A review of disease risk prediction methods and applications in the omics era.

Risk prediction and disease prevention are the innovative care challenges of the 21st century. Apart from freeing the individual from the pain of disease, it will lead to low medical costs for society. Until very recently, risk assessments have ushered in a new era with the emergence of omics technologies, including genomics, transcriptomics, epigenomics, proteomics, and so on, which potentially advance the ability of biomarkers to aid prediction models. While risk prediction has achieved great success, there are still some challenges and limitations. We reviewed the general process of omics-based disease risk model construction and the applications in four typical diseases. Meanwhile, we highlighted the problems in current studies and explored the potential opportunities and challenges for future clinical practice.

Identifying functional subtypes of IgA nephropathy based on three machine learning algorithms and WGCNA.

BACKGROUND: IgA nephropathy (IgAN) is one of the most common primary glomerulonephritis, which is a significant cause of renal failure. At present, the classification of IgAN is often limited to pathology, and its molecular mechanism has not been established. Therefore we aim to identify subtypes of IgAN at the molecular level and explore the heterogeneity of subtypes in terms of immune cell infiltration, functional level. METHODS: Two microarray datasets (GSE116626 and GSE115857) were downloaded from GEO. Differential expression genes (DEGs) for IgAN were screened with limma. Three unsupervised clustering algorithms (hclust, PAM, and ConsensusClusterPlus) were combined to develop a single-sample subtype random forest classifier (SSRC). Functional subtypes of IgAN were defined based on functional analysis and current IgAN findings. Then the correlation between IgAN subtypes and clinical features such as eGFR and proteinuria was evaluated by using Pearson method. Subsequently, subtype heterogeneity was verified by subtype-specific modules identification based on weighted gene co-expression network analysis(WGCNA) and immune cell infiltration analysis based on CIBERSORT algorithm. RESULTS: We identified 102 DEGs as marker genes for IgAN and three functional subtypes namely: viral-hormonal, bacterial-immune and mixed type. We screened seventeen genes specific to viral hormonal type (ATF3, JUN and FOS etc.), and seven genes specific to bacterial immune type (LIF, C19orf51 and SLPI etc.). The subtype-specific genes showed significantly high correlation with proteinuria and eGFR. The WGCNA modules were in keeping with functions of the IgAN subtypes where the MEcyan module was specific to the viral-hormonal type and the MElightgreen module was specific to the bacterial-immune type. The results of immune cell infiltration revealed subtype-specific cell heterogeneity which included significant differences in T follicular helper cells, resting NK cells between viral-hormone type and control group; significant differences in eosinophils, monocytes, macrophages, mast cells and other cells between bacterial-immune type and control. CONCLUSION: In this study, we identified three functional subtypes of IgAN for the first time and specific expressed genes for each subtype. Then we constructed a subtype classifier and classify IgAN patients into specific subtypes, which may be benefit for the precise treatment of IgAN patients in future.

Integrating multi-omics data to analyze the potential pathogenic mechanism of CTSH gene involved in type 1 diabetes in the exocrine pancreas.

Type 1 diabetes (T1D) is an autoimmune disease caused by the destruction of insulin-producing pancreatic islet beta cells. Despite significant advancements, the precise pathogenesis of the disease remains unknown. This work integrated data from expression quantitative trait locus (eQTL) studies with Genome wide association study (GWAS) summary data of T1D and single-cell transcriptome data to investigate the potential pathogenic mechanisms of the CTSH gene involved in T1D in exocrine pancreas. Using the summary data-based Mendelian randomization (SMR) approach, we obtained four potential causative genes associated with T1D: BTN3A2, PGAP3, SMARCE1 and CTSH. To further investigate these genes'roles in T1D development, we validated them using a scRNA-seq dataset from pancreatic tissues of both T1D patients and healthy controls. The analysis showed a significantly high expression of the CTSH gene in T1D acinar cells, whereas the other three genes showed no significant changes in the scRNA-seq data. Moreover, single-cell WGCNA analysis revealed the strongest positive correlation between the module containing CTSH and T1D. In addition, we found cellular ligand-receptor interactions between the acinar cells and different cell types, especially ductal cells. Finally, based on functional enrichment analysis, we hypothesized that the CTSH gene in the exocrine pancreas enhances the antiviral response, leading to the overexpression of pro-inflammatory cytokines and the development of an inflammatory microenvironment. This process promotes ß cells injury and ultimately the development of T1D. Our findings offer insights into the underlying pathogenic mechanisms of T1D.

Genome-wide identification of m6A-associated single nucleotide polymorphisms in complex diseases of nervous system.

N6-methyladenosine (m6A) is one of the most abundant chemical modifications on RNA and can affect the occurrence and development of diseases. Some studies have shown that the expressions of some m6A-related genes are significantly regulated by single nucleotide variants (SNV). However, the function of m6A-associated single nucleotide polymorphisms (m6A-SNP) remains unclear in multiple sclerosis (MS), Alzheimer's disease (AD) and Parkinson's disease (PD). Here, we identified the disease-associated m6A-SNPs by integrating genome-wide association study (GWAS) and m6A-SNPs from the RMVar database, and confirmed the relationship between these identified m6A-SNPs and their target genes in eQTL analysis and gene differential expression analysis. Finally, 26 genes corresponding to 20 m6A-SNPs with eQTL signals were identified and differentially expressed (P < 0.05) in MS, 15 genes corresponding to 12 m6A-SNPs (P < 1e-04) were differentially expressed in AD, and 27 PD-associated m6A-SNPs that regulated the expression of 31 genes were identified. There were 5 HLA genes with eQTL signals (HLA-DQB1, HLA-DRB1, HLA-DQA1, HLA-DQA2 and HLA-DQB1-AS1) to be detected in the three diseases. In summary, our study provided new insights into understanding the potential roles of these m6A-SNPs in disease pathogenesis as well as therapeutic target.

Identification of key genes associated with multiple sclerosis based on gene expression data from peripheral blood mononuclear cells.

The aim of this study was to identify the potential key candidate genes of multiple sclerosis (MS) and uncover mechanisms in MS. We combined data from the microarray expression profile of three MS stages and performed bioinformatics analysis. Differentially expressed genes (DEGs) were identified among the distinct stages of MS and healthy controls, and a total of 349 shared DEGs were identified. Gene ontology (GO) and pathway enrichment analyses showed that the DEGs were significantly enriched in the biological processes (BPs) of purine-related metabolic processes and signaling, especially the common DEGs, which were enriched in some immunological processes. Most of the DEGs were enriched in signaling pathways associated with the immune system, some immune diseases and infectious disease pathways. Through a protein-protein interaction (PPI) network analysis and a gene expression regulatory network constructed with MS-related miRNAs, we confirmed FOS, TP53, VEGFA, JUN, HIF1A, RB1, PTGS2, CXCL8, OAS2, NFKBIA and OAS1 as candidate genes of MS. Furthermore , we explored the potential SNPs associated with MS by database mining. In conclusion, this study provides the identified genes, SNPs, biological processes, and cellular pathways associated with MS. The uncovered candidate genes may be potential biomarkers involved in the diagnosis and therapy of MS.

The shared and specific mechanism of four autoimmune diseases.

Interaction between genetic and epigenetic mechanisms may lead to autoimmune diseases. The features of these diseases show familial aggregation. The generality and specificity are keys to studying pathogenesis and etiology of them. This research integrated data of genetics and epigenetics, to find disease-related genes based on the levels of expression and regulation, and explored then to the shared and specific mechanism of them by analyzing shared and specific pathways of common four autoimmune diseases, including Type 1 Diabetes Mellitus (T1D), Multiple Sclerosis (MS), Rheumatoid Arthritis (RA) and Systemic Lupus Erythematosus (SLE). The results showed that Lysosome and Fc gamma R-mediated phagocytosis are shared pathways of the four diseases. It means that the occurrence and development of them may associate with lysosomes and phagocytosis. And there were 2 pathways are shared pathways of three diseases, ribosome pathway associated with susceptibility to MS, RA and SLE, and Pathogenic Escherichia coli infection associated with susceptibility to T1D, MS and RA; 9 pathways are shared pathways of two diseases. The corporate underlying causes of these diseases may be these shared pathways activated. Furthermore, we found that T1D-related specific pathways (Insulin signaling,etc.) were 9, MS (Proteasome,etc.) is also 9, RA and SLE is 10 and 6 respectively. These pathways could help us to reveal shared and specific mechanisms of the four diseases.

Genome-wide pathway-based association analysis identifies risk pathways associated with Parkinson's disease.

Parkinson's disease (PD) is the second most common neurodegenerative disease. It is generally believed that it is influenced by both genetic and environmental factors, but the precise pathogenesis of PD is unknown to date. In this study, we performed a pathway analysis based on genome-wide association study (GWAS) to detect risk pathways of PD in three GWAS datasets. We first mapped all SNP markers to autosomal genes in each GWAS dataset. Then, we evaluated gene risk values using the minimum P-value of the tagSNPs. We took a pathway as a unit to identify the risk pathways based on the cumulative risks of the genes in the pathway. Finally, we combine the analysis results of the three datasets to detect the high risk pathways associated with PD. We found there were five same pathways in the three datasets. Besides, we also found there were five pathways which were shared in two datasets. Most of these pathways are associated with nervoussystem. Five pathways had been reported to be PD-related pathways in the previous literature. Our findings also implied that there was a close association between immune response and PD. Continued investigation of these pathways will further help us explain the pathogenesis of PD.

Genome-wide haplotype association study identify the FGFR2 gene as a risk gene for acute myeloid leukemia.

Acute myeloid leukemia (AML) is a cancer of the myeloid line of blood cells, and generally considered to be caused by environment and genetic factors. In this study, we combined a genome-wide haplotype association study (GWHAS) and gene prioritization strategy to mine AML-related genetic affect factors and understand its pathogenesis. A total of 175 AML patients were downloaded from the public GEO database (GSE32462) and 218 matched Caucasian controls were from the HapMap Project. We first identified the linkage disequilibrium (LD) blocks and performed a GWHAS to scan AML-related haplotypes. Then we mapped these haplotypes to the corresponding genes as candidate. And finally, we prioritized all the AML candidate genes based on the similarity with 38 known AML susceptibility genes. The results showed that 1754 haplotypes were significant associated with AML (P<1E-5) and mapped to 591 candidate genes. After prioritizing all 591 AML candidate genes, we obtained four genes ranking at the front as AML risk genes: RUNX1, JAK1, PDGFRA, and FGFR2. Among them, RUNX1, JAK1 and PDGFRA had been confirmed as AML risk genes. In particular, we found that the gene FGFR2 was a novel AML susceptibility gene with a haplotype TT (rs7090018 and rs2912759) showed significant association with AML (P-value = 7.07E-06). In a word, our findings might provide a new perspective to understand the pathogenesis of AML.

Integrative analysis of genome-wide association studies and gene expression analysis identifies pathways associated with rheumatoid arthritis.

Rheumatoid arthritis (RA) is a complex and systematic autoimmune disease, which is usually influenced by both genetic and environmental factors. Pathway analyses based on a single data type such as microarray data or SNP data have successfully revealed some biology pathways associated with RA. However, we found that the pathway analysis based on a single data type only provide limited understanding about the pathogenesis of RA. Gene-disease association is usually caused by many ways, such as genotype, gene expression and so on. Therefore, the integrative analysis method combining multiple levels of evidence can more precisely and comprehensively identify the pathway associations. In this study, we performed a pathway analysis by integrating GWAS and gene expression analysis to detect the RA-related pathways. The integrative analysis identified 28 pathways associated with RA. Among these pathways, 18 pathways were also found by both GWAS and gene expression analysis, 7 pathways are novel RA-related pathways, such as B cell receptor signaling pathway, Toll-like receptor signaling pathway, Fc gamma R-mediated phagocytosis and so on. Compared with pathway analyses using only one type genomic data, we found integrative analysis can increase the power to identify the real associations and provided more stable and accurate results. We believe these results will contribute to perform future genetic studies in RA pathogenesis and may promote the development of new therapeutic strategies by targeting these pathways.

The drug target genes show higher evolutionary conservation than non-target genes.

Although evidence indicates that drug target genes share some common evolutionary features, there have been few studies analyzing evolutionary features of drug targets from an overall level. Therefore, we conducted an analysis which aimed to investigate the evolutionary characteristics of drug target genes. We compared the evolutionary conservation between human drug target genes and non-target genes by combining both the evolutionary features and network topological properties in human protein-protein interaction network. The evolution rate, conservation score and the percentage of orthologous genes of 21 species were included in our study. Meanwhile, four topological features including the average shortest path length, betweenness centrality, clustering coefficient and degree were considered for comparison analysis. Then we got four results as following: compared with non-drug target genes, 1) drug target genes had lower evolutionary rates; 2) drug target genes had higher conservation scores; 3) drug target genes had higher percentages of orthologous genes and 4) drug target genes had a tighter network structure including higher degrees, betweenness centrality, clustering coefficients and lower average shortest path lengths. These results demonstrate that drug target genes are more evolutionarily conserved than non-drug target genes. We hope that our study will provide valuable information for other researchers who are interested in evolutionary conservation of drug targets.

Genome-wide haplotype association study identify TNFRSF1A, CASP7, LRP1B, CDH1 and TG genes associated with Alzheimer's disease in Caribbean Hispanic individuals.

Alzheimer's disease (AD) is an acquired disorder of cognitive and behavioral impairment. It is considered to be caused by variety of factors, such as age, environment and genetic factors. In order to identify the genetic affect factors of AD, we carried out a bioinformatic approach which combined genome-wide haplotype-based association study with gene prioritization. The raw SNP genotypes data was downloaded from GEO database (GSE33528). It contains 615 AD patients and 560 controls of Caribbean Hispanic individuals. Firstly, we identified the linkage disequilibrium (LD) haplotype blocks and performed genome-wide haplotype association study to screen significant haplotypes that were associated with AD. Then we mapped these significant haplotypes to genes and obtained candidate genes set for AD. At last, we prioritized AD candidate genes based on their similarity with 36 known AD genes, so as to identify AD related genes. The results showed that 141 haplotypes on 134 LD blocks were significantly associated with AD (P<1E-4), and these significant haplotypes were mapped to 132 AD candidate genes. After prioritizing these candidate genes, we found seven AD related genes: APOE, APOC1, TNFRSF1A, LRP1B, CDH1, TG and CASP7. Among these genes, APOE and APOC1 are known AD risk genes. For the other five genes TNFRSF1A, CDH1, CASP7, LRP1B and TG, this is the first genetic association study which showed the significant association between these five genes and AD susceptibility in Caribbean Hispanic individuals. We believe that our findings can provide a new perspective to understand the genetic affect factors of AD.

Genes with stable DNA methylation levels show higher evolutionary conservation than genes with fluctuant DNA methylation levels.

Different human genes often exhibit different degrees of stability in their DNA methylation levels between tissues, samples or cell types. This may be related to the evolution of human genome. Thus, we compared the evolutionary conservation between two types of genes: genes with stable DNA methylation levels (SM genes) and genes with fluctuant DNA methylation levels (FM genes). For long-term evolutionary characteristics between species, we compared the percentage of the orthologous genes, evolutionary rate dn/ds and protein sequence identity. We found that the SM genes had greater percentages of the orthologous genes, lower dn/ds, and higher protein sequence identities in all the 21 species. These results indicated that the SM genes were more evolutionarily conserved than the FM genes. For short-term evolutionary characteristics among human populations, we compared the single nucleotide polymorphism (SNP) density, and the linkage disequilibrium (LD) degree in HapMap populations and 1000 genomes project populations. We observed that the SM genes had lower SNP densities, and higher degrees of LD in all the 11 HapMap populations and 13 1000 genomes project populations. These results mean that the SM genes had more stable chromosome genetic structures, and were more conserved than the FM genes.

Comparing the evolutionary conservation between human essential genes, human orthologs of mouse essential genes and human housekeeping genes.

Human housekeeping genes are often confused with essential human genes, and several studies regard both types of genes as having the same level of evolutionary conservation. However, this is not necessarily the case. To clarify this, we compared the differences between human housekeeping genes and essential human genes with respect to four aspects: the evolutionary rate (dN/dS), protein sequence identity, single-nucleotide polymorphism (SNP) density and level of linkage disequilibrium (LD). The results showed that housekeeping genes had lower evolutionary rates, higher sequence identities, lower SNP densities and higher levels of LD compared with essential genes. Together, these findings indicate that housekeeping and essential genes are two distinct types of genes, and that housekeeping genes have a higher level of evolutionary conservation. Therefore, we suggest that researchers should pay careful attention to the distinctions between housekeeping genes and essential genes. Moreover, it is still controversial whether we should substitute human orthologs of mouse essential genes for human essential genes. Therefore, we compared the evolutionary features between human orthologs of mouse essential genes and human housekeeping genes and we got inconsistent results in long-term and short-term evolutionary characteristics implying the irrationality of simply replacing human essential genes with human orthologs of mouse essential genes.

RADB: a database of rheumatoid arthritis-related polymorphisms.

Rheumatoid arthritis (RA) is an autoimmune disease that has a complex genetic basis. Therefore, it is important to explore the genetic background of RA. The extensive recent application of polymorphic genetic markers, especially single nucleotide polymorphisms, has presented us with a large quantity of genetic data. In this study, we developed the Database of Rheumatoid Arthritis-related Polymorphisms (RADB), to integrate all the RA-related genetic polymorphisms and provide a useful resource for researchers. We manually extracted the RA-related polymorphisms from 686 published reports, including RA susceptibility loci, polymorphisms associated with particular clinical features of RA, polymorphisms associated with drug response in RA and polymorphisms associated with a higher risk of cardiovascular disease in RA. Currently, RADB V1.0 contains 3235 polymorphisms that are associated with 636 genes and refer to 68 countries. The detailed information extracted from the literature includes basic information about the articles (e.g., PubMed ID, title and abstract), population information (e.g., country, geographic area and sample size) and polymorphism information (e.g., polymorphism name, gene, genotype, odds ratio and 95% confidence interval, P-value and risk allele). Meanwhile, useful annotations, such as hyperlinks to dbSNP, GenBank, UCSC, Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway, are included. In addition, a tool for meta-analysis was developed to summarize the results of multiple studies. The database is freely available at http://www.bioapp.org/RADB. Database URL: http://www.bioapp.org/RADB.

Association between polymorphisms in the promoter region of interleukin-10 and susceptibility to inflammatory bowel disease.

The aim of this study was to assess the association of polymorphisms in the promoter region of the IL-10 gene with the risk of inflammatory bowel disease (IBD), including Crohn's disease (CD) and ulcerative colitis (UC). Fifteen studies (3,693 cases and 4,574 controls) were included in a meta-analysis of association between IL-10 -1082G/A, -819C/T and -592C/A polymorphisms, and IBD, CD and UC using allele contrast and the recessive, dominant, and additive models. Hardy-Weinberg equilibrium was confirmed for each study. Heterogeneity and study quality were investigated using stratification analyses and sensitivity analyses. Polymorphism -1082G/A showed significant association with CD, with odds ratios (ORs) for the GG + GA genotype and GG versus AA genotype of 1.278 (1.004-1.627) and 1.238 (1.027-1.492) in all subjects. Significant associations were found in the Caucasian subgroup using the allele contrast, dominant, and additive models. C-allele carriers of the -819C/T polymorphism were at increased risk of IBD (OR 1.093, 95% CI 1.004-1.190). Association with the -819C/T polymorphism was also found in Caucasians with CD (C vs. T: OR 1.104, 95% CI 1.010-1.206; CC + CT vs. TT: OR 1.328, 95% CI 1.006-1.754; CC vs. TT: OR 1.339, 95% CI 1.008-1.778), and with UC (CC vs. CT + TT: OR 1.188, 95% CI 1.019-1.385). No significant association was found between the -592C/A polymorphism and IBD, CD or UC. In conclusion, the meta-analysis demonstrated clear association between the IL-10 polymorphisms -1082G/A and -819C/T and the risk of IBD.

Prioritization of candidate genes for periodontitis using multiple computational tools.

BACKGROUND: Both genetic and environmental factors contribute to the development of periodontitis. Genetic studies identified a variety of candidate genes for periodontitis. The aim of the present study is to identify the most promising candidate genes for periodontitis using an integrative gene ranking method. METHODS: Seed genes that were confirmed to be associated with periodontitis were identified using text mining. Three types of candidate genes were then extracted from different resources (expression profiles, genome-wide association studies). Combining the seed genes, four freely available bioinformatics tools (ToppGene, DIR, Endeavour, and GPEC) were integrated for prioritization of candidate genes. Candidate genes that identified with at least three programs and ranked in the top 20 by each program were considered the most promising. RESULTS: Prioritization analysis resulted in 21 promising genes involved or potentially involved in periodontitis. Among them, IL18 (interleukin 18), CD44 (CD44 molecule), CXCL1 (chemokine [CXC motif] ligand 1), IL6ST (interleukin 6 signal transducer), MMP3 (matrix metallopeptidase 3), MMP7, CCR1 (chemokine [C-C motif] receptor 1), MMP13, and TLR9 (Toll-like receptor 9) had been associated with periodontitis. However, the roles of other genes, such as CSF3 (colony stimulating factor 3 receptor), CD40, TNFSF14 (tumor necrosis factor receptor superfamily, member 14), IFNB1 (interferon-ß1), TIRAP (toll-interleukin 1 receptor domain containing adaptor protein), IL2RA (interleukin 2 receptor α), ETS1 (v-ets avian erythroblastosis virus E26 oncogene homolog 1), GADD45B (growth arrest and DNA-damage-inducible 45 ß), BIRC3 (baculoviral IAP repeat containing 3), VAV1 (vav 1 guanine nucleotide exchange factor), COL5A1 (collagen, type V, α1), and C3 (complement component 3), have not been investigated thoroughly in the process of periodontitis. These genes are mainly involved in bacterial infection, immune response, and inflammatory reaction, suggesting that further characterizing their roles in periodontitis will be important. CONCLUSIONS: A combination of computational tools will be useful in mining candidate genes for periodontitis. These theoretical results provide new clues for experimental biologists to plan targeted experiments.

Meta-analysis of 125 rheumatoid arthritis-related single nucleotide polymorphisms studied in the past two decades.

OBJECTIVE: Candidate gene association studies and genome-wide association studies (GWAs) have identified a large number of single nucleotide polymorphisms (SNPs) loci affecting susceptibility to rheumatoid arthritis (RA). However, for the same locus, some studies have yielded inconsistent results. To assess all the available evidence for association, we performed a meta-analysis on previously published case-control studies investigating the association between SNPs and RA. METHODS: Two hundred and sixteen studies, involving 125 SNPs, were reviewed. For each SNP, three genetic models were considered: the allele, dominant and recessive effects models. For each model, the effect summary odds ratio (OR) and 95% CIs were calculated. Cochran's Q-statistics were used to assess heterogeneity. If the heterogeneity was high, a random effects model was used for meta-analysis, otherwise a fixed effects model was used. RESULTS: The meta-analysis results showed that: (1) 30, 28 and 26 SNPs were significantly associated with RA (P<0.01) for the allele, dominant, and recessive models, respectively. (2) rs2476601 (PTPN22) showed the strongest association for all the three models: OR = 1.605, 95% CI: 1.540-1.672, P<1.00E-15 for the T-allele; OR = 1.638, 95% CI: 1.565-1.714, P<1.00E-15 for the T/T+T/C genotype and OR = 2.544, 95% CI: 2.173-2.978, P<1.00E-15 for the T/T genotype. (3) Only 23 (18.4%), 13 (10.4%) and 15 (12.0%) SNPs had high heterogeneity (P<0.01) for the three models, respectively. (4) For some of the SNPs, there was no publication bias according to Funnel plots and Egger's regression tests (P<0.01). For the other SNPs, the associations were tested in only a few studies, and may have been subject to publication bias. More studies on these loci are required. CONCLUSION: Our meta-analysis provides a comprehensive evaluation of the RA association studies from the past two decades. The detailed meta-analysis results are available at: http://210.46.85.180/DRAP/index.php/Metaanalysis/index.

DBGSA: a novel method of distance-based gene set analysis.

When compared with single gene functional analysis, gene set analysis (GSA) can extract more information from gene expression profiles. Currently, several gene set methods have been proposed, but most of the methods cannot detect gene sets with a large number of minor-effect genes. Here, we propose a novel distance-based gene set analysis method. The distance between two groups of genes with different phenotypes based on gene expression should be larger if a certain gene set is significantly associated with the given phenotype. We calculated the distance between two groups with different phenotypes, estimated the significant P-values using two permutation methods and performed multiple hypothesis testing adjustments. This method was performed on one simulated data set and three real data sets. After a comparison and literature verification, we determined that the gene resampling-based permutation method is more suitable for GSA, and the centroid statistical and average linkage statistical distance methods are efficient, especially in detecting gene sets containing more minor-effect genes. We believe that this distance-based method will assist us in finding functional gene sets that are significantly related to a complex trait. Additionally, we have prepared a simple and publically available Perl and R package (http://bioinfo.hrbmu.edu.cn/dbgsa or http://cran.r-project.org/web/packages/DBGSA/).

Predict potential drug targets from the ion channel proteins based on SVM.

The identification of molecular targets is a critical step in the drug discovery and development process. Ion channel proteins represent highly attractive drug targets implicated in a diverse range of disorders, in particular in the cardiovascular and central nervous systems. Due to the limits of experimental technique and low-throughput nature of patch-clamp electrophysiology, they remain a target class waiting to be exploited. In our study, we combined three types of protein features, primary sequence, secondary structure and subcellular localization to predict potential drug targets from ion channel proteins applying classical support vector machine (SVM) method. In addition, our prediction comprised two stages. In stage 1, we predicted ion channel target proteins based on whole-genome target protein characteristics. Firstly, we performed feature selection by Mann-Whitney U test, then made predictions to identify potential ion channel targets by SVM and designed a new evaluating indicator Q to prioritize results. In stage 2, we made a prediction based on known ion channel target protein characteristics. Genetic algorithm was used to select features and SVM was used to predict ion channel targets. Then, we integrated results of two stages, and found that five ion channel proteins appeared in both prediction results including CGMP-gated cation channel beta subunit and Gamma-aminobutyric acid receptor subunit alpha-5, etc., and four of which were relative to some nerve diseases. It suggests that these five proteins are potential targets for drug discovery and our prediction strategies are effective.