersed over many databases, texts and prediction methods, which makes it difficult to get a comprehensive overview of the available evidence. To address this, we have developed STITCH (‘Search Tool for Interacting Chemicals’) that integrates these disparate data sources for 430 000 chemicals into a single, easy-to-use resource. In addition to the increased scope of the database, we have implemented a new network view that gives the user the ability to view binding affinities of chemicals in the interaction network. This enables the user to get a quick overview of the potential effects of the chemical on its interaction partners. For each organism, STITCH provides a global network; however, not all proteins have the same pattern of spatial expression. Therefore, only a certain subset of interactions can occur simultaneously. In the new, fifth release of STITCH, we have implemented functionality to filter out the proteins and chemicals not associated with a given tissue. The STITCH database can be downloaded in full, accessed programmatically via an extensive API, or searched via a redesigned web interface at http://stitch.embl.de.

Detecting Protein Complexes from Signed Protein-Protein Interaction Networks

Identification of protein complexes is fundamental for understanding the cellular functional organization. With the accumulation of physical protein-protein interaction (PPI) data, computational detection of protein complexes from available PPI networks has drawn a lot of attentions. While most of the existing protein complex detection algorithms focus on analyzing the physical protein-protein interaction network, none of them take into account the "signs"  (i.e., activation-inhibition relationships) of physical interactions. As the "signs"  of interactions reflect the way proteins communicate, considering the "signs"  of interactions can not only increase the accuracy of protein complex identification, but also deepen our understanding of the mechanisms of cell functions. In this study, we proposed a novel Signed Graph regularized Nonnegative Matrix Factorization (SGNMF) model to identify protein complexes from signed PPI networks. In our experiments, we compared the results collected by our model on signed PPI networks with those predicted by the state-of-the-art complex detection techniques on the original unsigned PPI networks. We observed that considering the "signs"  of interactions significantly benefits the detection of protein complexes. Furthermore, based on the predicted complexes, we predicted a set of signed complex-complex interactions for each dataset, which provides a novel insight of the higher level organization of the cell. All the experimental results and codes can be downloaded from http://mail.sysu.edu.cn/home/stsddq@mail.sysu.edu.cn/dai/others/SGNMF.zip.

A fast and high performance multiple data integration algorithm for identifying human disease genes

BACKGROUND

Integrating multiple data sources is indispensable in improving disease gene identification. It is not only due to the fact that disease genes associated with similar genetic diseases tend to lie close with each other in various biological networks, but also due to the fact that gene-disease associations are complex. Although various algorithms have been proposed to identify disease genes, their prediction performances and the computational time still should be further improved.

RESULTS

In this study, we propose a fast and high performance multiple data integration algorithm for identifying human disease genes. A posterior probability of each candidate gene associated with individual diseases is calculated by using a Bayesian analysis method and a binary logistic regression model. Two prior probability estimation strategies and two feature vector construction methods are developed to test the performance of the proposed algorithm.

CONCLUSIONS

The proposed algorithm is not only generated predictions with high AUC scores, but also runs very fast. When only a single PPI network is employed, the AUC score is 0.769 by using F2 as feature vectors. The average running time for each leave-one-out experiment is only around 1.5 seconds. When three biological networks are integrated, the AUC score using F3 as feature vectors increases to 0.830, and the average running time for each leave-one-out experiment takes only about 12.54 seconds. It is better than many existing algorithms.

ProSim: A Method for Prioritizing Disease Genes Based on Protein Proximity and Disease Similarity

Predicting disease genes for a particular genetic disease is very challenging in bioinformatics. Based on current research studies, this challenge can be tackled via network-based approaches. Furthermore, it has been highlighted that it is necessary to consider disease similarity along with the protein's proximity to disease genes in a protein-protein interaction (PPI) network in order to improve the accuracy of disease gene prioritization. In this study we propose a new algorithm called proximity disease similarity algorithm (ProSim), which takes both of the aforementioned properties into consideration, to prioritize disease genes. To illustrate the proposed algorithm, we have conducted six case studies, namely, prostate cancer, Alzheimer's disease, diabetes mellitus type 2, breast cancer, colorectal cancer, and lung cancer. We employed leave-one-out cross validation, mean enrichment, tenfold cross validation, and ROC curves to evaluate our proposed method and other existing methods. The results show that our proposed method outperforms existing methods such as PRINCE, RWR, and DADA.

Estimating a ranked list of human hereditary diseases for clinical phenotypes by using weighted bipartite network

With the availability of the huge medical knowledge data on the Internet such as the human disease network, protein-protein interaction (PPI) network, and phenotypegene, gene-disease bipartite networks, it becomes practical to help doctors by suggesting plausible hereditary diseases for a set of clinical phenotypes. However, identifying candidate diseases that best explain a set of clinical phenotypes by considering various heterogeneous networks is still a challenging task. In this paper, we propose a new method for estimating a ranked list of plausible diseases by associating phenotypegene with gene-disease bipartite networks. Our approach is to count the frequency of all the paths from a phenotype to a disease through their associated causative genes, and link the phenotype to the disease with path frequency in a new phenotype-disease bipartite (PDB) network. After that, we generate the candidate weights for the edges of phenotypes with diseases in PDB network. We evaluate our proposed method in terms of Normalized Discounted Cumulative Gain (NDCG), and demonstrate that we outperform the previously known disease ranking method called Phenomizer.