Next Article in Journal
Melatonin Secretion Is Increased in Children with Severe Traumatic Brain Injury
Next Article in Special Issue
Biochemical and Computational Insights on a Novel Acid-Resistant and Thermal-Stable Glucose 1-Dehydrogenase
Previous Article in Journal
Adiponectin Is Involved in Connective Tissue Growth Factor-Induced Proliferation, Migration and Overproduction of the Extracellular Matrix in Keloid Fibroblasts
Previous Article in Special Issue
PCVMZM: Using the Probabilistic Classification Vector Machines Model Combined with a Zernike Moments Descriptor to Predict Protein–Protein Interactions from Protein Sequences

Int. J. Mol. Sci. 2017, 18(5), 1045;

Determination of Genes Related to Uveitis by Utilization of the Random Walk with Restart Algorithm on a Protein–Protein Interaction Network
Department of Ophthalmology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai 200127, China
College of Information Engineering, Shanghai Maritime University, Shanghai 201306, China
School of Life Sciences, Shanghai University, Shanghai 200444, China
Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China
Author to whom correspondence should be addressed.
Academic Editor: Christo Z. Christov
Received: 9 April 2017 / Accepted: 9 May 2017 / Published: 13 May 2017


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.
uveitis; protein–protein interaction; random walk with restart algorithm

1. Introduction

Uveitis is defined as an inflammation of the uveal tract, which is composed of the ciliary body, iris and choroid [1,2]. Uveitis is one of the leading causes of permanent and irreversible blindness in young and middle-aged people and accounts for 10–15% of blindness in the Western world [1,2,3]. Uveitis can be caused by infectious and non-infectious factors; the latter include Vogt–Koyanagi–Harada (VKH) syndrome, Behcet’s disease (BD), acute anterior uveitis (AAU), birdshot chorioretinopathy (BCR) and some types of cancers. VKH is an autoimmune disease characterized by systemic disorders including poliosis, vitiligo, alopecia, auditory signs and disorders of the central nervous system [4,5]. BD is a chronic multi-systemic inflammatory disease characterized by nongranulomatous uveitis, oral ulcers and skin lesions [2,6]. AAU is the most common non-infectious cause of uveitis and is characterized by self-limiting and recurrent inflammation involving the ciliary and iris body [7]. BCR is a chronic, bilateral, and posterior uveitis that has an almost 100% genetic association with HLA-A29 [8]. Uveitis or uveitis masquerade syndrome could also be induced by some intraocular tumors, such as retinoblastoma and intraocular lymphoma, or their therapeutic approaches [9,10,11,12,13,14,15].
It has been reported that complex genetic mechanisms coupled with an aberrant immune response may be involved in the development of uveitis. In some cases, the pathogenesis of uveitis seemly has a different cause than those described above, such as sarcoidosis [16]. Mutations in different genes and gene families have been discovered in patients. In this study, we focused on the most important causes of uveitis and research for the putative genes involved in these processes. Human leukocyte antigens (HLAs) are the major molecules that are important for the development of uveitis, including uveitis associated with VKH (HLA-DR4, DRB1/DQA1), BD (HLA-B51), AAU (HLA-B27) and BCR (HLA-A29). In addition, genome-wide association studies revealed that abnormities of many non-HLA genes such as the interleukin (IL) family and the Signal transducer and activator of transcription 4 (STAT4) also participate in the progression of uveitis [17,18,19]. IL23R is associated with both VKH and AAU [20]. Furthermore, copy number variations (CNVs) of Toll-like receptors (TLRs), a family of cellular receptors that function in innate immune response, are associated with BD, VKH and AAU. These genes include TLRs 1–3, TLRs 5–7, and TLRs 9–10 [21]. SNPs of TLR4 were also shown to be involved in the development of BD [22]. In addition, it has been demonstrated that there is increased expression of T-bet and IFN-γ, two genes involved in the Th1 cell pathway, in uveitis patients [23]. The activator of STAT4 affects IL-17 production and is a shared risk factor for BD in different cohorts [17,24]. Finally, interleukins (notably IL-2, IL12B, IL18 and IL23R) are important cytokines that play a pathogenic role in the process of uveitis [2,17,25]. In this study, we mainly focused on the genes that play an important role in the immune system, transcription, or cell adhesion.
Using traditional methods, it is quite difficult to collect these large-scale data and analyze genes synthetically. The microarray is a widely used tool for the identification of novel genes. Microarray analysis has been used to determine a number of genes that are associated with uveitis, including the IL10 family and several other transcripts [16,26,27,28,29]. In recent years, computational analysis has been applied to identify virulence genes, but many of these genes were identified based on guilt by association (GBA) [30,31,32]. This approach assumes that the candidate genes, which are neighbors of disease genes, are more likely to be new virulence genes. Thus, the GBA-based methods only consider the neighbors of known disease genes to discover novel candidates. Therefore, these methods only examine part of the gene network. Random Walk with Restart (RWR) is another algorithm that identifies disease-related genes [33,34,35]. This algorithm utilizes a set of seed nodes that represent disease genes and performs random walking on the gene network. When the probabilities of all nodes are stable, the probability of a node gene correlating with disease is updated. The genes that correspond to nodes that have high probabilities may be potential novel candidate virulence genes. This method is useful for mining disease genes and to better explore the mechanism of disease. In addition, other studies have adopted the shortest path (SP) algorithm to identify novel disease genes [36,37,38,39,40,41]. By searching the shortest paths that connect any two validated disease genes, genes that are present in these paths could be extracted and considered as novel disease genes. An obvious advantage of the RWR or SP algorithms is that these algorithms utilize the entire gene network and consider more factors, therefore performing a more extensive and reliable analysis.
As discussed above, many genetic factors contribute to the pathogenesis of uveitis. In this study, we utilized computational analyses to build a genetic network based on previously known factors. A computational method was built to identify novel genes related to uveitis. First, a large network was constructed using human protein–protein interactions (PPIs). Next, the RWR algorithm was performed on the network using the validated uveitis-related genes as seed nodes, yielding several possible candidate genes. These candidate genes were filtered based on a set of criteria that were built by p-values and their associations with validated uveitis-related genes. To indicate the utility of the method, it was evaluated by five-fold cross-validation, resulting in the average F1-measure of 0.189. Furthermore, the proposed method was compared with a classic GBA-based method [30,31,32] to further prove its effectiveness for identification of uveitis-related genes. Through our method, 56 novel candidate genes were identified and extensively analyzed.

2. Results and Discussion

2.1. Results of Testing Random Walk with Restart (RWR)-Based Method

Before the RWR-based method was used to identify novel uveitis-related genes, five-fold cross-validation was adopted to evaluate its utility. For each part, the results yielded by the method on the rest four parts were counted as recall, precision and F1-measure, which are listed in Table 1. It can be observed that the average of recall, precision and F1-measure was 0.287, 0.141 and 0.189, respectively. Although these measurements are not very high, the RWR-based method is still acceptable due to the difficulties for identification of novel genes with given functions. Besides, the utility of the RWR-based method would be further proved by comparing it with other methods, which is described in Section 2.5.

2.2. RWR Genes

Based on the uveitis-related genes, the RWR algorithm yielded a probability for each gene in the PPI network, which indicated the likelihood of the gene being important for uveitis. Then, genes were selected that had probabilities larger than 10−5. From our analysis, we obtained 3641 RWR genes, which are provided with their RWR probabilities in Supplementary Table S1.

2.3. Candidate Genes

According to the RWR-based method detailed in Section 3.3, RWR genes were filtered using a permutation test. For each RWR gene, a p-value was assigned to indicate whether the RWR gene is specific for uveitis. The p-value for each of the 3641 RWR genes is also provided in Supplementary Table S1. We found 1231 candidate genes that had a p-value < 0.05 (see the first 1231 genes in Supplementary Table S1).
The 1231 candidate genes were then further analyzed using the criteria outlined in Section 3.3. For each candidate gene, MIS (cf. Equation (3)) and MFS (cf. Equation (5)) were calculated, and the values for each gene are available in Supplementary Table S1. The threshold for MIS was set at 900, while 0.8 was used as the threshold for MFS. Finally, we obtained 56 Ensembl IDs (listed in Table 2) corresponding to core candidate genes. These genes were deemed to be highly related to uveitis and could be considered novel candidate genes. As intuitionistic evidence, a sub-network was plotted in Figure 1, which contains the putative and validated genes. Each putative gene had strong associations with validated genes, implying that they had functions similar to those of the validated genes and may be novel uveitis-related genes with high probabilities.

2.4. Analysis of Novel Genes

In this study, the RWR-based method yielded fifty-six genes that were deemed to have a significant correlation with uveitis. Detailed information for these genes is provided in Table 2.

2.4.1. Immune System Regulation Genes

CCL4 (C-C motif chemokine ligand 4) belongs to the cytokine family and is involved in immunoregulation and inflammation. It has been reported that CCL4 is associated with BD immunopathogenesis [42]. In the majority of VKH cases, the expression of another family member CCL17 was lower in cerebrospinal fluid than in serum, which indicated its potential function in VKH [43]. CCL17 also could be inhibited by overexpression of SOCS1 in the retina to regulate the recruitment of inflammatory cells [44]. The cytokine CCL20 was considered to be a specific biomarker of HLA-B27-associated uveitis [45]. Our study revealed that CCL4, CCL17 and CCL20 likely play essential roles in uveitis.
CD40 ligand (also known as CD154) is a type II transmembrane glycoprotein that has structural homology to the proteins of the TNF (tumor necrosis factor) family [46,47,48,49]. The interaction between the CD40 and CD40 ligand is important for both cellular and humoral immune responses [50]. The CD40 and CD40 ligand interaction provides signals in T-cell priming and effecter functions [46,48,49,51,52,53], whereas monocyte and B-cell apoptosis could be inhibited by their interaction [54]. It has been demonstrated that the CD40 ligand is associated with the immune-pathogenesis of several autoimmune diseases including AU (anterior uveitis) [54,55]. The CD40 ligand is significantly expressed on T-cells in the peripheral blood of patients with AU [56]. The results of the RWR-based method revealed a MIS of 999 had a p-value < 0.001. Expression of CD80 on dendritic cells (DCs) could be induced by activation of NOD1 and NOD2 and is involved in the pathogenesis of VKH syndrome [57]. In another report, it was found that BBR downregulated the expression of costimulatory molecules CD40, CD80 and CD86 on DCs [58]. The MIS and p-value of CD80 were 999 and <0.001, respectively. We speculate that these molecules play key roles in uveitis, but their mechanism in uveitis must still be clarified.
CSF2 (colony stimulating factor 2) is a cytokine that functions as a hematological cell growth factor by stimulating stem cells to produce granulocytes and monocytes [59]. Three signaling pathways can be activated by CSF2: the JAK2/STAT pathway, the MAP pathway and the PI3K pathway [60,61,62,63,64]. CSF2 is a valuable prognostic indicator and a therapeutic target in tumors [59]. CSF2 expression in uveitis is reported as rare. However, in this study, the MIS of CSF2 was 992 with a p-value < 0.001. We speculate that CSF2 might be a key factor in the pathogenesis of uveitis.
Interleukins and their receptors are inflammatory cytokines that play an important role in immune system response. Many interleukins and their receptors are involved in uveitis, as discussed above. Our data showed that IL13, IL15, IL1A, IL1B, IL1RN, IL4, IL5, IL9, IL23A and IL2RA had MISs larger than 900 with p-values <0.05. It has been observed that the expression of IL1A is decreased in patients with clinically active BD, while the expression of IL1B is increased in patients with active, inactive or ocular BD [65]. IL1B has been associated with ocular Behcet’s disease [66]. IL-13 is a strong immunomodulatory cytokine which is a promising mode of treatment for uveitis [67,68,69,70]. IL-15 and its receptor system is involved in the inflammatory process and pathogenesis of BD and the IL-15/Fc fusion protein has been shown to inhibit IRBP1-20 specific CD80+ T cell to decrease the severity of EAU [71,72]. An aberrantly high CNV of IL23A is a common risk factor for VKH and BD [73]. In mice, IL-1RN suppresses immune-mediated ocular inflammation and is considered a potential biomarker in the management of patients with uveitis [74]. Interleukin 2 receptor α (IL2RA) is a risk locus in various autoimmune diseases and a variant of this gene, rs2104286, was demonstrated to be strongly associated with intermediated uveitis [75]. An antibody against IL2RA, daclizumab is used to reduce intermediated uveitis [76]. However, rs2104286 was not related to endogenous non-anterior uveitis [77]. EAU (experimental autoimmune uveoretinitis) disease severity was reduced in mice in which IL-1B expression was reduced in the retina through deletion of S100B, a Ca2+ binding protein [78]. In a Lewis rat model of EAU, IL-2 and IL-4 were produced in destructive foci in the retina and uveal tract. IL-2 is thought to act as a cytotoxic effector, while IL-4 is associated with a helper cell function [79]. In patients with BD, IL-2 is more highly expressed, while IL-4 is more lowly expressed [80]. Genetic findings suggest that more work should be done to evaluate both the molecular target and the inhibitor for personalized therapy.
TLR2, TLR3, TLR7 and TLR9 belong to the Toll-like receptor (TLR) family, which are key factors in pathogen recognition and activation of innate immunity. TLRs are thought to be associated with infection and auto-inflammatory or autoimmune diseases, including uveitis [81,82]. Several autoimmune diseases, including BD, are associated with certain TLR gene polymorphisms [83,84]. A significant association has been found between polymorphism of TLR2 and ocular BD patients [85]. The expression of TLR4 was significantly up-regulated in monocyte-derived macrophages from VKH patients [86]. The chitosan-mediated TLR3-siRNA transfection had a potential therapeutic effect on remitting uveitis [87]. In a Chinese Han population, a high copy number of TLR7 conferred risk for BD patients [88]. In the Japanese population, the homozygous genotypes and homozygous deplotype configuration of TLR9 SNPs was associated with the susceptibility to BD [89]. It has been reported that glucocorticoid could improve uveitis by downregulating TLR7 and TLR9 in peripheral blood of patients with uveitis [90]. In our analysis, TLR2, TLR3, TLR7 and TLR9 have MIS scores of 968, 966, 926 and 958, respectively. We argue that TLR2, TLR3, TLR7 and TLR8 play essential roles in uveitis and thus require more attention.

2.4.2. Transcription Associated Genes

Jun (also known as jun proto-oncogene) is a critical subunit of the transcription factor AP1, which is an important modulator of diverse biological processes such as cell proliferation, apoptosis and malignant transformation [91]. Jun is activated through phosphorylation at Ser 63 and Ser 73 by JNK [92,93]. A high level of Jun has been observed in various types of cancer including non-small cell lung cancer, oral squamous cell carcinoma, breast cancer and colorectal cancer [94,95,96,97,98]. Overexpression of Jun has led to aberrant tumor growth and progression and inhibited cell apoptosis [94]. The underlying mechanism of Jun as it relates to uveitis is still unclear. In a gene screen assay, it was found that expression of Jun showed a significantly higher index in experimental lens-induced uveitis rabbits [99]. In our analysis, Jun showed a significant index p-value and an MIS of 999; therefore, we propose that Jun may be an essential factor in uveitis.
STAT1 and STAT6 encode transcription factors that belong to the STAT family, where phosphorylation is activated by receptor associated kinases. Atopic dermatitis associated uveitis may be driven by TH2-mediated inflammation [100]. IL-4 is a TH2 cytokine, and binding with its receptor can activate STAT6 via the (Jak) Janus kinase/STAT signaling pathway to promote many immunomodulatory genes [101]. Furthermore, the Stat6 VT (STAT6 V547A/T548A) mouse model of atopic dermatitis exhibited uveitis symptoms [100]. In the STAT family, TH17 cells can be induced by IL-2 and suppressed by IL-27/STAT1 to contribute to uveitis [102]. In this study, STAT1 and STAT6 both had significant p-values and MIS of 999, and therefore we hypothesize that STAT1/6 function has a putative role in uveitis.

2.4.3. Cell Adhesion and Signal Transduction Related Genes

Matrix metalloproteinase (MMP) are key factors for the degradation of extracellular matrix components and modification of cytokines, protease inhibitors, and cell surface signaling systems [103,104,105,106]. Polymorphisms on the MMP-9 promoter can affect the development of visceral involvement in Korean people with BD [107]. In our RWR analysis, MMP9 had an MIS of 971 and a p-value < 0.001, which suggests that MMP-9 may be a novel susceptibility gene for uveitis.
VCAM1 (vascular cell adhesion molecule 1) belongs to the Ig superfamily and is a cell surface sialoglycoprotein expressed by cytokine-activated endothelium. This protein is mediated by leukocyte-endothelial cell adhesion and signal transduction [108,109]. VCAM1 can be regulated by inflammatory cytokines such as IL1B [108]. Uveitis is closely associated with the immune system and immune-related proteins including the interleukin family. In our study, the MIS of VCAM1 was 968 and the p-value was less than 0.05, which makes VCAM1 a candidate gene for uveitis.
We detected 56 novel uveitis-related genes using the RWR-based method. These genes can be clustered into three categories, shown in Figure 2. Among these 56 potential genes, eight (8/56, 14.3%) genes, IL13, IL1RN, JAK1, MYD88, NOD1, PTGS2, TGFB1 and TLR3, were considered as uveitis genes by experimental evidence [110,111,112,113,114,115,116,117], and 29 (29/56, 51.8%) genes (CASP8, CCL17, CCL20, CCL4, CD19, CD40 LG, CD80, CXCL1, CXCL11, CXCL13, CXCL16, CXCL5, CXCL9, CXCR2, IL15, IL1A, IL1B, IL23A, IL2RA, IL4, JUN, MMP9, STAT1, STAT6, TLR2, TLR7, TLR9, TNFRSF1A and VCAM1) had a correlation with uveitis. However, the pathogenesis is not clear. In our results, we found that these genes have a significant relationship with uveitis genes and therefore need more validation. There are few reports of the rest of the genes (19/56, 33.9%) (CCR8, CD276, CD8A, CSF2, CSF3, CXCR1, CXCR5, CXCR6, GAPDH, GZMB, HLA-DB5, IL5, IL9, KITLG, MAPK8, PTPRC, SELE, TGFB3 and TGFBI) which participate in the process of uveitis. We considered that they might be novel uveitis genes and merit attention. We argue that some of these may be critical putative virulence genes for uveitis and could be interesting agents for the treatment of human uveitis.

2.5. Comparison of Other Methods

The results listed in Section 2.1, Section 2.2, Section 2.3 and Section 2.4 can partly prove the effectiveness of the RWR-based method. In this section, we compared our method with a classic GBA-based method [30,31,32], i.e., a method like the nearest neighbor algorithm (NNA). This method identified novel genes from neighbors of the uveitis-related genes in a network. For convenience, we directly used the PPI network that was adopted in the RWR-based method. In addition, we called a neighbor of a node is a nearer neighbor if the edge between them was assigned a higher weight due to the definition of the interaction score reported in STRING. The GBA-based method selected the k nearest neighbors of each uveitis-related genes and collected them together as the predicted genes of the method, where k is a predefined parameter.
The five-fold cross-validation method was also adopted to test the GBA-based method, which used the same partition in testing the RWR-based method. Because we do not know the best value of k, we tried the following values: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100. The testing results are provided in Supplementary Table S2. The best performance of the GBA-based method with different parameter k on each part is shown in Table 3. Compared with the testing results of RWR-based method, also listed in Table 3 for convenience, we can see that GBA-based method provides higher recalls sometimes, however, it always provides lower precisions, indicating the GBA-based method can yield more false positive genes. If only considering the F1-measure, we can conclude that F1-measures of the RWR-based method are always higher than those of the GBA-based method. It is indicated that the RWR-based method is superior to GBA-based method for identification of uveitis-related genes.

3. Materials and Methods

3.1. Materials

Uveitis-related genes were collected from literatures indexed by PubMed ( The keywords “uveitis” and “genes” were used to search the literature in PubMed, which resulted in the collection of 744 papers. Among them, 98 review papers that generally summarized uveitis-related genes were manually reviewed. From those 98 papers, 121 genes were chosen from 96 reviews reporting the functional genes that may be important for uveitis or for specific uveitis symptoms. These genes are provided in Supplementary Table S3. In total, 146 Ensembl IDs for these genes were also determined and are provided in Supplementary Table S3.

3.2. Protein-protein Interaction (PPI) Network

PPIs are useful for the investigation of genetic disorders because they play an essential role in intracellular and intercellular biochemical processes. Many computational methods have been developed using this information, such as the prediction engines for the identification of protein functions [118,119,120] and methods for identification of novel disease genes [36,37,38]. Several methods were built based on the hypothesis that two proteins in a PPI are more likely to share similar functions. Thus, we can infer novel genes related to uveitis using PPI information and the uveitis-related genes mentioned in Section 3.1.
In this study, we used the PPI information retrieved from STRING (Search Tool for the Retrieval of Interacting Genes/Proteins, Version 9.1, [121] to construct the PPI network that the RWR algorithm can be applied. To access the PPI information in STRING, we downloaded the file “protein.links.v9.1 txt.gz”. Because “9606” is the organism code for the human interactome in STRING, lines in this file that started with “9606” were extracted, obtaining 2,425,314 human PPIs involving 20,770 proteins. According to STRING, these PPIs were derived from the following four sources: (1) genomic context; (2) high-throughput experiments; (3) (conserved) co-expression; and (4) previous knowledge. Thus, the information in STRING contained both the direct (physical) and the indirect (functional) association between proteins, therefore STRING could widely measure the associations between proteins. Each PPI contained two Ensembl IDs and one score that ranged between 150 and 999, which indicated the strength of the interaction. An interaction with a high score meant this interaction has a high probability of occurring. For each interaction containing proteins pa and pb, the score was denoted by S(pa,pb). The PPI network defined the 20,770 proteins as the nodes, and two nodes were adjacent if and only if their corresponding proteins can form a PPI. Additionally, each edge in the network represented a PPI; thus, we assigned a weight to each edge, which was defined as the score of its corresponding PPI. From our analysis, a PPI network containing 20,770 nodes and 2,425,314 edges was obtained.

3.3. RWR-Based Method

The RWR algorithm was executed on the PPI network using validated genes as seed nodes to search possible genes. Then, a permutation test was executed to exclude false discoveries found by RWR. The remaining candidate genes with strong associations to validated genes were selected for further analysis. The pseudo-codes of the RWR-based method are listed in Table 4.

3.3.1. Searching Possible Genes Using the RWR Algorithm

RWR is a type of ranking algorithm [33]. Based on a seed node or a set of seed nodes, it simulates a walker that starts from the nodes and randomly walks in a network. Here, 146 Ensembl IDs listed in Supplementary Table S3 were deemed as seed nodes. Starting from these nodes, we attempted to discover novel nodes (genes) related to uveitis. In the beginning of the RWR algorithm, a 20,770-D vector P0 was constructed, in which each composition represented the probability that a node in the network was a uveitis-related gene. Because the 146 Ensembl IDs represented validated uveitis-related genes, their compositions in P0 were set to 1/146, while others were set to zero. Then, the RWR algorithm repeatedly updated this probability vector until it became stable. We designated Pi to represent the probability vector after the i-th step was executed. The probability vector was updated according to the following equation:
P i + 1 = ( 1 r ) A T P i + r P 0
where A represented the column-wise normalized adjacency matrix of the PPI network and r was set to 0.8. When P t + 1 P t L 1 < 10 6 , the update procedure was stopped, and Pt+1 was the output of the RWR algorithm.
According to the probability vector yielded by the RWR algorithm, some nodes received high probabilities. It was apparent that their corresponding genes are more likely to be uveitis-related genes. To avoid missing possible uveitis-related genes, we set a probability threshold of 10−5. The corresponding genes of these nodes were designated as RWR genes.
In this study, we used the RWR program on the heterogeneous network that was implemented in Matlab and proposed by Li and Patra [122]. The code can be downloaded at By setting the special values of some parameters, this program could be used to execute the RWR algorithm on a single network.

3.3.2. Excluding False Discoveries Using the Permutation Test

Based on the validated uveitis-related genes and RWR algorithm, new RWR genes were accessed. However, this result was influenced by the structure of the constructed PPI network, i.e., some RWR genes were selected due to the structure of the network and they were not necessarily unique to uveitis. Furthermore, if we randomly selected some nodes in the network as seed nodes of the RWR algorithm, these genes were still selected for and were therefore deemed as likely to be false positive. To control for these genes, a permutation test was executed. We randomly constructed 1000 Ensembl ID sets, denoted by E 1 , E 2 , , E 1000 , consisting of 146 Ensembl IDs. For each set, the Ensembl IDs were deemed seed nodes of the RWR algorithm. Each RWR gene was given a probability. Thus, there were 1000 probabilities for 1000 sets and one probability for 146 Ensembl IDs of the uveitis-related genes for each RWR gene. Then, a measurement, called the p-value, was counted for each RWR gene g, which was defined as:
p - value ( g ) = Θ 1000
where Θ represented the number of randomly constructed sets where the probability assigned to g was larger than that for the 146 Ensembl IDs of uveitis-related genes. Clearly, an RWR gene with a high p-value indicated that the gene was not specific for uveitis and should be discarded. RWR genes with p-values less than 0.05 were selected for further analysis as potential candidate genes for uveitis.

3.3.3. Selection of Core Genes by Associations with Validated Genes

We hypothesized that, of the candidate genes, some may have a strong correlation with uveitis. To further select core candidate genes, two criteria were designed. Candidate genes satisfying both criteria were selected for additional analysis. Candidate genes that had the strongest associations with uveitis-related genes were more likely to be novel uveitis-related genes. Thus, for each candidate gene g, we calculated the maximum interaction score (MIS) as follows:
M I S ( g ) = max { S ( g , g ) : g is   a   uveitis - related   gene }
A high MIS suggested that the candidate gene was closely related to at least one uveitis-related gene, indicating that it was a novel uveitis-related gene with a high probability. According to STRING, a score of 900 was the cut-off for the highest confidence level. Therefore, candidate genes with MISs larger than 900 were selected.
Validated uveitis-related genes have strong associations with specific gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Therefore, candidate genes that had similar associations with uveitis GO terms and KEGG pathways were more likely to be novel uveitis-related genes. We performed GO term (KEGG pathway) [123,124,125,126] enrichment analysis for candidate genes and uveitis-related genes. The representation of a gene g on all GO terms and KEGG pathways was encoded into a vector ES (g) using this theory. This vector can be obtained by an in-house program using the R function phyper. The R code used was “score <− −log10 (phyper (numWdrawn− 1, numW, numB, numDrawn, lower.tail = FALSE)),” where numW, numB, and numDrawn are the number of genes annotated to the GO term or KEGG pathway, the number of genes not annotated to the GO term or KEGG pathway, and the number of neighbors of gene g and numWdrawn is the number of neighbors of gene g that are also annotated to the GO term or KEGG pathway. The relativity of the two genes g and g on GO terms and KEGG pathways was measured by
Γ ( g , g ) = E S ( g ) E S ( g ) E S ( g ) E S ( g )
A high outcome of Equation (4) indicated that g and g have a similar relationship in terms of GO terms and KEGG pathways. For any candidate gene g, we calculated the maximum function score (MFS) using the following equation:
M F S ( g ) = max { Γ ( g , g ) : g is   a   uveitis - related   gene }
Candidate genes with high MFSs were selected. In this equation, we set 0.8 as the threshold of MFS to select essential candidate genes.

3.4. Methods for Testing RWR-Based Method

In this study, we designed the RWR-based method to identify novel uveitis-related genes. However, it is necessary to test its effectiveness in advance. Here, the five-fold cross-validation [127] was employed. In detail, 146 Ensembl IDs of uveitis-related genes were randomly and equally divided into five parts. Then, Ensembl IDs in each part were singled out in turn and other Ensembl IDs in the rest four parts were used as the seed nodes in the RWR-based method. For each part, the results yielded by a good identification method on the rest four parts should satisfy the following conditions: (I) the results can recover a high proportion of the Ensembl IDs in the part; and (II) the results cannot contain several Ensembl IDs that are not in the part. Thus, recall and precision were employed to evaluate the results yielded by the RWR-based method, which can be calculated by
{ recall = T P T P + F N precision = T P T P + F P
where TP represented the number of Ensembl IDs in the part that can be recovered by the method, FN represented the number of Ensembl IDs in the part that cannot be recovered by the method and FP represented the number of Ensembl IDs that were yielded by the method and not in the part. In addition, to evaluate the predicted results on the whole, the F1-measure was also adopted, which can be computed by
F 1 - measure = 2 × recall × precision recall + precision
It is clear that a high F1-measure means the good performance of the method.

4. Conclusions

This study presented a computational method to determine novel uveitis-related genes. Using the RWR algorithm and certain screening criteria, 56 putative genes were accessed. Extensive analysis of the obtained genes confirmed that several genes are associated with the pathogenesis of uveitis. We hope that the identified novel genes may be used as material to study uveitis and that the proposed method can be extended to other diseases.

Supplementary Materials

Supplementary materials can be found at


This study was supported by the National Natural Science Foundation of China (31371335), Natural Science Foundation of Shanghai (17ZR1412500).

Author Contributions

Lin Liu conceived and designed the experiments; Shiheng Lu, Yan Yan performed the experiments; Shiheng Lu, Zhen Li, Lei Chen and Jing Yang analyzed the data; Shiheng Lu, Yuhang Zhang, Shaopeng Wang contributed reagents/materials/analysis tools; Shiheng Lu wrote the paper.

Conflicts of Interest

The authors declare no conflict of interest.


  1. Miserocchi, E.; Foqliato, G.; Modorati, G.; Bandello, F. Review on the worldwide epidemiology of uveitis. Eur. J. Ophthalmol. 2013, 23, 705–717. [Google Scholar] [CrossRef] [PubMed]
  2. Fraga, N.A.; Oliveira, M.F.; Follador, I.; Rocha, B.O.; Rêgo, V.R. Psoriasis and uveitis: A literature review. An. Bras. Dermatol. 2012, 87, 877–883. [Google Scholar] [CrossRef] [PubMed]
  3. Kulkarni, P. Review: Uveitis and immunosuppressive drugs. J. Ocul. Pharmacol. Ther. 2001, 17, 181–187. [Google Scholar] [CrossRef] [PubMed]
  4. Murakami, S.; Inaba, Y.; Mochizuki, M.; Nakajima, A.; Urayama, A. A nation-wide survey on the occurrence of Vogt-Koyanagi-Harada disease in Japan. Jpn. J. Ophthalmol. 1994, 98, 389–392. [Google Scholar]
  5. Moorthy, R.S.; Inomata, H.; Rao, N.A. Vogt-Koyanagi-Harada syndrome. Surv. Ophthalmol. 1995, 39, 265–292. [Google Scholar] [CrossRef]
  6. Evereklioglu, C. Current concepts in the etiology and treatment of Behcet disease. Surv. Ophthalmol. 2005, 50, 297–350. [Google Scholar] [CrossRef] [PubMed]
  7. Chang, J.H.; Wakefield, D. Uveitis: A global perspective. Ocul. Immunol. Inflamm. 2002, 10, 263–279. [Google Scholar] [CrossRef] [PubMed]
  8. Priem, H.A.; Kijlstra, A.; Noens, L.; Baarsma, G.S.; de Laey, J.J.; Oosterhuis, J.A. HLA typing in birdshot chorioretinopathy. Am. J. Ophthalmol. 1988, 105, 182–185. [Google Scholar] [CrossRef]
  9. Chaput, F.; Amer, R.; Baglivo, E.; Touitou, V.; Kozyreff, A.; Bron, D.; Bodaghi, B.; LeHoang, P.; Bergstrom, C.; Grossniklaus, H.E.; et al. Intraocular T-cell Lymphoma: Clinical Presentation, Diagnosis, Treatment, and Outcome. Ocul. Immunol. Inflamm. 2016, 22, 1–10. [Google Scholar] [CrossRef] [PubMed]
  10. Kitazawa, K.; Nagata, K.; Yamanaka, Y.; Kuwahara, Y.; Lehara, T.; Kinoshita, S.; Sotozono, C. Diffuse Anterior Retinoblastoma with Sarcoidosis-Like Nodule. Case Rep. Ophthalmol. 2015, 6, 443–447. [Google Scholar] [CrossRef] [PubMed]
  11. Catala-Mora, J.; Parareda-Salles, A.; Vicuña-Muñoz, C.G.; Medina-Zurinaga, M.; Prat-Bartomeu, J. [Uveitis masquerade syndrome as a presenting form of diffuse retinoblastoma]. Arch. Soc. Esp. Oftalmol. 2009, 84, 477–480. [Google Scholar] [PubMed]
  12. All-Ericsson, C.; Economou, M.A.; Landau, I.; Seregard, S.; Träisk, F. Uveitis masquerade syndromes: Diffuse retinoblastoma in an older child. Acta Ophthalmol. Scand. 2007, 85, 569–570. [Google Scholar] [CrossRef] [PubMed]
  13. Jovanovic, S.; Jovanović, Z.; Paović, J.; Teperković, V.S.; Pesić, S.; Marković, V. Two cases of uveitis masquerade syndrome caused by bilateral intraocular large B-cell lymphoma. Vojnosanit. Pregl. 2013, 70, 1151–1154. [Google Scholar] [CrossRef] [PubMed]
  14. Shen, K.; Smith, S.V.; Lee, A.G. Acute myelogenous leukemia presenting with uveitis, optic disc edema, and granuloma annulare: Case report. Can. J. Ophthalmol. 2016, 51, e153–e155. [Google Scholar] [CrossRef] [PubMed]
  15. Miserocchi, E.; Cimminiello, C.; Mazzola, M.; Russo, V.; Modorati, G.M. New-onset uveitis during CTLA-4 blockade therapy with ipilimumab in metastatic melanoma patient. Can. J. Ophthalmol. 2015, 50, e2–e4. [Google Scholar] [CrossRef] [PubMed]
  16. Rosenbaum, J.T.; Pasadhika, S.; Crouser, E.D.; Choi, D.; Harrington, C.A.; Lewis, J.A.; Austin, C.R.; Diebel, T.N.; Vance, E.E.; Braziel, R.M.; et al. Hypothesis: Sarcoidosis is a STAT1-mediated disease. Clin. Immunol. 2009, 132, 174–183. [Google Scholar] [CrossRef] [PubMed]
  17. Hou, S.; Yang, Z.; Du, L.; Jiang, Z.; Shu, Q.; Chen, Y.; Li, F.; Zhou, Q.; Ohno, S.; Chen, R.; et al. Identification of a susceptibility locus in STAT4 for Behcet’s disease in Han Chinese in a genome-wide association study. Arthritis. Rheum. 2012, 64, 4104–4113. [Google Scholar] [CrossRef] [PubMed]
  18. Remmers, E.F.; Cosan, F.; Kirino, Y.; Ombrello, M.J.; Abaci, N.; Satorius, C.; Le, J.M.; Yang, B.; Korman, B.D.; Cakiris, A.; et al. Genome-wide association study identifies variants in the MHC class I, IL10, and IL23R-IL12RB2 regions associated with Behcet’s disease. Nat. Genet. 2010, 42, 698–702. [Google Scholar] [CrossRef] [PubMed]
  19. Mizuki, N.; Meguro, A.; Ota, M.; Ohno, S.; Shiota, T.; Kawagoe, T.; Ito, N.; Kera, J.; Okada, E.; Yatsu, K.; et al. Genome-wide association studies identify IL23R-IL12RB2 and IL10 as Behcet’s disease susceptibility loci. Nat. Genet. 2010, 42, 703–706. [Google Scholar] [CrossRef] [PubMed]
  20. Robinson, P.C.; Claushuis, T.A.; Cortes, A.; Martin, T.M.; Evans, D.M.; Leo, P.; Mukhopadhyay, P.; Bradbury, L.A.; Cremin, K.; Harris, J.; et al. Genetic dissection of acute anterior uveitis reveals similarities and differences in associations observed with ankylosing spondylitis. Arthritis Rheumatol. 2015, 67, 140–151. [Google Scholar] [CrossRef] [PubMed]
  21. Fang, J.; Chen, L.; Tang, J.; Hou, S.; Liao, D.; Ye, Z.; Wang, C.; Cao, Q.; Kijlstra, A.; Yang, P. Association Between Copy Number Variations of TLR7 and Ocular Behcet’s Disease in a Chinese Han Population. Investig. Ophthalmol. Vis. Sci. 2015, 56, 1517–1523. [Google Scholar] [CrossRef] [PubMed]
  22. Kirino, Y.; Zhou, Q.; Ishigatsubo, Y.; Mizuki, N.; Tugal-Tutkun, I.; Seyahi, E.; Özyazgan, Y.; Ugurlu, S.; Erer, B.; Abaci, N. Targeted resequencing implicates the familial Mediterranean fever gene MEFV and the toll-like receptor 4 gene TLR4 in Behcet disease. Proc. Natl. Acad. Sci. USA 2013, 110, 8134–8139. [Google Scholar] [CrossRef] [PubMed]
  23. Li, B.; Yang, P.; Chu, L.; Zhou, H.; Huang, X.; Zhu, L.; Kijlstra, A. T-bet expression in the iris and spleen parallels disease expression during endotoxin-induced uveitis. Graefe’s Arch. Clin. Exp. Ophthalmol. 2007, 245, 407–413. [Google Scholar] [CrossRef] [PubMed]
  24. Kirino, Y.; Bertsias, G.; Ishigatsubo, Y.; Mizuki, N.; Tugal-Tutkun, I.; Seyahi, E.; Ozyazgan, Y.; Sacli, F.S.; Erer, B.; Inoko, H.; et al. Genome-wide association analysis identifies new susceptibility loci for Behcet’s disease and epistasis between HLA-B*51 and ERAP1. Nat. Genet. 2013, 45, 202–207. [Google Scholar] [CrossRef] [PubMed]
  25. Jiang, Z.; Yang, P.; Hou, S.; Du, L.; Xie, L.; Zhou, H.; Kijlstra, A. IL-23R gene confers susceptibility to Behcet’s disease in a Chinese Han population. Ann. Rheum. Dis. 2010, 69, 1325–1328. [Google Scholar] [CrossRef] [PubMed]
  26. Smith, J.R.; Choi, D.; Chipps, T.J.; Pan, Y.; Zamora, D.O.; Davies, M.H.; Babra, B.; Powers, M.R.; Planck, S.R.; Rosenbaum, J.T. Unique gene expression profiles of donor-matched human retinal and choroidal vascular endothelial cells. Investig. Ophthalmol. Vis. Sci. 2007, 48, 2676–2684. [Google Scholar] [CrossRef] [PubMed]
  27. Li, Z.; Liu, B.; Maminishkis, A.; Mahesh, S.P.; Yeh, S.; Lew, J.; Lim, W.K.; Sen, H.N.; Clarke, G.; Buggage, R. Gene expression profiling in autoimmune noninfectious uveitis disease. J. Immunol. 2008, 181, 5147–5157. [Google Scholar] [CrossRef] [PubMed]
  28. Ohta, K.; Kikuchi, T.; Miyahara, T.; Yoshimura, N. DNA microarray analysis of gene expression in iris and ciliary body of rat eyes with endotoxin-induced uveitis. Exp. Eye Res. 2005, 80, 401–412. [Google Scholar] [CrossRef] [PubMed]
  29. Li, Z.; Mzhesh, S.P.; Liu, B.; Yeh, S.; Lew, J.; Lim, W.; Levy Clarke, G.; Buggage, R.; Nussenblatt, R.B. Gene Expression Profiling of Non-infectious Uveitis Patients Using Pathway Specific cDNA Microarray Analysis. Investig. Ophthalmol. Vis. Sci. 2007, 48, 1505. [Google Scholar] [CrossRef]
  30. Oliver, S. Guilt-by-association goes global. Nature 2000, 403, 601–603. [Google Scholar] [CrossRef] [PubMed]
  31. Oti, M.; Snel, B.; Huynen, M.A.; Brunner, H.G. Predicting disease genes using protein-protein interactions. J. Méd. Genet. 2006, 43, 691–698. [Google Scholar] [CrossRef] [PubMed]
  32. Krauthammer, M.; Kaufmann, C.A.; Conrad Gilliam, T.; Rzhetsky, A. Molecular triangulation: Bridging linkage and molecular-network information for identifying candidate genes in Alzheimer’s disease. Proc. Natl. Acad. Sci. USA 2004, 101, 15148–15153. [Google Scholar] [CrossRef] [PubMed]
  33. Kohler, S.; Bauer, S.; Horn, D.; Robinson, P.N. Walking the interactome for prioritization of candidate disease genes. Am. J. Hum. Genet. 2008, 82, 949–958. [Google Scholar] [CrossRef] [PubMed]
  34. Li, Y.; Li, J. Disease gene identification by random walk on multigraphs merging heterogeneous genomic and phenotype data. BMC Genom. 2012, 13, S27. [Google Scholar] [CrossRef] [PubMed]
  35. Jiang, R.; Gan, M.X.; He, P. Constructing a gene semantic similarity network for the inference of disease genes. BMC Syst. Biol. 2011, 5, S2. [Google Scholar] [CrossRef] [PubMed]
  36. Chen, L.; Hao, X.Z.; Huang, T.; Shu, Y.; Huang, G.H.; Li, H.P. Application of the shortest path algorithm for the discovery of breast cancer related genes. Curr. Bioinform. 2016, 11, 51–58. [Google Scholar] [CrossRef]
  37. Zhang, J.; Yang, J.; Yang, T.; Huang, T.; Shu, Y.; Chen, L. Identification of novel proliferative diabetic retinopathy related genes on protein-protein interaction network. Neurocomputing 2016, 217, 63–72. [Google Scholar] [CrossRef]
  38. Gui, T.; Dong, X.; Li, R.; Li, Y.; Wang, Z. Identification of Hepatocellular Carcinoma–Related Genes with a Machine Learning and Network Analysis. J. Comput. Biol. 2015, 22, 63–71. [Google Scholar] [CrossRef] [PubMed]
  39. Chen, L.; Yang, J.; Huang, T.; Kong, X.; Lu, L.; Cai, Y.D. Mining for novel tumor suppressor genes using a shortest path approach. J. Biomol. Struct. Dyn. 2016, 34, 664–675. [Google Scholar] [CrossRef] [PubMed]
  40. Chen, L.; Huang, T.; Zhang, Y.H.; Jiang, Y.; Zheng, M.; Cai, Y.D. Identification of novel candidate drivers connecting different dysfunctional levels for lung adenocarcinoma using protein-protein interactions and a shortest path approach. Sci. Rep. 2016, 6, 29849. [Google Scholar] [CrossRef] [PubMed]
  41. Chen, L.; Wang, B.; Wang, S.; Yang, J.; Hu, J.; Xie, Z.; Wang, Y.; Huang, T.; Cai, Y.D. OPMSP: A computational method integrating protein interaction and sequence information for the identification of novel putative oncogenes. Protein. Pept. Lett. 2016, 23, 1081–1094. [Google Scholar] [CrossRef] [PubMed]
  42. Oguz, A.K.; Yılmaz, S.T.; Oygür, Ç.Ş.; Çandar, T.; Sayın, I.; Kılıçoğlu, S.S.; Ergün, İ.; Ateş, A.; Özdağ, H.; Akar, N. Behcet’s: A Disease or a Syndrome? Answer from an Expression Profiling Study. PLoS ONE 2016, 11, e0149052. [Google Scholar] [CrossRef] [PubMed]
  43. Miyazawa, I.; Abe, T.; Narikawa, K.; Feng, J.; Misu, T.; Nakashima, I.; Fujimori, J.; Tamai, M.; Fujihara, K.; Itoyama, Y. Chemokine profile in the cerebrospinal fluid and serum of Vogt-Koyanagi-Harada disease. J. Neuroimmunol. 2005, 158, 240–244. [Google Scholar] [CrossRef] [PubMed]
  44. Yu, C.R.; Mahdi, R.R.; Oh, H.M.; Amadi-Obi, A.; Levy-Clarke, G.; Burton, J.; Eseonu, A.; Lee, Y.; Chan, C.C.; Egwuagu, C.E. Suppressor of cytokine signaling-1 (SOCS1) inhibits lymphocyte recruitment into the retina and protects SOCS1 transgenic rats and mice from ocular inflammation. Investig. Ophthalmol. Vis. Sci. 2011, 52, 6978–6986. [Google Scholar] [CrossRef] [PubMed]
  45. Abu El-Asrar, A.M.; Berghmans, N.; Al-Obeidan, S.A.; Mousa, A.; Opdenakker, G.; van Damme, J.; Struyf, S. The Cytokine Interleukin-6 and the Chemokines CCL20 and CXCL13 Are Novel Biomarkers of Specific Endogenous Uveitic Entities. Investig. Ophthalmol. Vis. Sci. 2016, 57, 4606–4613. [Google Scholar] [CrossRef] [PubMed]
  46. Hollenbaugh, D.; Grosmaire, L.S.; Kullas, C.D.; Chalupny, N.J.; Braesch-Andersen, S.; Noelle, R.J.; Stamenkovic, I.; Ledbetter, J.A.; Aruffo, A. The human T cell antigen gp39, a member of the TNF gene family, is a ligand for the CD40 receptor: Expression of a soluble form of gp39 with B cell co-stimulatory activity. EMBO J. 1992, 11, 4313–4321. [Google Scholar] [PubMed]
  47. Lane, P.; Traunecker, A.; Hubele, S.; Inui, S.; Lanzavecchia, A.; Gray, D. Activated human T cells express a ligand for the human B cell-associated antigen CD40 which participates in T cell-dependent activation of B lymphocytes. Eur. J. Immunol. 1992, 22, 2573–2578. [Google Scholar] [CrossRef] [PubMed]
  48. Noelle, R.J.; Ledbetter, J.A.; Aruffo, A. CD40 and its ligand, an essential ligand-receptor pair for thymus-dependent B-cell activation. Immunol. Today 1992, 13, 431–433. [Google Scholar] [CrossRef]
  49. Fanslow, W.C.; Srinivasan, S.; Paxton, R.; Gibson, M.G.; Spriggs, M.K.; Armitage, R.J. Structural characteristics of CD40 ligand that determine biological function. Semin. Immunol. 1994, 6, 267–278. [Google Scholar] [CrossRef] [PubMed]
  50. Howard, L.M.; Miga, A.J.; Vanderlugt, C.L.; Dal Canto, M.C.; Laman, J.D.; Noelle, R.J.; Miller, S.D. Mechanisms of immunotherapeutic intervention by anti-CD40L (CD154) antibody in an animal model of multiple sclerosis. J. Clin. Investig. 1999, 103, 281–290. [Google Scholar] [CrossRef] [PubMed]
  51. Casamayor-Palleja, M.; Khan, M.; MacLennan, I.C. A subset of CD4+ memory T cells contains preformed CD40 ligand that is rapidly but transiently expressed on their surface after activation through the T cell receptor complex. J. Exp. Med. 1995, 181, 1293–1301. [Google Scholar] [CrossRef] [PubMed]
  52. Stuber, E.; Strober, W.; Neurath, M. Blocking the CD40L-CD40 interaction in vivo specifically prevents the priming of T helper 1 cells through the inhibition of interleukin 12 secretion. J. Exp. Med. 1996, 183, 693–698. [Google Scholar] [CrossRef] [PubMed]
  53. Grewal, I.S.; Flavell, R.A. CD40 and CD154 in cell-mediated immunity. Annu. Rev. Immunol. 1998, 16, 111–135. [Google Scholar] [CrossRef] [PubMed]
  54. Ogard, C.; Sorensen, T.L.; Krogh, E. Increased CD40 ligand in patients with acute anterior uveitis. Acta Ophthalmol. Scand. 2005, 83, 370–373. [Google Scholar] [CrossRef] [PubMed]
  55. Balashov, K.E.; Smith, D.R.; Khoury, S.J.; Hafler, D.A.; Weiner, H.L. Increased interleukin 12 production in progressive multiple sclerosis: Induction by activated CD4+ T cells via CD40 ligand. Proc. Natl. Acad. Sci. USA 1997, 94, 599–603. [Google Scholar] [CrossRef] [PubMed]
  56. Ang, M.; Cheung, G.; Vania, M.; Chen, J.; Yang, H.; Li, J.; Chee, S.P. Aqueous cytokine and chemokine analysis in uveitis associated with tuberculosis. Mol. Vis. 2012, 18, 565–573. [Google Scholar] [PubMed]
  57. Deng, B.; Ye, Z.; Li, L.; Zhang, D.; Zhu, Y.; He, Y.; Wang, C.; Wu, L.; Kijlstra, A.; Yang, P. Higher Expression of NOD1 and NOD2 is Associated with Vogt-Koyanagi-Harada (VKH) Syndrome But Not Behcet’s Disease (BD). Curr. Mol. Med. 2016, 16, 424–435. [Google Scholar] [CrossRef] [PubMed]
  58. Yang, Y.; Qi, J.; Wang, Q.; Du, L.; Zhou, Y.; Yu, H.; Kijlstra, A.; Yang, P. Berberine suppresses Th17 and dendritic cell responses. Investig. Ophthalmol. Vis. Sci. 2013, 54, 2516–2522. [Google Scholar] [CrossRef] [PubMed]
  59. Lee, Y.Y.; Wu, W.J.; Huang, C.N.; Li, C.C.; Li, W.M.; Yeh, B.W.; Liang, P.I.; Wu, T.F.; Li, C.F. CSF2 Overexpression Is Associated with STAT5 Phosphorylation and Poor Prognosis in Patients with Urothelial Carcinoma. J. Cancer 2016, 7, 711–721. [Google Scholar] [CrossRef] [PubMed]
  60. Jucker, M.; Feldman, R.A. Identification of a new adapter protein that may link the common β subunit of the receptor for granulocyte/macrophage colony-stimulating factor, interleukin (IL)-3, and IL-5 to phosphatidylinositol 3-kinase. J. Biol. Chem. 1995, 270, 27817–27822. [Google Scholar] [PubMed]
  61. Bittorf, T.; Jaster, R.; Brock, J. Rapid activation of the MAP kinase pathway in hematopoietic cells by erythropoietin, granulocyte-macrophage colony-stimulating factor and interleukin-3. Cell Signal. 1994, 6, 305–311. [Google Scholar] [CrossRef]
  62. Kimura, A.; Rieger, M.A.; Simone, J.M.; Chen, W.; Wickre, M.C.; Zhu, B.M.; Hoppe, P.S.; O’Shea, J.J.; Schroeder, T.; Hennighausen, L. The transcription factors STAT5A/B regulate GM-CSF-mediated granulopoiesis. Blood 2009, 114, 4721–4728. [Google Scholar] [CrossRef] [PubMed]
  63. Mui, A.L.; Wakao, H.; Harada, N.; O’Farrell, A.M.; Miyajima, A. Interleukin-3, granulocyte-macrophage colony-stimulating factor, and interleukin-5 transduce signals through two forms of STAT5. J. Leukoc. Biol. 1995, 57, 799–803. [Google Scholar] [PubMed]
  64. Feldman, G.M.; Rosenthal, L.A.; Liu, X.; Hayes, M.P.; Wynshaw-Boris, A.; Leonard, W.J.; Hennighausen, L.; Finbloom, D.S. STAT5A-deficient mice demonstrate a defect in granulocyte-macrophage colony-stimulating factor-induced proliferation and gene expression. Blood 1997, 90, 1768–1776. [Google Scholar] [PubMed]
  65. Taheri, S.; Borlu, M.; Evereklioglu, C.; Ozdemir, S.Y.; Ozkul, Y. mRNA Expression Level of Interleukin Genes in the Determining Phases of Behcet’s Disease. Ann. Dermatol. 2015, 27, 291–297. [Google Scholar] [CrossRef] [PubMed]
  66. Liang, L.; Tan, X.; Zhou, Q.; Zhu, Y.; Tian, Y.; Yu, H.; Kijlstra, A.; Yang, P. IL-1β triggered by peptidoglycan and lipopolysaccharide through TLR2/4 and ROS-NLRP3 inflammasome-dependent pathways is involved in ocular Behcet’s disease. Investig. Ophthalmol. Vis. Sci. 2013, 54, 402–414. [Google Scholar] [CrossRef] [PubMed]
  67. Roberge, F.G.; de Smet, M.D.; Benichou, J.; Kriete, M.F.; Raber, J.; Hakimi, J. Treatment of uveitis with recombinant human interleukin-13. Br. J. Ophthalmol. 1998, 82, 1195–1198. [Google Scholar] [CrossRef] [PubMed]
  68. Marie, O.; Thillaye-Goldenberg, B.; Naud, M.C.; de Kozak, Y. Inhibition of endotoxin-induced uveitis and potentiation of local TNF-α and interleukin-6 mRNA expression by interleukin-13. Investig. Ophthalmol. Vis. Sci. 1999, 40, 2275–2282. [Google Scholar]
  69. Lemaitre, C.; Thillaye-Goldenberg, B.; Naud, M.C.; de Kozak, Y. The effects of intraocular injection of interleukin-13 on endotoxin-induced uveitis in rats. Investig. Ophthalmol. Vis. Sci. 2001, 42, 2022–2030. [Google Scholar]
  70. De Kozak, Y.; Omri, B.; Smith, J.R.; Naud, M.C.; Thillaye-Goldenberg, B.; Crisanti, P. Protein kinase C ζ (PKC ζ) regulates ocular inflammation and apoptosis in endotoxin-induced uveitis (EIU)—Signaling molecules involved in EIU resolution by PKC ζ inhibitor and interleukin-13. Am. J. Pathol. 2007, 170, 1241–1257. [Google Scholar] [CrossRef] [PubMed]
  71. Xia, Z.J.; Kong, X.L.; Zhang, P. [In vivo effect of recombined IL-15/Fc fusion protein on EAU]. Sichuan Da Xue Xue Bao Yi Xue Ban 2008, 39, 944–949. [Google Scholar] [PubMed]
  72. Choe, J.Y.; Lee, H.; Kim, S.G.; Kim, M.J.; Park, S.H.; Kim, S.K. The distinct expressions of interleukin-15 and interleukin-15 receptor α in Behcet’s disease. Rheumatol. Int. 2013, 33, 2109–2115. [Google Scholar] [CrossRef] [PubMed]
  73. Hou, S.; Liao, D.; Zhang, J.; Fang, J.; Chen, L.; Qi, J.; Zhang, Q.; Liu, Y.; Bai, L.; Zhou, Y.; et al. Genetic variations of IL17F and IL23A show associations with Behcet’s disease and Vogt-Koyanagi-Harada syndrome. Ophthalmology 2015, 122, 518–523. [Google Scholar] [CrossRef] [PubMed]
  74. Lim, W.K.; Fujimoto, C.; Ursea, R.; Mahesh, S.P.; Silver, P.; Chan, C.C.; Gery, I.; Nussenblatt, R.B. Suppression of immune-mediated ocular inflammation in mice by interleukin 1 receptor antagonist administration. Arch. Ophthalmol. 2005, 123, 957–963. [Google Scholar] [CrossRef] [PubMed]
  75. Lindner, E.; Weger, M.; Steinwender, G.; Griesbacher, A.; Posch, U.; Ulrich, S.; Wegscheider, B.; Ardjomand, N.; El-Shabrawi, Y. IL2RA gene polymorphism rs2104286 A>G seen in multiple sclerosis is associated with intermediate uveitis: Possible parallel pathways? Investig. Ophthalmol. Vis. Sci. 2011, 52, 8295–8299. [Google Scholar] [CrossRef] [PubMed]
  76. Nussenblatt, R.B.; Fortin, E.; Schiffman, R.; Rizzo, L.; Smith, J.; van Veldhuisen, P.; Sran, P.; Yaffe, A.; Goldman, C.K.; Waldmann, T.A.; et al. Treatment of noninfectious intermediate and posterior uveitis with the humanized anti-Tac mAb: A phase I/II clinical trial. Proc. Natl. Acad. Sci. USA 1999, 96, 7462–7466. [Google Scholar] [CrossRef] [PubMed]
  77. Cenit, M.C.; Marquez, A.; Cordero-Coma, M.; Fonollosa, A.; Adan, A.; Martinez-Berriotxoa, A.; Llorenc, V.; Diaz Valle, D.; Blanco, R.; Canal, J.; et al. Evaluation of the IL2/IL21, IL2RA and IL2RB genetic variants influence on the endogenous non-anterior uveitis genetic predisposition. BMC Med. Genet. 2013, 14, 52. [Google Scholar] [CrossRef] [PubMed]
  78. Niven, J.; Hoare, J.; McGowan, D.; Devarajan, G.; Itohara, S.; Gannage, M.; Teismann, P.; Crane, I. S100B Up-Regulates Macrophage Production of IL1β and CCL22 and Influences Severity of Retinal Inflammation. PLoS ONE 2015, 10, e0132688. [Google Scholar] [CrossRef] [PubMed]
  79. Charteris, D.G.; Lightman, S.L. Comparison of the expression of interferon gamma, IL2, IL4, and lymphotoxin mRNA in experimental autoimmune uveoretinitis. Br. J. Ophthalmol. 1994, 78, 786–790. [Google Scholar] [CrossRef] [PubMed]
  80. Shahram, F.; Nikoopour, E.; Rezaei, N.; Saeedfar, K.; Ziaei, N.; Davatchi, F.; Amirzargar, A. Association of interleukin-2, interleukin-4 and transforming growth factor-β gene polymorphisms with Behcet’s disease. Clin. Exp. Rheumatol. 2011, 29, S28–S31. [Google Scholar] [PubMed]
  81. Chang, J.H.; McCluskey, P.; Wakefield, D. Expression of toll-like receptor 4 and its associated lipopolysaccharide receptor complex by resident antigen-presenting cells in the human uvea. Investig. Ophthalmol. Vis. Sci. 2004, 45, 1871–1878. [Google Scholar] [CrossRef]
  82. Chang, J.H.; McCluskey, P.J.; Wakefield, D. Toll-like receptors in ocular immunity and the immunopathogenesis of inflammatory eye disease. Br. J. Ophthalmol. 2006, 90, 103–108. [Google Scholar] [CrossRef] [PubMed]
  83. Meguro, A.; Ota, M.; Katsuyama, Y.; Oka, A.; Ohno, S.; Inoko, H.; Mizuki, N. Association of the toll-like receptor 4 gene polymorphisms with Behcet’s disease. Ann. Rheum. Dis. 2008, 67, 725–727. [Google Scholar] [CrossRef] [PubMed]
  84. Song, G.G.; Choi, S.J.; Ji, J.D.; Lee, Y.H. Toll-like receptor polymorphisms and vasculitis susceptibility: Meta-analysis and systematic review. Mol. Biol. Rep. 2013, 40, 1315–1323. [Google Scholar] [CrossRef] [PubMed]
  85. Fang, J.; Hu, R.; Hou, S.; Ye, Z.; Xiang, Q.; Qi, J.; Zhou, Y.; Kijlstra, A.; Yang, P. Association of TLR2 gene polymorphisms with ocular Behcet’s disease in a Chinese Han population. Investig. Ophthalmol. Vis. Sci. 2013, 54, 8384–8392. [Google Scholar] [CrossRef] [PubMed]
  86. Liang, L.; Tan, X.; Zhou, Q.; Tian, Y.; Kijlstra, A.; Yang, P. TLR3 and TLR4 But not TLR2 are Involved in Vogt-Koyanagi-Harada Disease by Triggering Proinflammatory Cytokines Production Through Promoting the Production of Mitochondrial Reactive Oxygen Species. Curr. Mol. Med. 2015, 15, 529–542. [Google Scholar] [CrossRef] [PubMed]
  87. Chen, S.; Yan, H.; Sun, B.; Zuo, A.; Liang, D. Subretinal transfection of chitosan-loaded TLR3-siRNA for the treatment of experimental autoimmune uveitis. Eur. J. Pharm. Biopharm. 2013, 85, 726–735. [Google Scholar] [CrossRef] [PubMed]
  88. Fang, J.; Chen, L.; Tang, J.H.; Hou, S.P.; Liao, D.; Ye, Z.; Wang, C.K.; Cao, Q.F.; Kijlstra, A.; Yang, P.Z. Association Between Copy Number Variations of TLR7 and Ocular Behcet’s Disease in a Chinese Han Population. Investig. Ophthalmol. Vis. Sci. 2015, 56, 1517–1523. [Google Scholar] [CrossRef] [PubMed]
  89. Sakamoto, N.; Sekine, H.; Kobayashi, H.; Sato, Y.; Ohira, H. Association of the toll-like receptor 9 gene polymorphisms with Behcet’s disease in a Japanese population. Fukushima J. Med. Sci. 2012, 58, 127–135. [Google Scholar] [CrossRef] [PubMed]
  90. Cui, H.P.; Pei, Y.X.; Li, G.F.; Lou, Y.R. Effect of glucocorticoid on cytokines TLR9 and TLR7 in peripheral blood for patients with uveitis. Exp. Ther. Med. 2016, 12, 3893–3896. [Google Scholar] [CrossRef] [PubMed]
  91. Shaulian, E. AP-1—The Jun proteins: Oncogenes or tumor suppressors in disguise? Cell Signal. 2010, 22, 894–899. [Google Scholar] [CrossRef] [PubMed]
  92. Smeal, T.; Binetruy, B.; Mercola, D.; Grover-Bardwick, A.; Heidecker, G.; Rapp, U.R.; Karin, M. Oncoprotein-mediated signalling cascade stimulates c-Jun activity by phosphorylation of serines 63 and 73. Mol. Cell Biol. 1992, 12, 3507–3513. [Google Scholar] [CrossRef] [PubMed]
  93. Pulverer, B.J.; Kyriakis, J.M.; Avruch, J.; Nikolakaki, E.; Woodgett, J.R. Phosphorylation of c-jun mediated by MAP kinases. Nature 1991, 353, 670–674. [Google Scholar] [CrossRef] [PubMed]
  94. Qing, H.; Gong, W.; Che, Y.; Wang, X.; Peng, L.; Liang, Y.; Wang, W.; Deng, Q.; Zhang, H.; Jiang, B. PAK1-dependent MAPK pathway activation is required for colorectal cancer cell proliferation. Tumour Biol. 2012, 33, 985–994. [Google Scholar] [CrossRef] [PubMed]
  95. Wang, C.Y.; Chen, C.L.; Tseng, Y.L.; Fang, Y.T.; Lin, Y.S.; Su, W.C.; Chen, C.C.; Chang, K.C.; Wang, Y.C.; Lin, C.F. Annexin A2 silencing induces G2 arrest of non-small cell lung cancer cells through p53-dependent and -independent mechanisms. J. Biol. Chem. 2012, 287, 32512–32524. [Google Scholar] [CrossRef] [PubMed]
  96. Gonzalez-Villasana, V.; Gutierrez-Puente, Y.; Tari, A.M. Cyclooxygenase-2 utilizes Jun N-terminal kinases to induce invasion, but not tamoxifen resistance, in MCF-7 breast cancer cells. Oncol. Rep. 2013, 30, 1506–1510. [Google Scholar] [PubMed]
  97. Gao, L.; Huang, S.; Ren, W.; Zhao, L.; Li, J.; Zhi, K.; Zhang, Y.; Qi, H.; Huang, C. Jun activation domain-binding protein 1 expression in oral squamous cell carcinomas inversely correlates with the cell cycle inhibitor p27. Med. Oncol. 2012, 29, 2499–2504. [Google Scholar] [CrossRef] [PubMed]
  98. Song, X.; Tao, Y.G.; Deng, X.Y.; Jin, X.; Tan, Y.N.; Tang, M.; Wu, Q.; Lee, L.M.; Cao, Y. Heterodimer formation between c-Jun and Jun B proteins mediated by Epstein-Barr virus encoded latent membrane protein 1. Cell Signal. 2004, 16, 1153–1162. [Google Scholar] [CrossRef] [PubMed]
  99. Rocha, G.; Duclos, A.; Chalifour, L.E.; Baines, M.G.; Antecka, E.; Deschenes, J. Analysis of gene expression during experimental uveitis in the rabbit. Can. J. Ophthalmol. 1996, 31, 228–233. [Google Scholar] [PubMed]
  100. Turner, M.J.; DaSilva-Arnold, S.; Luo, N.; Hu, X.; West, C.C.; Sun, L.; Hall, C.; Bradish, J.; Kaplan, M.H.; Travers, J.B.; et al. STAT6-mediated keratitis and blepharitis: A novel murine model of ocular atopic dermatitis. Investig. Ophthalmol. Vis. Sci. 2014, 55, 3803–3808. [Google Scholar] [CrossRef] [PubMed]
  101. Tepper, R.I.; Levinson, D.A.; Stanger, B.Z.; Campos-Torres, J.; Abbas, A.K.; Leder, P. IL-4 induces allergic-like inflammatory disease and alters T cell development in transgenic mice. Cell 1990, 62, 457–467. [Google Scholar] [CrossRef]
  102. Amadi-Obi, A.; Yu, C.R.; Liu, X.; Mahdi, R.M.; Clarke, G.L.; Nussenblatt, R.B.; Gery, I.; Lee, Y.S.; Egwuagu, C.E. TH17 cells contribute to uveitis and scleritis and are expanded by IL-2 and inhibited by IL-27/STAT1. Nat. Med. 2007, 13, 711–718. [Google Scholar] [CrossRef] [PubMed]
  103. Malla, N.; Sjoli, S.; Winberg, J.O.; Hadler-Olsen, E.; Uhlin-Hansen, L. Biological and pathobiological functions of gelatinase dimers and complexes. Connect. Tissue Res. 2008, 49, 180–184. [Google Scholar] [CrossRef] [PubMed]
  104. Murphy, G.; Nagase, H. Progress in matrix metalloproteinase research. Mol. Asp. Med. 2008, 29, 290–308. [Google Scholar] [CrossRef] [PubMed]
  105. Sivak, J.M.; Fini, M.E. Mmps in the eye: Emerging roles for matrix metalloproteinases in ocular physiology. Prog. Retin. Eye Res. 2002, 21, 1–14. [Google Scholar] [CrossRef]
  106. Nagase, H.; Visse, R.; Murphy, G. Structure and function of matrix metalloproteinases and TIMPs. Cardiovasc. Res. 2006, 69, 562–573. [Google Scholar] [CrossRef] [PubMed]
  107. Lee, Y.J.; Kang, S.W.; Baek, H.J.; Choi, H.J.; Bae, Y.D.; Kang, E.H.; Lee, E.Y.; Lee, E.B.; Song, Y.W. Association between matrix metalloproteinase 9 promoter polymorphisms and Behcet’s disease. Hum. Immunol. 2010, 71, 717–722. [Google Scholar] [CrossRef] [PubMed]
  108. Quillard, T.; Coupel, S.; Coulon, F.; Fitau, J.; Chatelais, M.; Cuturi, M.C.; Chiffoleau, E.; Charreau, B. Impaired Notch4 activity elicits endothelial cell activation and apoptosis: Implication for transplant arteriosclerosis. Arterioscler. Thromb. Vasc. Biol. 2008, 28, 2258–2265. [Google Scholar] [CrossRef] [PubMed]
  109. Verginelli, F.; Adesso, L.; Limon, I.; Alisi, A.; Gueguen, M.; Panera, N.; Giorda, E.; Raimondi, L.; Ciarapica, R.; Campese, A.F.; et al. Activation of an endothelial Notch1-Jagged1 circuit induces VCAM1 expression, an effect amplified by interleukin-1β. Oncotarget 2015, 6, 43216–43229. [Google Scholar] [PubMed]
  110. Crosson, J.N.; Laird, P.W.; Debiec, M.; Bergstrom, C.S.; Lawson, D.H.; Yeh, S. Vogt-Koyanagi-Harada-like syndrome after CTLA-4 inhibition with ipilimumab for metastatic melanoma. J. Immunother. 2015, 38, 80–84. [Google Scholar] [CrossRef] [PubMed]
  111. Yu, C.R.; Kim, S.H.; Mahdi, R.M.; Egwuagu, C.E. SOCS3 deletion in T lymphocytes suppresses development of chronic ocular inflammation via upregulation of CTLA-4 and expansion of regulatory T cells. J. Immunol. 2013, 191, 5036–5043. [Google Scholar] [CrossRef] [PubMed]
  112. Shimizu, J.; Izumi, T.; Arimitsu, N.; Fujiwara, N.; Ueda, Y.; Wakisaka, S.; Yoshikawa, H.; Kaneko, F.; Suzuki, T.; Takai, K.; et al. Skewed TGFβ/Smad signalling pathway in T cells in patients with Behcet’s disease. Clin. Exp. Rheumatol. 2012, 30, S35–S39. [Google Scholar] [PubMed]
  113. Li, Q.; Sun, B.; Dastgheib, K.; Chan, C.C. Suppressive effect of transforming growth factor β1 on the recurrence of experimental melanin protein-induced uveitis: Upregulation of ocular interleukin-10. Clin. Immunol. Immunopathol. 1996, 81, 55–61. [Google Scholar] [CrossRef] [PubMed]
  114. Sharma, R.K.; Gupta, A.; Kamal, S.; Bansal, R.; Singh, N.; Sharma, K.; Virk, S.; Sachdeva, N. Role of Regulatory T Cells in Tubercular Uveitis. Ocul. Immunol. Inflamm. 2016, 1–10. [Google Scholar] [CrossRef] [PubMed]
  115. Fabiani, C.; Vitale, A.; Lopalco, G.; Iannone, F.; Frediani, B.; Cantarini, L. Different roles of TNF inhibitors in acute anterior uveitis associated with ankylosing spondylitis: State of the art. Clin. Rheumatol. 2016, 35, 2631–2638. [Google Scholar] [CrossRef] [PubMed]
  116. Hatemi, I.; Hatemi, G.; Pamuk, O.N.; Erzin, Y.; Celik, A.F. TNF-α antagonists and thalidomide for the management of gastrointestinal Behcet’s syndrome refractory to the conventional treatment modalities: A case series and review of the literature. Clin. Exp. Rheumatol. 2015, 33, S129–S137. [Google Scholar] [PubMed]
  117. Bharadwaj, A.S.; Schewitz-Bowers, L.P.; Wei, L.; Lee, R.W.; Smith, J.R. Intercellular adhesion molecule 1 mediates migration of Th1 and Th17 cells across human retinal vascular endothelium. Investig. Ophthalmol. Vis. Sci. 2013, 54, 6917–6925. [Google Scholar] [CrossRef] [PubMed]
  118. Hu, L.; Huang, T.; Shi, X.; Lu, W.C.; Cai, Y.D.; Chou, K.C. Predicting functions of proteins in mouse based on weighted protein-protein interaction network and protein hybrid properties. PLoS ONE 2011, 6, e14556. [Google Scholar] [CrossRef] [PubMed]
  119. Hu, L.L.; Huang, T.; Cai, Y.D.; Chou, K.C. Prediction of Body Fluids where Proteins are Secreted into Based on Protein Interaction Network. PLoS ONE 2011, 6, e22989. [Google Scholar] [CrossRef] [PubMed]
  120. Chen, L.; Zhang, Y.H.; Huang, T.; Cai, Y.D. Identifying novel protein phenotype annotations by hybridizing protein-protein interactions and protein sequence similarities. Mol. Genet. Genom. 2016, 291, 913–934. [Google Scholar] [CrossRef] [PubMed]
  121. Jensen, L.J.; Kuhn, M.; Stark, M.; Chaffron, S.; Creevey, C.; Muller, J.; Doerks, T.; Julien, P.; Roth, A.; Simonovic, M. STRING 8-a global view on proteins and their functional interactions in 630 organisms. Nucleic Acids Res. 2009, 37, D412–D416. [Google Scholar] [CrossRef] [PubMed]
  122. Li, Y.; Patra, J.C. Genome-wide inferring genephenotype relationship by walking on the heterogeneous network. Bioinformatics 2010, 26, 1219–1224. [Google Scholar] [CrossRef] [PubMed]
  123. Yang, J.; Chen, L.; Kong, X.; Huang, T.; Cai, Y.-D. Analysis of Tumor Suppressor Genes Based on Gene Ontology and the KEGG Pathway. PLoS ONE 2014, 9, e107202. [Google Scholar] [CrossRef] [PubMed]
  124. Huang, T.; Zhang, J.; Xu, Z.P.; Hu, L.L.; Chen, L.; Shao, J.L.; Zhang, L.; Kong, X.Y.; Cai, Y.D.; Chou, K.C. Deciphering the effects of gene deletion on yeast longevity using network and machine learning approaches. Biochimie 2012, 94, 1017–1025. [Google Scholar] [CrossRef] [PubMed]
  125. Zhang, J.; Xing, Z.; Ma, M.; Wang, N.; Cai, Y.-D.; Chen, L.; Xu, X. Gene Ontology and KEGG Enrichment Analyses of Genes Related to Age-Related Macular Degeneration. BioMed Res. Int. 2014, 2014, 450386. [Google Scholar] [CrossRef] [PubMed]
  126. Chen, L.; Zhang, Y.-H.; Zheng, M.; Huang, T.; Cai, Y.-D. Identification of compound-protein interactions through the analysis of gene ontology, KEGG enrichment for proteins and molecular fragments of compounds. Mol. Genet. Genom. 2016, 291, 2065–2079. [Google Scholar] [CrossRef] [PubMed]
  127. Kohavi, R. A study of cross-validation and bootstrap for accuracy estimation and model selection. In Proceedings of the 14th International joint Conference on artificial intelligence, Montreal, QC, Canada, 20–25 August 1995. [Google Scholar]
Figure 1. The sub-network of the large network containing Ensembl Identifications (IDs) of validated and putative uveitis-related genes. Blue nodes represent Ensembl IDs of validated uveitis-related genes. Green nodes represent Ensembl IDs of putative uveitis-related genes.
Figure 1. The sub-network of the large network containing Ensembl Identifications (IDs) of validated and putative uveitis-related genes. Blue nodes represent Ensembl IDs of validated uveitis-related genes. Green nodes represent Ensembl IDs of putative uveitis-related genes.
Ijms 18 01045 g001
Figure 2. Clustering results of the 56 novel genes according to their evidences for being novel uveitis-related genes. Among 56 novel genes, eight have experiment evidence, 29 have significant relationship with uveitis but without experiment evidence, while no evidence can be found for the rest genes.
Figure 2. Clustering results of the 56 novel genes according to their evidences for being novel uveitis-related genes. Among 56 novel genes, eight have experiment evidence, 29 have significant relationship with uveitis but without experiment evidence, while no evidence can be found for the rest genes.
Ijms 18 01045 g002
Table 1. The performance of the Random Walk with Restrart (RWR)-based method yielded by five-fold cross-validation.
Table 1. The performance of the Random Walk with Restrart (RWR)-based method yielded by five-fold cross-validation.
Index of PartRecallPrecisionF1-Measure
Table 2. Novel genes inferred by Random Walk with Restrart (RWR)-based method.
Table 2. Novel genes inferred by Random Walk with Restrart (RWR)-based method.
Ensembl IDGene SymbolDescriptionProbabilityp-ValueMISMFS
ENSP00000351671 bCCL20C-C motif chemokine ligand 201.65 × 10−4<0.0019990.841
ENSP00000250151 bCCL4C-C motif chemokine ligand 41.64 × 10−4<0.0019940.820
ENSP00000326432 cCCR8C-C motif chemokine receptor 88.90 × 10−5<0.0019510.816
ENSP00000313419 bCD19CD19 molecule2.15 × 10−4<0.0019470.837
ENSP00000320084 cCD276CD276 molecule1.91 × 10−4<0.0019550.823
ENSP00000359663 bCD40LGCD40 ligand1.97 × 10−4<0.0019990.839
ENSP00000264246 bCD80CD80 molecule2.18 × 10−4<0.0019990.820
ENSP00000283635 cCD8ACD8a molecule1.91 × 10−4<0.0019900.815
ENSP00000296871 cCSF2Colony stimulating factor 22.71 × 10−4<0.0019920.875
ENSP00000225474 cCSF3Colony stimulating factor 31.55 × 10−4<0.0019160.829
ENSP00000379110 bCXCL1C-X-C motif chemokine ligand 11.69 × 10−4<0.0019730.827
ENSP00000306884 bCXCL11C-X-C motif chemokine ligand 111.28 × 10−4<0.0019990.818
ENSP00000286758 bCXCL13C-X-C motif chemokine ligand 131.49 × 10−4<0.0019860.806
ENSP00000293778 bCXCL16C-X-C motif chemokine ligand 161.02 × 10−4<0.0019520.800
ENSP00000296027 bCXCL5C-X-C motif chemokine ligand 51.11 × 10−4<0.0019580.811
ENSP00000354901 bCXCL9C-X-C motif chemokine ligand 92.13 × 10−4<0.0019990.883
ENSP00000295683 cCXCR1C-X-C motif chemokine receptor 18.67 × 10−5<0.0019990.833
ENSP00000319635 bCXCR2C-X-C motif chemokine receptor 21.02 × 10−4<0.0019990.851
ENSP00000229239 cGAPDHGlyceraldehyde-3-phosphate dehydrogenase2.12 × 10−4<0.0019220.824
ENSP00000216341 cGZMBGranzyme B2.46 × 10−4<0.0019910.829
ENSP00000364114 cHLA-DRB5Major histocompatibility complex, class II, DR β 52.27 × 10−4<0.0019480.822
ENSP00000304915 aIL13Interleukin 131.31 × 10−4<0.0019990.813
ENSP00000296545 bIL15Interleukin 151.85 × 10−4<0.0019460.806
ENSP00000263339 bIL1AInterleukin 1 α1.82 × 10−4<0.0019960.820
ENSP00000263341 bIL1BInterleukin 1 β3.58 × 10−4<0.0019990.873
ENSP00000259206 aIL1RNInterleukin 1 receptor antagonist1.68 × 10−4<0.0019990.836
ENSP00000228534 bIL23AInterleukin 23 subunit Α2.87 × 10−4<0.0019980.844
ENSP00000369293 bIL2RAInterleukin 2 receptor subunit Α2.46 × 10−4<0.0019990.866
ENSP00000274520 cIL9Interleukin 91.27 × 10−4<0.0019650.806
ENSP00000360266 bJUNJun proto-oncogene, AP-1 transcription factor subunit3.22 × 10−4<0.0019990.831
ENSP00000361405 bMMP9Matrix metallopeptidase 91.70 × 10−4<0.0019710.833
ENSP00000379625 aMYD88Myeloid differentiation primary response 881.82 × 10−4<0.0019990.882
ENSP00000356346 cPTPRCProtein tyrosine phosphatase, receptor type C2.18 × 10−4<0.0019940.826
ENSP00000331736 cSELESelectin E1.46 × 10−4<0.0019780.830
ENSP00000354394 bSTAT1Signal transducer and activator of transcription 12.63 × 10−4<0.0019990.852
ENSP00000300134 bSTAT6Signal transducer and activator of transcription 61.77 × 10−4<0.0019990.804
ENSP00000221930 aTGFB1Transforming growth factor β 12.90 × 10−4<0.0019970.832
ENSP00000416330 cTGFBITransforming growth factor β induced1.91 × 10−4<0.0019170.813
ENSP00000260010 bTLR2Toll like receptor 22.25 × 10−4<0.0019680.888
ENSP00000370034 bTLR7Toll like receptor 71.26 × 10−4<0.0019260.819
ENSP00000353874 bTLR9Toll like receptor 91.55 × 10−4<0.0019580.854
ENSP00000294728 bVCAM1Vascular cell adhesion molecule 12.23 × 10−4<0.0019680.882
ENSP00000292174 cCXCR5C-X-C motif chemokine receptor 51.14 × 10−40.0019760.820
ENSP00000343204 aJAK1Janus kinase 11.21 × 10−40.0019990.818
ENSP00000162749 bTNFRSF1ATNF Receptor superfamily member 1A2.30 × 10−40.0019990.826
ENSP00000304414 cCXCR6C-X-C motif chemokine receptor 69.27 × 10−50.0029640.803
ENSP00000296795 aTLR3Toll like receptor 31.58 × 10−40.0029660.858
ENSP00000231454 cIL5Interleukin 51.13 × 10−40.0049910.803
ENSP00000222823 aNOD1Nucleotide binding oligomerization domain containing 17.72 × 10−50.0049910.866
ENSP00000231449 bIL4Interleukin 42.55 × 10−40.0059990.852
ENSP00000356438 aPTGS2Prostaglandin-endoperoxide synthase 21.92 × 10−40.0099720.864
ENSP00000219244 bCCL17C-C motif chemokine ligand 171.20 × 10−40.019840.808
ENSP00000351273 bCASP8Caspase 89.66 × 10−50.0279990.821
ENSP00000353483 cMAPK8Mitogen-activated protein kinase 81.03 × 10−40.0349250.847
ENSP00000228280 cKITLGKIT ligand9.60 × 10−50.0399580.810
ENSP00000238682 cTGFB3Transforming growth factor β 35.37 × 10−50.0499610.850
a: Genes with experiment evidence; b: Genes without experiment evidence but have significant relationship with uveitis; c: Genes without any evidence.
Table 3. Comparison of the RWR-based method and GBA-based method.
Table 3. Comparison of the RWR-based method and GBA-based method.
Index of PartRWR-Based MethodGBA-Based Method
RecallPrecisionF1-MeasureBest Value of kRecallPrecisionF1-Measure
Table 4. The pseudo-code of the RWR-based method.
Table 4. The pseudo-code of the RWR-based method.
RWR-Based Method
Input: Ensembl IDs of uveitis-related genes, a PPI network
Output: A number of putative uveitis-related genes
  • Execute the RWR algorithm on the PPI network using the Ensembl IDs of uveitis-related genes as seed nodes, yielding a probability for each gene in the network; genes with probabilities higher than 10−5 were selected and called RWR genes;
  • Execute a permutation test, producing the p-value for each RWR gene; select RWR genes with p-values less than 0.05; the remaining genes were called candidate genes;
  • For each candidate gene, calculate its MIS (cf. Equation (3)) and MFS (cf. Equation (5)); select candidate genes with MISs no less than 900 and MFSs larger than 0.8;
  • Output the remaining candidate genes as the putative uveitis-related genes.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (
Int. J. Mol. Sci. EISSN 1422-0067 Published by MDPI AG, Basel, Switzerland RSS E-Mail Table of Contents Alert
Back to Top