New Insights into Schizophrenia Disease Genes Interactome in the Human Brain: Emerging Targets and Therapeutic Implications in the Postgenomics Era

Abstract

Schizophrenia, a complex neurological disorder, is comprised of interactions between multiple genetic and environmental factors wherein each of the factors individually exhibits a small effect. In this regard a network-based strategy is best suited to capture the combined effect of multiple genes with their definite pattern of interactions. Given that schizophrenia affects multiple regions of the brain, we postulated that instead of any single specific tissue, a mutual set of interactions occurs between different regions of brain in a well-defined pattern responsible for the disease phenotype. To validate, we constructed and compared tissue specific co-expression networks of schizophrenia candidate genes in twenty diverse brain tissues. As predicted, we observed a common interaction network of certain genes in all the studied brain tissues. We examined fundamental network topologies of the common network to sequester essential common candidates for schizophrenia. We also performed a gene set analysis to identify the essential biological pathways enriched by the common candidates in the network. Finally, the candidate drug targets were prioritized and scored against known available schizophrenic drugs by molecular docking studies. We distinctively identified protein kinases as the top candidates in the network that can serve as probable drug targets for the disease. Conclusively, we propose that a comprehensive study of the connectivity amongst the disease genes themselves may turn out to be more informative to understand schizophrenia disease etiology and the underlying complexity.

Introduction

Schizophrenia is a complex neurological disorder where interactions between multiple genetic and environmental factors contribute to its etiology (Harrison et al., 2012; Kumarasinghe et al., 2012; Pelayo-Terán et al., 2012; Schultz et al., 2007; Sebat et al., 2009). Apart from direct influence of the genetic variants, several twin studies have shown that environmental factors such as low socioeconomic status, physical stress, and drug abuse can also modulate the functionality of the neurotransmitters and their receptors (Critchley et al., 2013; Evins et al., 2012; Fabi et al., 2013; Murray et al., 2013; Rabin et al., 2013; Van et al., 2088; Werner et al., 2007). In search of the genetic alterations, robust high-throughput technologies, including Genome Wide Association Studies (GWAS), have revealed numerous disease susceptibility loci. It is now clear that complex diseases such as schizophrenia result from a combined effect of multiple genes wherein each of the individual gene exhibits a small effect (Bergen et al., 2012; Schwab et al., 2013; Singh et al., 2014). Despite ample efforts worldwide, the research findings have not yet proved fruitful. Several association signals still await the biological interpretation in the disease pathophysiology. Therefore, an integrated systems biology approach is a prerequisite for understanding the disease complexity. A network-based strategy can best capture the combined effect of multiple genes along with specific pattern of interactions (Jia et al., 2011; Jiang et al., 2013; Sun et al., 2010; 2012).

Recent network-based studies done in schizophrenia and other neurological complex disorders focus on a protein–protein interaction network (PPIN) where a network is constructed using disease genes and their interacting partners or nearest neighbors from the human interactome (Camargo et al., 2007; Lee et al., 2011; Luo et al., 2014). Diverse approaches are being used to select candidate genes for network construction. The use of differentially expressed and co-expressed gene information is one of the popular strategies for network generation (Hsu et al., 2008; Sreedevi and Danail, 2012; Torkamani et al., 2010). Undoubtedly, these expression networks reveal interaction among important genes, but the approach is constrained by data normalization and false discovery rate (FDR) correction methods. Similarly, mapping GWAS identified disease-associated variants in the schizophrenia risk gene network highlights crucial nodes in the network (Gilman et al., 2012; Jia et al., 2012; Yu et al., 2014). However, the population sample size and stringent statistical parameters: p-value distribution, multiple testing correction methods, etc. warrant the utility of the method. Another method used is a gene prioritization approach, where candidate genes reported from genetic studies such as linkage, association, and gene expression are scored and used to structure a molecular gene network (Sun et al., 2009). However, the prioritization approach is also limited by accuracy of the scoring method. Recently, gene regulatory networks are also being employed to understand the disease complexity (Guo et al., 2010; Potkin et al., 2010). Regulatory networks are certainly meaningful in understanding the regulation of disease genes; still the strategy is constrained by valid experimental proof for regulation of disease genes.

Several histochemical studies reveal that schizophrenia primarily affects two core areas of the cerebral cortex: the parietal lobe and the frontal lobe (Huang et al., 2009; Premkumar et al., 2006; Tu et al., 2013; Yildiz et al., 2011). Schizophrenic patients exhibit significantly reduced expression of glutaminergic and dopaminergic receptors in the frontal lobe (Carlsson et al., 1990; Tseng et al., 2007), while the parietal lobe undergoes significant decrease in gray matter volume (Zhou et al., 2007). Also, corpus callosum (the white matter connecting the two cerebral hemispheres) changes in shape and size (Woodruff et al., 1995). Apart from this, reduction in the volume of temporal lobe has been associated with positive symptoms of schizophrenia (Andreasen et al., 1990; Barta et al., 1990, and the thalamus reduces in size leading to hypometabolism in schizophrenia (Andreasen et al., 1994).

In view of involvement of different regions of the brain, we hypothesize that instead of any individual specific tissue, a mutual set of interactions operate between these regions in a defined pattern that lead to a disease phenotype. To validate, we constructed a tissue specific co-expression network of schizophrenia candidate genes for various parts of the brain. As predicted, we found a set of interacting candidate pairs that are common in each tissue, suggestive of a shared interaction network for schizophrenia in brain. To support the above finding, we also studied the network topology of nodal genes in the common network in contrast to tissue-specific networks so as to identify probable drug targets in schizophrenia. In conclusion, we suggest that the disease genes operate via a distinct pattern while interacting with each other in different biological pathways. Understanding interactions amongst these candidate disease genes will be more meaningful to ascertain the underlying complexity of this disease.

Methods

Source of the dataset

Schizophrenia disease genes (SDG) were retrieved from four well-characterized public databases. The initial dataset was extracted from the OMIM MorbidMap (Amberger et al., 2009) consisting of 63 SDGs. The second dataset was derived from DisGeNet database (Bauer-Mehren et al., 2011), a comprehensive resource of gene-disease association that encompasses biomedical aspects of many diseases and provides vast information of disease genes. We found 553 SDG on searching the database with the keyword ‘schizophrenia.’ The third and fourth datasets were from NeuroDNet (Vasaikar et al., 2013) and Schizophrenia Gene Resource (SZGR) (Jia et al., 2010), two well-known resources for human neurological disorders and associated genes. A total number of 25 and 38 candidates SDG were found in these two databases, respectively. The compilation of SDG from all four datasets and on removal of redundant genes resulted in 556 unique disease genes for schizophrenia. Ethical approvals were not considered as the datasets were retrieved from publically available databases and no clinical samples have been used in this study.

Construction of disease gene interaction network

We constructed a schizophrenia disease gene co-expression network using a modified version of CRG interactome (Bossi and Lehner, 2009). CRG interactome is by far the largest protein–protein interaction network (PPIN) that integrates data from 21 known PPIN database resources to define a network of 80,922 physical interactions that can occur between 10,229 human proteins. Each pair of interactions was further extrapolated with their co-expression information in 78 different human tissues from GNF Atlas database (Su et al., 2004). An interaction is reported for a pair considering a normalized threshold expression value of 200. Any two genes are classified as co-expressed (1) if their normalized expression value is more than 200 and as not co-expressed (0) if the normalized expression is less than 200 (Bossi and Lehner, 2009). To confirm proper annotation, we retrieved Ensembl gene identifiers for 556 SDG and mapped them to CRG interactome. Resultant interactions were plotted as an undirected graph where each node is represented by the SDG and the two nodes are connected by an edge, if their partners are co-expressed in at least one tissue type in CRG data. We used an open source software platform, Cytoscape; v2.8.7 (Shannon et al., 2003) for visualization and further analysis of the network. The importance of a node in a molecular network is often correlated to its centrality (Jeong et al., 2001). There are different measures that capture the centrality of a node in a network. We investigated three different centrality measures namely Degree, Betweenness, and Closeness to prioritize the importance of SDG in the network. Details of these parameters have already been discussed in our earlier study (Podder et al., 2014).

All the measurements describing the network topology were calculated using NetworkAnalyzer (Assenov et al., 2008) which is a java based application (plugin) for Cytoscape.

Classification of housekeeping and non-housekeeping genes

Housekeeping (HK) genes are universally expressed in all tissues and their constitutive expression is essential to maintain the cellular functions (Butte et al., 2001). The expression levels of HK genes have been found to be constant in many diseased tissues as well as in normal tissues (Cheng et al., 2011; Morgan et al., 2010). Therefore, it is necessary to look at HK genes if they contribute fairly in disease pathophysiology. A list of human housekeeping genes (HK) was compiled from two different datasets by Chang et al., (2011) and Eisenberg and Levanon, (2013). In the first study (Chang et al., 2011), a meta-analysis has been performed, where genes from 104 different microarray datasets have been catalogued from 43 normal human tissues that identified 2064 housekeeping genes. On the other hand, in the second study (Eisenberg and Levanon, 2013), an expression profiling of publicly available RNA-seq data identified 3804 genes as HK genes whose uniform expression was observed across a panel of 16 normal human tissues. In the present work, we have considered the HK genes from both these studies. It was observed that less than 50% of HK genes overlapped in both studies. The redundant genes were removed, which resulted in a total of 4694 genes as HK in humans. Schizophrenia disease gene interactions were so classified based upon the presence of housekeeping (HK) disease genes and non-housekeeping (NHK) disease genes in the network.

Identification of candidate pathways

Biological pathways were identified from KEGG database (Kanehisa et al., 2012) using gene set enrichment analysis. We implemented the method described in ClueGo plugin (Bindea et al., 2009) for enrichment calculations. The p-values were calculated by a two-sided hypergeometric test followed by the Bonferroni correction (p≤0.05) for each KEGG term to determine the list of significant pathways (Rivals et al., 2007).

Identification of drug targets and docking studies

DrugBank3.0 (http://www.drugbank.ca/) (Knox et al., 2011) is currently the most comprehensive drug database that facilitates the study of multiple genes and drugs involved in a specific biological process. To identify the drugs related to schizophrenia, we used ‘schizophrenia’ as a search term and retrieved the specific list of drugs. A compilation of schizophrenic drugs together with their targets have been reported in a recent study (Putnam et al., 2011). We finally enlisted 48 drugs along with their targets. Additionally, docking studies were performed for a set of prospective drug targets in the network against all available schizophrenia drugs by using extra-precision (XP) mode of the Glide cross docking program (Schrödinger suite, Maestro version 9.3) (Friesner et al., 2006). Binding free energy (ΔG) for each docked pose was further estimated using MM-GBSA calculations (Prime, version 3.1) (Rastelli et al., 2010).

Results

Tissue-specific Disease Genes Interaction Network (DGNet)

We considered 556 schizophrenia disease genes (SDG) from the current databases to construct the disease gene co-expression network. Schizophrenia being a neurological disorder, we mapped SDG in twenty brain-related tissues in CRG data to compute the key interactions in different parts of human brain. In order to check the co-expression, an interaction pair was selected if both the genes in the SDG list correspond to a co-expression value of 1. We employed an in-house perl script to retrieve such interaction pairs in all the twenty tissues. Further, interaction pairs for each tissue were mapped together to form an undirected co-expression network of disease genes in schizophrenia (DGNet). We constructed twenty such DGNet. The numbers of disease genes and their interaction details for all the twenty tissues are presented in Figure 1. Majority of the interactions were seen in hypothalamus (206 SDG along with 322 interactions) followed by other tissues in brain. We noticed that the number of SDG (187–206) and their interactions (277–322) in each tissue did not differ remarkably, hence indicating that a set of SDG might share common interactions across those tissues.

FIG. 1.

SDG interaction details in twenty different tissues of brain. Tissues are represented in descending order based upon the number of SDG (Schizophrenia Disease Gene) present in each of them followed by their aggregate interaction statistics.

We also calculated the degree distribution (Diestel, 2005) of SDG in each of the twenty networks. The degree is the number of connections or edges that a SDG makes with any other node in the network. A comparison of degree distribution plot within the DGNet revealed that all of them (DGNet) followed the power law distribution that is a scale-free property of any biological network (Barabasi et al., 1999). We also observed that the average number of nodes having similar degree in the network is merely the same (Fig. 2) reflecting a commonality within the DGNet.

FIG. 2.

Degree distribution of DGNet in twenty different tissues of brain. The degree distribution curve of DGNet follows the power law distribution in each of the twenty tissues of brain. The comparison plot signified, the average number of nodes those shared similar degree across the networks are nearly overlapping with each other. The power curve shows the distribution of DGNet in hypothalamus with R² value 0.899.

Commonality within DGNet and expansion to (DG+NDG)Net

We evaluated the commonality within the SDG interaction pairs in all twenty tissues by an in-house developed perl script. We identified a total of 155 paired interactions among 113 SDG that were common to all twenty tissues. Based on the above information, a common interaction co-expression network of disease genes in schizophrenia was designed (Fig. 3). To investigate the diverse role of SDG, we further expanded the common DGNet in the human interactome where we included the interaction information of SDG as well as NSDG (Non-Schizophrenia Disease Genes) to structure the common ‘Disease Genes+Non-Disease Genes’ interaction Network (DG+NDG)Net. The mutual interactions between NSDG were ignored in the resultant network to reduce the complexity. Similar networks were also constructed in tissue specific manner for hypothalamus (Supplementary Fig. S1). We finally arrived at four separate but interrelated networks namely the DGNet and (DG+NDG)Net for hypothalamus and the common tissue, respectively, for further investigations. Details about the information on networks have been provided in Supplementary Figure S2. We observed that in (DG+NDG)Net, less than 8% genes belonged to SDG set, and rest of the genes were NDG (not for schizophrenia). Interestingly, a subsequent reduction of the network density from the tissue-specific network (hypothalamus) to the common interacting network of SDG also resulted in the same ratio.

FIG. 3.

Common interacting (co-expression) network of SDG in brain. (A) The common interacting DGNet comprises of 113 SDG and 155 defined interactions between them. The node size and color are drawn on a scale based on the centrality value of each node. (B) A snapshot of common (DG+NDG)Net displaying the whole interactome of 1419 genes and 2867 interactions in schizophrenia.

Comparison Between DGNet and (DG+NDG)Net

Network centrality measurement

Basic topological parameters of each network were taken for network centrality measurement. We calculated Node Degree, Betweeness Centrality (BC), and Closeness Centrality (CC) of each node in all four networks and compared the results. BC of a node calculates the number of shortest paths that passes through a particular node in the network (Brandes, 2001), whereas nodes with high CC value have the ability to contact any node in the network in the shortest possible path (Freeman, 1979). Altogether, these parameters evaluate the physical importance of any node in the network. In biological networks, genes with rich topological values mostly tend to work as a central regulatory “hub” (Vázquez et al., 2004). We plotted the BC and CC values together for the corresponding genes of both the DGNet and (DG+NDG)Net present in common interaction network. We found in both the cases that the genes with higher topological (BC and CC) values were mostly shared (Fig. 4). We further compared the degree value of each node between the networks in hypothalamus and the common interacting network of SDG (Table 1). We found certain genes like HSPA8, CSNK2A1, CALM1, ATXN1, SRSF1/6, YWHAB, and JUN were retained as highly connected nodes in all four networks. On decreasing the network density, the number of interactions made by each node altered without affecting the priority of nodes. Interestingly, we identified few known candidate genes for schizophrenia such as DISC1 and YWHAG as important members in hypothalamus network that lacked in the common interacting network of SDG.

FIG. 4.

Comparison of topological parameters between DGNet and (DG+NDG)Net. The Betweenness centrality (BC) plot (green triangle at the bottom) and the Closeness centrality (CC) plot (red cross at the top) in each graph shows the distribution of nodes in corresponding network. Topologically enriched nodes became common and highlighted in both networks.

Table 1.

Candidate Genes Prioritized by Comparing the Degree Values in Different Networks

Hypothalamus				Common tissues
(DG+NDG)Net		DGNet		(DG+NDG)Net		DGNet
Gene	Degree	Gene	Degree	Gene	Degree	Gene	Degree
YWHAG	222	YWHAG	19	HSPA8	136	HSPA8	10
HSPA8	200	CALM1	17	SRSF1	121	CSNK2A1	10
TP53	178	DISC1	14	HNRNPU	120	CALM1	9
SRSF1	164	HSPA8	13	SRSF6	109	ATXN1	9
CSNK2A1	163	YWHAZ	13	CSNK2A1	103	ACTG1	7
HNRNPU	155	CSNK2A1	13	SRSF7	92	CAMK2A	7
ESR1	152	ATXN1	12	ATXN1	84	DLG1	7
SRSF6	147	ESR1	12	HNRNPK	80	SRSF1	7
ATXN1	126	TP53	12	YWHAB	77	YWHAB	6
SRSF7	123	GRIN1	10	PPM1G	73	GAPDH	6
CALM1	118	DLG1	9	CALM1	64	YWHAE	6
HNRNPK	117	YWHAB	9	JUN	59	SRSF6	6
YWHAZ	116	SRSF1	8	ACTG1	58	PRNP	6
YWHAB	110	JUN	8	ATN1	50	JUN	6
PPM1G	99	CAMK2A	8	NR3C1	49	MAP1A	5
JUN	90	YWHAH	8	VIM	48	YWHAH	5
VIM	80	ACTG1	8	HNRNPA3	45	HNRNPU	5
ACTG1	79	YWHAE	8	GSK3B	43	NR3C1	5
AKT1	78	PICK1	7	YWHAE	42	APOE	4
NR3C1	73	PRNP	7	IPO5	39	GRIN2B	4
ATN1	69	STX1A	7	GAPDH	38	CAMK2B	4
HNRNPA3	66	SRSF6	7	EIF5	37	VIM	4
GSK3B	65	VIM	7	NDUFS3	34	STX1A	4
TRRAP	60	AKT1	7	DLG1	33	L1CAM	4
YWHAE	59	NR3C1	7	NDUFS6	33	PLEC	4

A comparison of node degree values between DGNet and (DG+NDG)Net in both hypothalamus and common interacting network of SDG. Degree distribution of top 25 genes in each network reveal that essential SDGs appear highly connected in all four different networks (highlighted in bold).

Presence of Housekeeping and Non-housekeeping Genes

To further investigate, we classified the DGNet and (DG+NDG)Net of common interacting network based on the presence of housekeeping (HK) and non-housekeeping (NHK) genes. In both cases, the distribution of HK genes clearly reflected that when we incorporate the interactions of NDG in DGNet of SDG the percentage of HK genes gradually increased from 30% to 66% (Fig. 5). On examination of the topologically enriched genes in both the common interacting networks (Fig. 4), we found that the top identified genes like HSPA8, CALM1, DLG1, and genes from SRSF families fall in the HK category, reflecting their universal role to maintain certain biological functions in each and every tissue of the brain. Also, throughout our study, we were able to isolate common NHK gene-rich interactions (CSNK2A1, CAMK2A, CAMK2B, ATXN1, and PPM1G) that were retained in both the common interacting network of brain.

FIG. 5.

Distribution of HK and NHK genes in DGNet and (DG+NDG)Net. 4694 human housekeeping genes (HK) were compiled from two different datasets (Chang et al., 2011; Eisenberg et al., 2013). Genes present in both the DGNet and (DG+NDG)Net were systematically classified into two separate groups: housekeeping (HK) and non-housekeeping (NHK). An increased ratio of HK genes was observed when the interactions of NDG were included with SDG.

Knowledge of candidate pathways

We next performed gene set enrichment analysis of DGNet present in both hypothalamus and common interacting network of SDG. Significant biological pathways for schizophrenia enriched by SDG in each network were identified from KEGG database (Table 2). The comparison chart clearly highlights important candidate pathways for schizophrenia such as amphetamine addiction (KEGG ID: 5031) (p=4.31E-11 and 1.62E-08) and dopaminergic synapse (KEGG ID: 4728) (p=1.36E-09 and 1.07E-05), respectively, in both the networks. Other essential pathways, such as neuroactive ligand-receptor interaction (KEGG ID: 4080), calcium signaling (KEGG ID: 4020), and neurotropin signaling (KEGG ID: 4722) were also sustained in both the networks throughout the process. The above observation led to the conclusion that identification of common interaction network of disease genes will help in prioritizing significant pathways similar to those obtained in tissue specific network. Hence, we suggest that the common DGNet is robust enough to capture the important pathways in brain that are generally affected in various tissues in schizophrenia. Interestingly, we also observed some significant overlapping gene sets with other disease pathways like Alzheimer's (KEGG ID: 5010) and amyotrophic lateral sclerosis (ALS) (KEGG ID: 5014). The overlapping SDG might act as ‘pleotropic genes’ that are characteristic of disease genes present in phenotypically similar complex disorders (Aerts et al., 2006).

Table 2.

Candidate Pathways Prioritized in Hypothalamus and Common Interacting Network of SDG

		DGNet (Hypothalamus)			DGNet (Common brain tissues)
KEGG ID	Pathways	Nr. Genes	% Associated Genes	p-value^*	Nr. Genes	% Associated Genes	p-value^*
5031	Amphetamine addiction	16	22.86	4.31E-11	11	15.71	1.62E-08
4728	Dopaminergic synapse	19	14.50	1.36E-09	11	8.46	1.07E-05
5030	Cocaine addiction	12	24.00	2.19E-08	7	14.00	5.03E-05
5033	Nicotine addiction	10	25.00	4.71E-07	NA	NA	NA
4080	Neuroactive ligand-receptor interaction	23	8.36	6.71E-07	9	4.40	0.03
4724	Glutamatergic synapse	15	12.71	1.35E-06	NA	NA	NA
5014	Amyotrophic lateral sclerosis (ALS)	10	18.52	1.00E-05	6	11.32	8.92E-04
4720	Long-term potentiation	11	15.49	1.54E-05	9	12.86	3.72E-06
4012	ErbB signaling pathway	12	13.64	1.76E-05	NA	NA	NA
5010	Alzheimer's disease	16	9.41	3.02E-05	13	7.74	2.69E-06
4713	Circadian entrainment	11	11.34	3.63E-04	NA	NA	NA
4020	Calcium signaling pathway	14	7.65	0.002	11	6.15	2.30E-04
4722	Neurotrophin signaling pathway	11	9.17	0.003	8	6.30	0.003
4151	PI3K-Akt signaling pathway	18	5.19	0.020	NA	NA	NA
4010	MAPK signaling pathway	15	5.77	0.022	NA	NA	NA
5016	Huntington's disease	12	6.56	0.028	8	4.35	0.03

p-value was calculated by two-sided hypergeometric test. Correction of p-value has been done using Bonferroni correction method. P-value less than 0.05 considered significant. The KEGG database search and analysis was performed using ClueGO (Cytoscape plugin).

Prioritization of drugs and drug-targets

To prioritize important drugs and their targets for the DGNet, we designed a drug–target interaction network based on available drug information for schizophrenia from DrugBank (Knox et al., 2011). We identified 48 known schizophrenia-specific drugs along with their targets. Further, each candidate SDG was matched with the targets in the DGNet. Two receptor protein encoded genes from dopamine family (DRD2) and serotonin family (HTR2A) were identified as major drug targets in the common DGNet. Currently, there are 37 and 32 drugs available for both the targets, respectively. Few G-protein encoded genes (GRIN1, GRIN2A, GRIN2B, and GRIA) were found to be targeted by the drug L-glutamic acid (DB00142). Additionally, calcium signaling protein (CALM1) was also identified to be targeted by the known drugs like fluphenazine (DB00623) and trifluoperazine (DB00831).

However, based on the existing databases in the public domain, we were unable to detect any drug for the novel candidate genes that were found to be topologically enriched in the identified common DGNet. Therefore, we employed a molecular docking strategy where the top ranked genes were docked against all available schizophrenia specific drugs from DrugBank. We performed a comprehensive search of all 113 genes of common DGNet in the Protein Data Bank (PDB) (Berman et al., 2000). A total of 23 crystal structures were identified, out of which only 11 were NHK genes. Information on the PDB structures and drug targets is listed in Supplementary Table S1. Top candidates identified from the common DGNet belonged to the protein kinase family (CSNK2A1, CAMK2A, and CAMK2B). Availability of the crystal structures of these protein kinases allowed us to perform molecular docking and followed by calculation of binding affinities with schizophrenia specific drugs (Fig. 6). We found that drugs such as lurasidone (DB08815), aripiprazole (DB01238), and bromocriptine (DB01200) exhibited greater binding affinity for all the three kinases (ΔG>− 80 kcal/mol). A few bridge drugs such as haloperidol (DB00502), trazodone (DB00656), and trifluoperazine (DB00831) also showed significant binding energy (ΔG>− 70 kcal/mol).

FIG. 6.

Molecular docking of top three kinases (CSNK2A1, CAMK2A, and CAMK2B) for all available schizophrenia drugs. Prioritization of 48 known schizophrenia-specific drugs based on molecular docking targeting top three proteins kinases as prospective drug targets. Red arrow indicates drugs having higher affinity for all the kinases in the study. Blue arrows depict a respected docking score and reveal the polypharmacological properties of few bridge drugs available in the market.

Discussion

A complex disorder is the result of interplay between disease and non-disease genes. Such interactions of genes often follow unique combinations and patterns based upon their presence in different pathways and tissues (Craig, 2008; Janssens et al., 2008). In this study, we have tried to identify the respective pattern of disease gene interactions in schizophrenia by constructing two different networks: one specific to co-expression information among known schizophrenia disease genes (SDG) named DGNet, and another by incorporating the co-expression information of non-disease gene (NDG) as well along with SDG, which we referred to as (DG+NDG)Net.

In order to resolve the complexity of neuropsychiatric disorders such as schizophrenia, we propose that it is essential to identify the common disease elements in various tissues across the brain. With the available SDG, we identified disease gene co-expression networks for twenty different tissues in human brain. Further, a subset of these networks was identified as common interacting network of SDG. The common network was validated with various computational analyses in order to establish its importance in disease process. We observed that majority of the SDG interactions were present in hypothalamus. Earlier studies have also reported the association of hypothalamus with schizophrenia, though its specific role in disease etiology is still not clear (Hans-Gert et al., 2010; Klomp et al., 2012; Tognin et al., 2012). In view of this, we suggest that the hypothalamus plays a crucial role in understanding the tissue specificity of the disease.

Topologically enriched genes are believed to be important regulatory elements in any biological network (Jeong et al., 2001). Most of the hub genes in the common interacting network were identified with high centrality value in both the tissue specific DGNet (hypothalamus) and the overall (DG+NDG)Net. Our data strongly reflect that important candidate genes for schizophrenia can be prioritized and detected based upon their co-expression and interaction patterns in all the brain tissues. Knowing that human HK genes are important to maintain normal biological functions however, ubiquitous expression of HK genes in both normal and disease tissues downgrade their prominence for further disease etiology (Butte et al., 2001). NHK genes are tissue specific, and alterations in NHK genes have repeatedly been shown to cause disease, thereby potential targets for drugs in any disease (Dezso et al., 2008). Perhaps an overall increase of HK genes in the (DG+NDG)Net indicated the adequacy of DGNet to represent the important NHK genes for schizophrenia. The local arrangement of these NHK genes in common interacting network of SDG is also important for the functionality of many candidate pathways.

We identified amphetamine addiction (KEGG ID: 5031) and dopaminergic synapse (KEGG ID: 4728) as most significant pathways enriched by SDG in the common DGNet. Undoubtedly, drug abuse has been a common concern in schizophrenia, though its role as a cause or as an effect of disease is still not clear (Bühler et al., 2002; Dixon et al., 1991). Prolonged use of amphetamine, cocaine, and similar drugs enhance dopamine levels in brain and lead to symptoms which resemble positively with those in schizophrenia (Jacobs and Silverstone, 1986). Chemical imbalance in brain due to malfunctioning of the dopaminergic system has been shown as a fundamental cause of many neurological diseases. A highly debatable hypothesis—“Dopamine hypothesis”, proposes dopamine to be a causative agent for schizophrenia, and drugs that increase dopamine activity worsen schizophrenic's symptoms (Abi-Dargham, 2004; Howes and Kapur, 2009; Lau et al., 2013). It has also been established that drugs like amphetamine increase dopamine activity, which in turn is the key contributing factor for the etiology of schizophrenia.

However, it is not clear how the corresponding pathways of these molecules crosstalk with each other in the disease process (Boileau et al., 2013; Seeman, 2013; Sunday, 2011). We speculated that there should be a set of interacting SDG that share common pattern of interactions across diverse pathways associated with schizophrenia. We mapped down a subset of interactions for dopaminergic synapse (KEGG ID: 4728) and amphetamine addiction (KEGG ID: 5031) pathways from common interacting network of SDG (Fig. 7). We observed that a group of HK and NHK genes together exhibited a similar pattern of interactions in both pathways. Co-expression pattern between one of the HK gene, calmodulin 1 (CALM1) and two separate families of NHK genes; calmodulin-dependent protein kinase (CAMK2A and CAMK2B) and NMDA receptors (GRIN2A and GRIN2B) were found to be common not only in the above two pathways but also in other pathways such as calcium signaling (KEGG ID: 4020), neurotropin signaling (KEGG ID: 4722), and long-term potentiation (KEGG ID: 4720) (data not shown). Hence, we conclude that candidate pathways are interlinked in schizophrenia through a common co-expression distribution of SDG.

FIG. 7.

Pathway specific interactions pattern of SDG in common interacting DGNet. A subset of interactions were dragged from DGNet for both (A) amphetamine addiction (KEGG ID: 5031) and (B) dopaminergic synapse (KEGG ID: 4728) pathway. SDG and their interactions present in both pathways are highlighted. which clearly indicates a common pattern of connectivity between a set of SDG.

Further enlisting the biological significance of such common co-expression pattern of SDG, we highlighted their positions in KEGG pathway map (Supplementary Fig. S3). It was observed that the genes which exhibited common co-expression pattern are closely interconnected and are mostly involved in the end terminal process of the pathways (highlighted red). However, genes that did not follow any pattern were found to be positioned randomly in the pathway (highlighted yellow and pink). An increase in extracellular concentration of dopamine triggers ‘dopamine-mediated NMDA receptors trafficking’ systems in brain (Aperia and Greengard, 2006; Gao and Wolf, 2008). NMDA receptors are glutamate-gated cation channels that are involved in the modulation of a synaptic response (Dingledine et al., 1999). Activation of NMDA receptors (GRIN2A and GRIN2B) leads to an influx of calcium (Ca2⁺) into the postsynaptic region where it activates several signaling cascades. Elevated Ca2⁺ levels allow activation of calmodulin (CALM1) and the kinase activity of calmodulin-dependent kinases (CAMK2A and CAMK2B) which phosphorylate the transcription factor encoded by cAMP responsive element binding (CREB1) gene. As a result, an induced transcription of cAMP-responsive element (CRE) leads to the dysregulation of cyclic adenosine monophosphate (cAMP) that result in development of various symptoms related to schizophrenia and other neurological disorders (Fig. 8).

FIG. 8.

Role of common network elements in cAMP regulation. Common pattern of interactions between SDG play an important role in cyclic AMP regulation process. A schematic representation shows the combined role of major SDG along with other members in downstream cAMP regulation and indicates a central role of calmodulin kinases (CAMK2A and CAMK2B).

Genes that followed a common pattern of interaction are highly organized at functional level and thus are important candidates as drug targets in schizophrenia. Here, NMDA receptors were seen to share common binding site for schizophrenic drugs such as L-glutamic acid (DB00142), glycine (DB00145), and haloperidol (DB00502) (Supplementary Fig. S4). Earlier studies have shown that drug targeted network in schizophrenia is mostly centered with disease genes from the dopamine and serotonin family (Potkin et al., 2010) that also appeared as an important common SDG in majority of the pathways in our study (DRD2 and HTR2A). However, the pharmacological field is still limited by drugs for proteins such as calmodulin-dependent kinases (CAMK2A and CAMK2B) that are key mediators for the secondary messenger-mediated signal transduction in the disease pathways. It is well established that abnormal protein phosphorylation by kinases is an underlying cause or a consequence in several diseases. Hence, apart from G-protein couple receptors, protein kinases have also been considered as important drug targets in the last decade (Badrinarayan and Sastry, 2013; Cohen, 2002; McGuire et al., 2014). We identified three different protein kinases (CSNK2A1, CAMK2A, and CAMK2B) as chief candidate SDG in both the topological and pathway based analysis. Surprisingly, none of these have been targeted by any of the available schizophrenia-specific drugs designed so far.

Therefore, we carried out docking studies of the three protein kinases with known schizophrenic drugs. It was observed that haloperidol features similar binding affinities with all the three kinases apart from targeting GRN2B, DRD2, and HTR2A. We also observed that few bridge drugs such as trazodone and trifluoperazine show similar interaction energetics with all three kinases in addition to the known targets such as SLC6A4, HTR2A and CALM1, DRD2, respectively (Supplementary Fig. S4). On the other hand, the drugs lurasidone, aripiprazole, and bromocriptine also appear to bind significantly with all three kinases. These observations reflect the therapeutic potential of protein kinases for further drug development in schizophrenia.

Association of the available marketed antipsychotic drugs with systemic side effects (Divac et al., 2014; Werner and Coveñas, 2014) can further guide us to design novel drugs by optimizing off-target binding at early stages of drug development. Given the multifactorial nature of schizophrenia, a systematic analysis of disease genes based on genetic variants also will further help in better understanding of disease pathophysiology.

Conclusions

We constructed a schizophrenia disease gene interaction network based upon the co-expression statistics of known disease genes that are common in twenty different zones of human brain tissue. We highlighted the common network elements based on the rich topological properties of SDG and their common pattern of distribution in disease candidate pathways. Interactions amongst the SDG appeared functionally more relevant than their interactions with other non-disease genes for schizophrenia. We isolated a few bridge drugs and evaluated their binding affinity for some important candidates in the network through molecular docking studies. We also suggested that identification of common interaction patterns of genes in the schizophrenia disease–gene interaction network will help to prioritize the essential targets for future drug discovery endeavours in the treatment of schizophrenia.

Footnotes

Acknowledgments

This work is supported by the Department of Biotechnology (DBT), Ministry of Science and Technology, Government of India. We also thank all anonymous reviewers for their critical comments and suggestions that helped us to improve this article.

Author Disclosure Statement

Both authors declare that they have no competing financial interests.

Abbreviations Used

References

Abi-Dargham

. (2004). Do we still believe in the dopamine hypothesis? New data bring new evidence. Int J Neuropsychopharmacol, 1, S1–S5.

Aerts

, Lambrechts

, Maity

, et al. (2006). Gene prioritization through genomic data fusion. Nat Biotechnol, 24, 537–544.

Amberger

, Bocchini

, Scott

, and Hamosh

. (2009). McKusick's Online Mendelian Inheritance in Man (OMIM). Nucleic Acids Res, 37, D793–D796.

Andreasen

, Arndt

, Swayze

, et al. (1994). Thalamic abnormalities in schizophrenia visualized through magnetic resonance image averaging. Science, 266, 294–298.

Andreasen

, Ehrhardt

, Swayze

, et al. (1990). Magnetic resonance imaging of the brain in schizophrenia: The pathophysiologic significance of structural abnormalities. Arch Gen Psychiatry, 47, 35–44.

Aperia

, and Greengard

. (2006). Dopamine receptor response to NMDA stimulation. Am J Psychiatry, 163, 1682–1682.

Assenov

, Ramírez

, Schelhorn

, Lengauer

, and Albrecht

. (2008). Computing topological parameters of biological networks. Bioinformatics, 24, 282–284.

Badrinarayan

, and Sastry

. (2013). Rational approaches towards lead optimization of kinase inhibitors: The issue of specificity. Curr Pharm Des, 19, 4714–4738.

Barabasi

, and Albert

. (1999). Emergence of scaling in random networks. Science, 286, 509–512.

10.

Barta

, Pearlson

, Powers

, et al. (1990). Auditory hallucinations and smaller superior temporal gyral volume in schizophrenia. Am J Psychiatry, 147, 1457–1462.

11.

Bauer-Mehren

, Bundschus

, Rautschka

, Mayer

, Sanz

, and Furlong

. (2011). Gene-disease network analysis reveals functional modules in mendelian, complex and environmental diseases. PLoS One, 6, e20284.

12.

Bergen

, and Petryshen

. (2012). Genome-wide association studies of schizophrenia: Does bigger lead to better results?. Curr Opin Psychiatry, 25, 76–82.

13.

Berman

, Westbrook

, Feng

, Gilliland

, Bhat

, Weissig

, Shindyalov

, and Bourne

. (2000). The Protein Data Bank. Nucleic Acids Res, 28, 235–242.

14.

Bindea

, Mlecnik

, Hackl

, et al. (2009). ClueGO: A Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics, 25, 1091–1093.

15.

Boileau

, Payer

, Chugani

, et al. (2013). In vivo evidence for greater amphetamine-induced dopamine release in pathological gambling: A positron emission tomography study with [11C]-(+)-PHNO. Mol Psychiatry. doi: 10.1038/mp.2013.163. [Epub ahead of print.]

16.

Bossi

, and Lehner

. (2009). Tissue specificity and the human protein interaction network. Mol Syst Biol, 5, 260.

17.

Brandes

. (2001). A faster algorithm for betweenness centrality. J Math Sociol, 25, 163–117.

18.

Bühler

, Hambrecht

, Löffler

, an der Heiden

, and Häfner

. (2002). Precipitation and determination of the onset and course of schizophrenia by substance abuse: A retrospective and prospective study of 232 population-based first illness episodes. Schizophrenia Res, 54, 243–252.

19.

Butte

, Dzau

, and Glueck

. (2001). Further defining housekeeping, or maintenance, genes Focus on A compendium of gene expression in normal human tissues. Physiol Genomics, 7, 95–96.

20.

Camargo

, Collura

, Rain

, Mizuguchi

, Hermjakob

, Kerrien

, Bonnert

, Whiting

, and Brandon

. (2007). Disrupted in Schizophrenia 1 Interactome: Evidence for the close connectivity of risk genes and a potential synaptic basis for schizophrenia. Mol Psychiatry, 12, 74–86.

21.

Carlsson

, and Carlsson

. (1990). Interactions between glutamatergic and monoaminergic systems within the basal ganglia—Implications for schizophrenia and Parkinson's disease. Trends Neurosci, 13, 272–276.

22.

Chang

, Cheng

, Chen

, Shu

, Tsai

, Huang

, and Hsu

. (2011). Identification of human housekeeping genes and tissue-selective genes by microarray meta-analysis. PLoS One, 6, e22859.

23.

Cheng

, Chang

, Chen

, et al. (2011). Identification of reference genes across physiological states for qRT-PCR through microarray meta-analysis. PLoS ONE, 6, e17347.

24.

Cohen

. (2002). Protein kinases—The major drug targets of the twenty-first century?. Nat Rev Drug Discov, 1, 309–315.

25.

Craig

. (2008). Complex diseases: Research and applications. Nat Educ, 1, 1.

26.

Critchley

, Eccles

, and Garfinkel

. (2013). Interaction between cognition, emotion, and the autonomic nervous system. Handb Clin Neurol, 117, 59–77.

27.

Dezso

, Nikolsky

, Sviridov

, et al. (2008). A comprehensive functional analysis of tissue specificity of human gene expression. BMC Biol, 6, 49.

28.

Diestel

. (2005). Graph Theory. Springer-Verlag, Heidelberg. ISBN 3-540-26182-6.

29.

Dingledine

, Borges

, Bowie

, et al. (1999). The glutamate receptor ion channels. Pharmacol Rev, 51, 7–61.

30.

Divac

, Prostran

, Jakovcevski

, and Cerovac

. (2014). Second-generation antipsychotics and extrapyramidal adverse effects. Biomed Res Int, 2014, 656370.

31.

Dixon

, Haas

, Weide

, Sweeney

, and Frances

. (1991). Drug abuse in schizophrenic patients: Clinical correlates and reasons for use. Am J Psychiatry, 148, 224–230.

32.

Eisenberg

, and Levanon

. (2013). Human housekeeping genes, revisited. Trends Genet, 29, 569–574.

33.

Evins

, Green

, Kane

, and Murray

. (2012). The effect of marijuana use on the risk for schizophrenia. J Clin Psychiatry, 73, 1463–1468.

34.

Fabi

, Fusco

, Valiante

, and Celli

. (2013). Genetics and epigenetics of schizophrenia. Clin Ter, 164, e319–24.

35.

Freeman

. (1979). Centrality in social networks: Conceptual clarification. Soc Networks, 1, 215–239.

36.

Friesner

, Murphy

, Repasky

, et al. (2006). Extra Precision Glide: Docking and scoring incorporating a model of hydrophobic enclosure for protein-ligand complexes. J Med Chem, 49, 6177–6196.

37.

Gao

, and Wolf

. (2008). Dopamine receptors regulate NMDA receptor surface expression in prefrontal cortex neurons. J Neurochem, 106, 2489–2501.

38.

Gilman

, Chang

, Xu

, Bawa

, Gogos

, Karayiorgou

, and Vitkup

. (2012). Diverse types of genetic variation converge on functional gene networks involved in schizophrenia. Nat Neurosci, 15, 1723–1728.

39.

Guo

, Sun

, Jia

, and Zhao

. (2010). A novel microRNA and transcription factor mediated regulatory network in schizophrenia. BMC Syst Biol, 4, 10.

40.

Hans-Gert

, Gerburg

, Johann

, Henrik

, and Bernhard

. (2010). The hypothalamus in schizophrenia research: No longer a wallflower existence. Open Neuroendocrinol J, 3, 59–67.

41.

Harrison

, Pritchett

, Stumpenhorst

, et al. (2012). Genetic mouse models relevant to schizophrenia: Taking stock and looking forward. Neuropharmacology, 62, 1164–1167.

42.

Howes

, and Kapur

. (2009). The dopamine hypothesis of schizophrenia: Version III—The final common pathway. Schizophr Bull, 35, 549–562.

43.

Hsu

, Yang

, Shih

, Liu

, and Hwu

. (2008). A protein interaction based model for schizophrenia study. BMC Bioinformatics, 9, S23.

44.

Huang

, Deng

, Chen

, et al. (2009). Brain structure abnormality as genetic endophenotype of schizophrenia. Zhonghua Yi Xue Yi Chuan Xue Za Zhi, 26, 490–494.

45.

Jacobs

, and Silverstone

. (1986). Dextroamphetamine-induced arousal in human subjects as a model for mania. Psychol Med, 16, 323–329.

46.

Janssens

, and van Duijn

. (2008). Genome-based prediction of common diseases: Advances and prospects. Hum Mol Genet, 17, R166–173.

47.

Jeong

, Mason

, Barabasi

, and Oltvai

. (2001). Lethality and centrality in protein networks. Nature, 411, 41–42.

48.

Jia

, Kao

, Kuo

, and Zhao

. (2011). A comprehensive network and pathway analysis of candidate genes in major depressive disorder. BMC Syst Biol, 23, S12.

49.

Jia

, Sun

, Guo

, and Zhao

. (2010). SZGR: A comprehensive schizophrenia gene resource. Mol Psychiatry, 15, 453–462.

50.

Jia

, Wang

, Fanous

, Pato

, Edwards

; International Schizophrenia Consortium, and Zhao

. (2012). Network-assisted investigation of combined causal signals from genome-wide association studies in schizophrenia. PLoS Comput Biol, 8, e1002587.

51.

Jiang

, Zhou

, Liu

, and Song

. (2013). Brainnetome-wide association studies in schizophrenia: The advances and future. Neurosci Biobehav Rev, 37, 2818–2835.

52.

Kanehisa

, Goto

, Sato

, Furumichi

, and Tanabe

. (2012). KEGG for integration and interpretation of large-scale molecular data sets. Nucleic Acids Res, 40, D109–114.

53.

Klomp

, Koolschijn

, Hulshoff Pol

, Kahn

, and Haren

. (2012). Hypothalamus and pituitary volume in schizophrenia: a structural MRI study. Int J Neuropsychopharmacol, 15, 281–288.

54.

Knox

, Law

, Jewison

, et al. (2011). DrugBank 3.0: A comprehensive resource for ‘omics’ research on drugs. Nucleic Acids Res, 39, D1035–1041.

55.

Kumarasinghe

, Tooney

, and Schall

. (2012). Finding the needle in the haystack: A review of microarray gene expression research into schizophrenia. Aust N Z J Psychiatry, 46, 598–610.

56.

Lau

, Wang

, Hsu

, and Liu

. (2013). Does the dopamine hypothesis explain schizophrenia?. Rev Neurosci, 24, 389–400.

57.

Lee

, Tsao

, Yang

, et al. (2011). Construction and analysis of the protein-protein interaction networks for schizophrenia, bipolar disorder, and major depression. BMC Bioinformatics, 12, S20.

58.

Luo

, Huang

, Jia

, Li

, Su

, Zhao

, and Gan

. (2014). Protein-protein interaction and pathway analyses of top schizophrenia genes reveal schizophrenia susceptibility genes converge on common molecular networks and enrichment of nucleosome (chromatin) assembly genes in schizophrenia susceptibility loci. Schizophr Bull, 40, 39–49.

59.

McGuire

, Hammond

, Yates

, Chen

, Haroutunian

, Meador-Woodruff

, and McCullumsmith

. (2014). Altered serine/threonine kinase activity in schizophrenia. Brain Res, 1568, 42–54.

60.

Morgan

, Dudley

, Deshpande

, and Butte

. (2010). Dynamism in gene expression across multiple studies. Physiol Genomics, 40, 128–140.

61.

Murray

, Paparelli

, Morrison

, Marconi

, and Di Forti

. (2013). What can we learn about schizophrenia from studying the human model, drug-induced psychosis?. Am J Med Genet B Neuropsychiatr Genet, 162B, 661–670.

62.

Pelayo-Terán

, Suárez-Pinilla

, Chadi

, and Crespo-Facorro

. (2012). Gene-environment interactions underlying the effect of cannabis in first episode psychosis. Curr Pharm Des, 18, 5024–5035.

63.

Podder

, Jatana

, and Latha

. (2014). Human Dopamine Receptors Interaction Network (DRIN): A systems biology perspective on topology, stability and functionality of the network. J Theor Biol, 357, 169–183.

64.

Potkin

, Macciardi

, Guffanti

, et al. (2010). Identifying gene regulatory networks in schizophrenia. Neuroimage, 53, 839–847.

65.

Premkumar

, Kumari

, Corr

, and Sharma

. (2006). Frontal lobe volumes in schizophrenia: Effects of stage and duration of illness. J Psychiatr Res, 40, 627–637.

66.

Putnam

, Sun

, and Zhao

. (2011). Exploring schizophrenia drug-gene interactions through molecular network and pathway modeling. AMIA Annu Symp Proc, 2011, 1127–1133.

67.

Rabin

, Zakzanis

, Daskalakis

, and George

. (2013). Effects of cannabis use status on cognitive function, in males with schizophrenia. Psychiatry Res, 206, 158–165.

68.

Rastelli

, Del Rio

, Degliesposti

, and Sgobba

. (2010). Fast and accurate predictions of binding free energies using MM-PBSA and MM-GBSA. J Comput Chem, 31, 797–810.

69.

Rivals

, Personnaz

, Taing

, and Potier

. (2007). Enrichment or depletion of a GO category within a class of genes: Which test?. Bioinformatics, 23, 401–407.

70.

Schultz

, North

, and Shields

. (2007). Schizophrenia: A review. Am Fam Physician, 75, 1821–1829.

71.

Schwab

, and Wildenauer

. (2013). Genetics of psychiatric disorders in the GWAS era: An update on schizophrenia. Eur Arch Psychiatry Clin Neurosci, 2, S147–154.

72.

Sebat

, Levy

, and McCarthy

. (2009). Rare structural variants in schizophrenia: One disorder, multiple mutations; one mutation, multiple disorders. Trends Genet, 25, 528–535.

73.

Seeman

. (2013). Are dopamine D2 receptors out of control in psychosis?. Prog Neuropsychopharmacol Biol Psychiatry, 4, 146–152.

74.

Shannon

, Markiel

, Ozier

, et al. (2003). Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res, 13, 2498–2504.

75.

Singh

, Kumar

, Agarwal

, Phadke

, and Jaiswal

. (2014). Genetic insight of schizophrenia: Past and future perspectives. Gene, 535, 97–100.

76.

Sreedevi

, and Danail

. (2012). A network view on schizophrenia related genes. Network Biol, 2, 16–25.

77.

, Wiltshire

, Batalov

, et al. (2004). A gene atlas of the mouse and human protein-encoding transcriptomes. Proc Natl Acad Sci, 101, 6062–6067.

78.

Sun

, Jia

, Fanous

, et al. (2010). Schizophrenia gene networks and pathways and their applications for novel candidate gene selection. PLoS One, 5, e11351.

79.

Sun

, Jia

, Fanous

, et al. (2009). A multi-dimensional evidence-based candidate gene prioritization approach for complex diseases–schizophrenia as a case. Bioinformatics, 25, 2595–2602.

80.

Sun

, Xu

, and Zhao

. (2012). Network-assisted investigation of antipsychotic drugs and their targets. Chem Biodivers, 9, 900–910.

81.

Sunday

. (2011). Can the action of amphetamine on dopamine cause schizophrenia?. Psychol Edu J. Vol. 39, No. 1.

82.

Tognin

, Rambaldelli

, Perlini

, et al. (2012). Enlarged hypothalamic volumes in schizophrenia. Psychiatry Res, 204, 75–81.

83.

Torkamani

, Brian

, Nicholas

, and Elizabeth

. (2010). Coexpression network analysis of neural tissue reveals perturbations in developmental processes in schizophrenia. Genome Res, 20, 403–412.

84.

Tseng

, Lewis

, Lipska

, and O'Donnell

. (2007). Post-pubertal disruption of medial prefrontal cortical dopamine-glutamate interactions in a developmental animal model of schizophrenia. Biol Psychiatry, 62, 730–738.

85.

, Lee

, Chen

, Li

, and Su

. (2013). Schizophrenia and the brain's control network: Aberrant within- and between-network connectivity of the frontoparietal network in schizophrenia. Schizophr Res, 147, 339–347.

86.

Van Os

, Rutten

, and Poulton

. (2008). Gene-environment interactions in schizophrenia: Review of epidemiological findings and future directions. Schizophr Bull, 34, 1066–1082.

87.

Vasaikar

, Padhi

, Jayaram

, and Gomes

. (2013). NeuroDNet—An open source platform for constructing and analyzing neurodegenerative disease networks. BMC Neurosci, 14, 3.

88.

Vázquez

, Dobrin

, Sergi

, Eckmann

, Oltvai

, and Barabási

. (2004). The topological relationship between the large-scale attributes and local interaction patterns of complex networks. Proc Natl Acad Sci USA, 101, 17940–17945.

89.

Werner

, and Coveñas

. (2014). Safety of antipsychotic drugs: Focus on therapeutic and adverse effects. Expert Opin Drug Saf, 13, 1031–1042.

90.

Werner

, Malaspina

, and Rabinowitz

. (2007). Socioeconomic status at birth is associated with risk of schizophrenia: Population-based multilevel study. Schizophr Bull, 33, 1373–1378.

91.

Woodruff

, McManus

, and David

. (1995). Meta-analysis of corpus callosum size in schizophrenia. J Neurol Neurosurg Psychiatry, 58, 457–461.

92.

Yildiz

, Borgwardt

, and Berger

. (2011). Parietal lobes in schizophrenia: Do they matter?. Schizophr Res Treatment, 2011, 581686.

93.

, Bi

, Liu

, Zhao

, Zhang

JF2

, Zhang

, and Yue

. (2014). Protein-interaction-network-based analysis for genome-wide association analysis of schizophrenia in Han Chinese population. J Psychiatr Res, 50, 73–78.

94.

Zhou

, Suzuki

, Takahashi

, et al. (2007). Parietal lobe volume deficits in schizophrenia spectrum disorders. Schizophrenia Res, 89, 35–48.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB

0.08 MB

0.13 MB

0.09 MB

0.03 MB