Prediction of osteoporosis candidate genes by computational disease-gene identification strategy

Liuwei Dihuang pill cures postmenopausal osteoporosis with kidney-Yin deficiency: Potential therapeutic targets identified based on gene expression profiling

This study aimed to investigate the potential therapeutic targets of Liuwei Dihuang pill (LDP) in the treatment of postmenopausal osteoporosis with kidney-Yin deficiency (PMO-KY).Gene expression data were downloaded from the GEO database, including 4 PMO-KY samples and 3 healthy postmenopausal controls from GSE56116, as well as 3 PMO-KY samples before LDP treatment and 3 PMO-KY samples after three months of LDP treatment from GSE57273. Limma package was used to identify differentially expressed genes (DEGs). Afterwards, the potential target genes of LDP (namely key DEGs) were identified according to the comparison of DEGs in PMO-KY group and the DEGs in LDP treatment groups. Subsequently, iRegulon plugin in Cytoscape software was used to predict potential transcription factors (TFs) that regulated the key DEGs, and Comparative Toxicogenomics Database was utilized to identify known PMO-related genes among the key DEGs.Totally, 202 and 2066 DEGs were identified between PMO-KY and controls, as well as after-treatment and before-treatment groups, respectively. Among them, 52 DEGs were up-regulated in PMO-KY but down-regulated after LDP treatment, and 8 TFs were predicted to these DEGs. Furthermore, 34 DEGs were down-regulated in PMO-KY but up-regulated after treatment, and 7 TFs were predicted to regulate these DEGs. Additionally, 43 of the 86 key DEGs were known PMO-related genes.NCOA3, TCF4, DUSP6, PELI2, and STX7 were predicted to be regulated by HOXA13. In the PMO-KY treatment, NCOA3, TCF4, DUSP6, PELI2, and STX7 might be the potential therapeutic targets of LDP. However, further investigation is required to confirm these genes.

Matrilin-3 inhibits chondrocyte hypertrophy as a bone morphogenetic protein-2 antagonist

Increased chondrocyte hypertrophy is often associated with cartilage joint degeneration in human osteoarthritis patients. Matrilin-3 knock-out (Matn3 KO) mice exhibit these features. However, the underlying mechanism is unknown. In this study, we sought a molecular explanation for increased chondrocyte hypertrophy in the mice prone to cartilage degeneration. We analyzed the effects of Matn3 on chondrocyte hypertrophy and bone morphogenetic protein (Bmp) signaling by quantifying the hypertrophic marker collagen type X (Col X) gene expression and Smad1 activity in Matn3 KO mice in vivo and in Matn3-overexpressing chondrocytes in vitro. The effect of Matn3 and its specific domains on BMP activity were quantified by Col X promoter activity containing the Bmp-responsive element. Binding of MATN3 with BMP-2 was determined by immunoprecipitation, solid phase binding, and surface plasmon resonance assays. In Matn3 KO mice, Smad1 activity was increased more in growth plate chondrocytes than in wild-type mice. Conversely, Matn3 overexpression in hypertrophic chondrocytes led to inhibition of Bmp-2-stimulated, BMP-responsive element-dependent Col X expression and Smad1 activity. MATN3 bound BMP-2 in a dose-dependent manner. Multiple epidermal growth factor (EGF)-like domains clustered together by the coiled coil of Matn3 is required for Smad1 inhibition. Hence, as a novel BMP-2-binding protein and antagonist in the cartilage extracellular matrix, MATN3 may have the inherent ability to inhibit premature chondrocyte hypertrophy by suppressing BMP-2/Smad1 activity.

A systems biological approach reveals multiple crosstalk mechanism between gram-positive and negative bacterial infections: an insight into core mechanism and unique molecular signatures

Background

Bacterial infections remain a major threat and a leading cause of death worldwide. Most of the bacterial infections are caused by gram-positive and negative bacteria, which are recognized by Toll-like receptor (TLR) 2 and 4, respectively. Activation of these TLRs initiates multiple pathways that subsequently lead to effective immune response. Although, both the TLRs share common signaling mechanism yet they may exhibit specificity as well, resulting in the release of diverse range of inflammatory mediators which could be used as candidate biomolecules for bacterial infections.

Results

We adopted systems biological approach to identify signaling pathways mediated by TLRs to determine candidate molecules associated with bacterial infections. We used bioinformatics concepts, including literature mining to construct protein-protein interaction network, prioritization of TLRs specific nodes using microarray data and pathway analysis. Our constructed PPI network for TLR 2 (nodes: 4091 and edges: 66068) and TLR 4 (node: 4076 and edges: 67898) showed 3207 common nodes, indicating that both the TLRs might share similar signaling events that are attributed to cell migration, MAPK pathway and several inflammatory cascades. Our results propose the potential collaboration between the shared signaling pathways of both the receptors may enhance the immune response against invading pathogens. Further, to identify candidate molecules, the TLRs specific nodes were prioritized using microarray differential expressed genes. Of the top prioritized TLR 2 molecules, 70% were co-expressed. A similar trend was also observed within TLR 4 nodes. Further, most of these molecules were preferentially found in blood plasma for feasible diagnosis.

Conclusions

The analysis reveals the common and unique mechanism regulated by both the TLRs that provide a broad perspective of signaling events in bacterial infections. Further, the identified candidate biomolecules could potentially aid future research efforts concerning the possibility in differential diagnosis of gram-positive and negative bacterial infections.

Candidate gene identification for systemic lupus erythematosus using network centrality measures and gene ontology

Systemic lupus erythematosus (SLE) commonly accredited as “the great imitator” is a highly complex disease involving multiple gene susceptibility with non-specific symptoms. Many experimental and computational approaches have been used to investigate the disease related candidate genes. But the limited knowledge of gene function and disease correlation and also lack of complete functional details about the majority of genes in susceptible locus, encumbrances the identification of SLE related candidate genes. In this paper, we have studied the human immunome network (undirected) using various graph theoretical centrality measures integrated with the gene ontology terms to predict the new candidate genes. As a result, we have identified 8 candidate genes, which may act as potential targets for SLE disease. We have also carried out the same analysis by replacing the human immunome network with human immunome signaling network (directed) and as an outcome we have obtained 5 candidate genes as potential targets for SLE disease. From the comparison study, we have found these two approaches are complementary in nature.

Long-term resveratrol treatment prevents ovariectomy-induced osteopenia in rats without hyperplastic effects on the uterus

Resveratrol (Res), a polyphenol that is abundant in many medicinal plants and is a selective oestrogen receptor modulator, exhibits multiple biological activities. In the present study, we determined whether Res prevents oestrogen deficiency-induced osteopenia and whether Res administration decreases pathological changes in the endometrium and lumen of the uterus compared with oestradiol replacement therapy (ERT). A total of sixty 3-4-month-old female Wistar rats were randomly divided into a sham-operated group (Sham) and five ovariectomy (OVX) subgroups, i.e. OVX rats as a control group (OVX); OVX rats receiving oestradiol valerate (ERT, 0·8 mg/kg); and OVX rats receiving Res 20, 40 and 80 mg/kg. Daily oral administration was initiated at week 2 after OVX for 12 weeks. A dose-response difference was observed in the effects of Res on bone mineral density (BMD) and trabecular microarchitecture. Only at the highest dose, bone loss was almost equivalent to that observed in the ERT group. The dose-response effects of Res on the biochemical parameters (alkaline phosphatase, IL-6, TNF-α and transforming growth factor-β1 concentrations in the serum as well as urinary Ca and P excretion) and the expressions of receptor activator of nuclear factor κB ligand (RANKL) and the RANKL:osteoprotegerin protein ratio in the femur were also observed. Furthermore, the thickening of the endometrium and the infiltration of lymphocytes were prevented in all the three Res-treated groups compared with the ERT group. In conclusion, Res treatment not only improves BMD and trabecular microarchitecture but also does not affect the uterus and Res might be a potential remedy for the treatment of postmenopausal osteoporosis.

Prioritization of retinal disease genes: an integrative approach

The discovery of novel disease-associated variations in genes is often a daunting task in highly heterogeneous disease classes. We seek a generalizable algorithm that integrates multiple publicly available genomic data sources in a machine-learning model for the prioritization of candidates identified in patients with retinal disease. To approach this problem, we generate a set of feature vectors from publicly available microarray, RNA-seq, and ChIP-seq datasets of biological relevance to retinal disease, to observe patterns in gene expression specificity among tissues of the body and the eye, in addition to photoreceptor-specific signals by the CRX transcription factor. Using these features, we describe a novel algorithm, positive and unlabeled learning for prioritization (PULP). This article compares several popular supervised learning techniques as the regression function for PULP. The results demonstrate a highly significant enrichment for previously characterized disease genes using a logistic regression method. Finally, a comparison of PULP with the popular gene prioritization tool ENDEAVOUR shows superior prioritization of retinal disease genes from previous studies. The java source code, compiled binary, assembled feature vectors, and instructions are available online at https://github.com/ahwagner/PULP.

A haplotype of MATN3 is associated with vertebral fracture in Chinese postmenopausal women: Peking Vertebral Fracture (PK-VF) study

The Matrilin3 gene (MATN3) encodes an extracellular matrix protein, which modulates chondrocyte differentiation. The aim of this study was to test for association of MATN3 polymorphisms with bone mineral density (BMD), fracture, vertebral fracture, bone turnover or 25-hydroxyvitamin D [25(OH)D] in postmenopausal women. A community-based population of 1488 postmenopausal women was randomly selected in Beijing. The history of fracture and vertebral fracture was obtained via questionnaire and vertebral X-ray respectively. BMD of lumbar spine (2-4), femoral neck and total hip were measured by dual energy X-ray absorptiometry. Serum N-terminal procollagen of type 1 collagen (P1NP), β-isomerized type I collagen C-telopeptide breakdown products (β-CTX) and 25(OH)D were quantified. Binary logistic regression revealed that Haplotype-4 was significantly associated with vertebral fracture risk in both additive model (p=0.023, OR=1.521) and dominant model (p=0.028, OR=1.623). The significance remained after 10,000 permutation tests to correct multiple testing (p=0.042). Re-selected age matched vertebral fracture case-control groups revealed similar associations in additive model (p=0.014, OR=1.927, 95%CI=1.142-3.253) and in dominant model (p=0.011, OR=2.231, 95%CI=1.200-4.148). However, no significant association was found between MATN3 polymorphisms and serum β-CTX, P1NP, 25(OH)D levels, or BMD. In linear regression, Haplotype-2 approached marginal significance in association with femoral neck BMD T-score (p=0.050), but this would account for only 0.2% of BMD variation in our sample. This study suggests that Haplotype-4 of MATN3 is associated with vertebral fracture risk independent of BMD in Chinese postmenopausal women. Efforts should be made to replicate our finding in other, similar and ethnically diverse, populations.

Computational approaches to disease-gene prediction: rationale, classification and successes

The identification of genes involved in human hereditary diseases often requires the time-consuming and expensive examination of a great number of possible candidate genes, since genome-wide techniques such as linkage analysis and association studies frequently select many hundreds of 'positional' candidates. Even considering the positive impact of next-generation sequencing technologies, the prioritization of candidate genes may be an important step for disease-gene identification. In this paper we develop a basic classification scheme for computational approaches to disease-gene prediction and apply it to exhaustively review bioinformatics tools that have been developed for this purpose, focusing on conceptual aspects rather than technical detail and performance. Finally, we discuss some past successes obtained by computational approaches to illustrate their beneficial contribution to medical research.

A methodology based on molecular interactions and pathways to find candidate genes associated to diseases: its application to schizophrenia and Alzheimer's disease

Experimental techniques for the identification of genes associated with diseases are expensive and have certain limitations. In this scenario, computational methods are useful tools to identify lists of promising genes for further experimental verification. This paper describes a flexible methodology for the in silico prediction of genes associated with diseases combining the use of available tools for gene enrichment analysis, gene network generation and gene prioritization. A set of reference genes, with a known association to a disease, is used as bait to extract candidate genes from molecular interaction networks and enriched pathways. In a second step, prioritization methods are applied to evaluate the similarities between previously selected candidates and the set of reference genes. The top genes obtained by these programs are grouped into a single list sorted by the number of methods that have selected each gene. As a proof of concept, top genes reported a few years ago in SzGene and AlzGene databases were used as references to predict genes associated to schizophrenia and Alzheimer's disease, respectively. In both cases, we were able to predict a statistically significant amount of genes belonging to the updated lists.

Identification of genes for bone mineral density variation by computational disease gene identification strategy

We previously used five freely available bioinformatics tools (Prioritizer, Geneseeker, PROSPECTR and SUSPECTS, Disease Gene Prediction, and Endeavour) to analyze the thirteen well-replicated osteoporosis susceptibility loci and identify a subset of most likely candidate osteoporosis susceptibility genes (Huang et al. in J Hum Genet 53:644-655, 2008). In the current study, we experimentally tested the association between bone mineral density (BMD) and the 9 most likely candidate genes [LAMC2(1q25-q31), MATN3(2p24-p23), ITGAV(2q31-q32), ACVR1(2q23-q24), TDGF1(3p21.31), EGF(4q25), IGF1(12q22-q23), ZIC2(13q32), BMP2(20p12)] which were pinpointed by 4 or more bioinformatics tools. Forty tag SNPs in nine candidate genes were genotyped in a southern Chinese female case-control cohort consisting of 1643 subjects. Single- and multi-marker association analyses were performed using logistic regression analysis implemented by PLINK. Potential transcription factor binding sites were predicted by MatInspector. The strongest association was observed between rs10178256 (MATN3) and trochanter (P < 0.001) and total hip BMD (P = 0.002). The SNP rs6214 (IGF1) showed consistent association with BMD at all the four measured skeletal sites (P = 0.005-0.044). Prediction of transcription factor binding suggested that the minor allele G of rs10178256 might abolish the binding of MESP1 and MESP2 which play vital roles in bone homeostasis, whereas the minor allele G of rs6214 might create an additional binding site for XBP1, a constitutive regulator of endoplasmic reticulum stress response. Our data suggested that variants in MATN3 and IGF1 were involved in BMD regulation in southern Chinese women.

Bone mineral density is linked to 1p36 and 7p15-13 in a southern Chinese population

Genome-wide linkage scans have identified a number of quantitative trait loci (QTLs) affecting bone mineral density (BMD), mainly in the Caucasian population. In this study, we aim to determine whether seven well-replicated QTLs also contribute to BMD variation in the southern Han Chinese population. Thirty-three microsatellite markers in the proximity of seven QTLs were genotyped in 1,459 subjects from 306 families ascertained through a proband with BMD Z-score equal to or less than -1.3 at either the lumbar spine or hip. Regression-based multipoint linkage analysis was performed. In the entire study population, good linkage evidence of total hip BMD to 7p14 [maximum log of odds (LOD) score (MLS) = 2.75; nominal P = 0.0002] and 1p36 (MLS = 1.6, P = 0.003) was revealed. In the subgroup analysis of 1,166 female subjects, MLS of 3.42, 2.65, 2.42, and 1.54 were obtained on 7p12 for total hip, lumbar spine, trochanter, and femoral neck BMD, respectively. A suggestive linkage signal was achieved at 7p14-15 with a MLS of 3.38 and 3.15 for trochanter and total hip BMD in the 678 premenopausal women, and at 7p12 for femoral neck and total hip BMD with MLS of 2.22 and 3.04 in postmenopausal women. Subgroup analysis of premenopausal women also provided additional evidence of suggestive linkage of total hip BMD to 1p36, with a MLS of 2.84 at 17.07 cM. Thus, linkage of BMD to 1p36 and 7p15-13 is confirmed in southern Chinese. Computational prioritization strategy and published genome-wide association studies suggested RERE and SFRP4 as two promising candidate genes in which variants responsible for the linkage signal may be identified by follow-up gene-wide association studies.

Common variants in FLNB/CRTAP, not ARHGEF3 at 3p, are associated with osteoporosis in southern Chinese women

SUMMARY

We performed an association study of five candidate genes within chromosome 3p14-25 in 1,080 Chinese female subjects. Polymorphisms in FLNB/CRTAP are associated with bone mineral density (BMD) in Chinese.

INTRODUCTION

Chromosomal region 3p14-25 has shown strong evidence of linkage to BMD in genome-wide linkage scans. The variants responsible for this linkage signal, nonetheless, remain obscure.

METHODS

Thirty SNPs in five positional and functional candidate genes within 3p14-25 (PPARG, CRTAP, TDGF1, PTHR1, and FLNB) and rs7646054 in the ARHGEF3 gene were genotyped in a case-control cohort of 1,080 Chinese females. Allelic and haplotypic association were tested using logistic regression analysis implemented in PLINK software. Potential transcription factor binding sites were predicted with MatInspector.

RESULTS

Multiple SNPs and haplotypes in FLNB were significantly associated with BMDs, with the strongest association between lumbar spine BMD and rs9828717 (p = 0.005). SNP rs7623768 and the haplotype G-C of rs4076086-rs7623768 in CRTAP were associated with femoral neck BMD (p = 0.009 and p = 0.003, respectively). PTHR1 showed haplotypic associations with lumbar spine and femoral neck BMD (p = 0.02 and p = 0.044, respectively). Nevertheless, the association between rs7646054 in ARHGEF3 and BMD observed in Caucasians was not replicated in our samples. Comparative genomics analysis indicated that rs9828717 is located within a highly conserved region. The minor T allele at rs9828717 may lead to loss of binding site for nuclear factor of activated T cells which binds and triggers the transcriptional program of osteoblasts.

CONCLUSIONS

Our data suggest that variants in FLNB and CRTAP at 3p are involved in BMD regulation in southern Chinese.

Outcome of array CGH analysis for 255 subjects with intellectual disability and search for candidate genes using bioinformatics

Array CGH enables the detection of pathogenic copy number variants (CNVs) in 5-15% of individuals with intellectual disability (ID), making it a promising tool for uncovering ID candidate genes. However, most CNVs encompass multiple genes, making it difficult to identify key disease gene(s) underlying ID etiology. Using array CGH we identified 47 previously unreported unique CNVs in 45/255 probands. We prioritized ID candidate genes using five bioinformatic gene prioritization web tools. Gene priority lists were created by comparing integral genes from each CNV from our ID cohort with sets of training genes specific either to ID or randomly selected. Our findings suggest that different training sets alter gene prioritization only moderately; however, only the ID gene training set resulted in significant enrichment of genes with nervous system function (19%) in prioritized versus non-prioritized genes from the same de novo CNVs (7%, p < 0.05). This enrichment further increased to 31% when the five web tools were used in concert and included genes within mitogen-activated protein kinase (MAPK) and neuroactive ligand-receptor interaction pathways. Gene prioritization web tools enrich for genes with relevant function in ID and more readily facilitate the selection of ID candidate genes for functional studies, particularly for large CNVs.

Refined QTLs of osteoporosis-related traits by linkage analysis with genome-wide SNPs: Framingham SHARe

Genome-wide association studies (GWAS) using high-density array of single-nucleotide polymorphisms (SNPs) offer an unbiased strategy to identify new candidate genes for osteoporosis. We used a subset of autosomal SNPs from the Affymetrix 500K+50K SNP GeneChip marker set to examine genetic linkage with multiple highly heritable osteoporosis-related traits, including BMD of the hip and spine, heel ultrasound (attenuation and speed of sound), and geometric indices of the hip, in two generations from the Framingham Osteoporosis Study. Variance component linkage analysis was performed using normalized residuals (adjusted for age, height, BMI, and estrogen status in women). Multipoint linkage analyses produced LOD scores > or =3.0 for BMD on chromosomes (chr.) 9 and 11 and for ultrasound speed of sound on chr. 5. Hip geometric traits were linked with higher LOD scores, such as with shaft width on chr. 4 (LOD=3.9) and chr. 16 (LOD=3.8) and with shaft section modulus on chr. 22 (LOD=4.0). LOD score > or =5.0 was obtained for femoral neck width on chr. 7. In conclusion, with an SNP-based linkage approach, we identified several novel potential QTLs and confirmed previously identified chromosomal regions linked to bone mass and geometry. Subsequent focus on the spectrum of genetic polymorphisms in these refined regions may contribute to finding variants predisposing to osteoporosis.

Peroxisome proliferator-activated receptor gamma polymorphism is related to peak bone mass: the JPOS study

We analyzed 1,217 women to examine the effect of peroxisome proliferator-activated receptors gamma (PPARgamma) C161 --> T on bone status. Among 664 premenopausal women, the C161 --> T is associated with low bone mineral density (BMD) at the total hip and femoral neck. Moreover, the odds ratio for osteopenia or osteoporosis at the femoral neck was 1.98 for premenopausal CT/TT genotypes.

INTRODUCTION

The impact of PPARgamma on BMD has not been conclusively established. We examined if PPARgamma C161T polymorphism is associated with BMD and its change.

METHODS

We conducted a baseline survey in 1996 and a 10-year follow-up survey, Japanese Population-based Osteoporosis Study, with a sample population representative of Japanese women. Of these, 1,217 participants in the 1996 survey were analyzed cross-sectionally, while longitudinal analysis was performed on 563 women. A P value < 0.0042 (=0.05/12 for three menstrual statuses and four skeletal sites) was considered statistically significant after Bonferroni correction in multiple testing for cross-sectional analysis.

RESULTS

The total hip and femoral neck BMDs were significantly higher for CC genotype than for CT/TT genotypes among 664 premenopausal women (P = 0.0020, P = 0.0022, respectively). Compared to the CC genotype, the odds ratio for osteopenia or osteoporosis (T-scores below -1) at the femoral neck was 1.98 for premenopausal CT/TT genotypes with statistical significance (P = 0.0041). Change of BMD at either skeletal site during the follow-up period was not significantly different for either menstrual status.

CONCLUSIONS

We conclude that the PPARgamma C161T is associated with low peak bone mass.

Exploring Candidate Genes for Epilepsy by Computational Disease-Gene Identification Strategy

Epilepsy is a complex disease with a strong genetic component. So far, studies have focused on experimental validation or genome-wide linkage scans for epilepsy susceptibility genes in multiple populations. We have used four bioinformatic tools (SNPs3D, PROSPECTR and SUSPECTS, GenWanderer, PosMed) to analyze 16 susceptibility loci selected from a literature search. Pathways and regulatory network analyses were performed using the Ingenuity Pathways Analysis (IPA) software. We identified a subset of 48 candidate epilepsy susceptibility genes. Five significant canonical pathways, in four typical networks, were identified: GABA receptor signaling, interleukin-6 (IL-6) signaling, G-protein coupled receptor signaling, type 2 diabetes mellitus signaling and airway inflammation in asthma. We concluded that online analytical tools provide a powerful way to reveal candidate genes which can greatly reduce experimental time. Our study contributes to further experimental tests for epilepsy susceptibility genes.

Linking genes to diseases: it's all in the data

Genome-wide association analyses on large patient cohorts are generating large sets of candidate disease genes. This is coupled with the availability of ever-increasing genomic databases and a rapidly expanding repository of biomedical literature. Computational approaches to disease-gene association attempt to harness these data sources to identify the most likely disease gene candidates for further empirical analysis by translational researchers, resulting in efficient identification of genes of diagnostic, prognostic and therapeutic value. Existing computational methods analyze gene structure and sequence, functional annotation of candidate genes, characteristics of known disease genes, gene regulatory networks, protein-protein interactions, data from animal models and disease phenotype. To date, a few studies have successfully applied computational analysis of clinical phenotype data for specific diseases and shown genetic associations. In the near future, computational strategies will be facilitated by improved integration of clinical and computational research, and by increased availability of clinical phenotype data in a format accessible to computational approaches.