Towards finding the linkage between metabolic and age-related disorders using semantic gene data network analysis.

A metabolic disorder (MD) occurs when the metabolic process is disturbed. This process is carried out by thousands of enzymes participating in numerous inter-dependent metabolic pathways. Critical biochemical reactions that involve the processing and transportation of carbohydrates, proteins and lipids are affected in metabolic diseases. Therefore, it is of interest to identify the common pathways of metabolic disorders by building protein-protein interactions (PPI) for network analysis. The molecular network linkages between MD and age related diseases (ARD) are intriguing. Hence, we created networks of protein-protein interactions that are related with MD and ARD using relevant known data in the public domain. The network analysis identified known MD associated proteins and predicted genes and or its products of ARD in common pathways. The genes in the common pathways were isolated from the network and further analyzed for their co-localization and shared domains. Thus, a model hypothesis is proposed using interaction networks that are linked between MD and ARD. This data even if less conclusive finds application in understanding the molecular mechanism of known diseases in relation to observed molecular events.

Background

Metabolic disorder (MD) is a cluster of metabolic risk factors characterized by obesity, elevated blood pressure, increased plasma glucose (fasting), high triglycerides in serum and decreased high-density cholesterol levels [1]. Metabolic disorder affected people are at increased risk for atherosclerosis, peripheral vascular disease, coronary heart disease, myocardial infarction, stroke, and type 2 diabetes [2-5]. These are the leading causes of disability worldwide [6]. The consequences of metabolic disorders are often treated by healthy weight, diet and physical activities [7,8]. So, evaluation of metabolic risk factors and the identification of population groups at risk of chronic diseases are essential for developing prevention strategies. Hence, the dynamic modeling of biological systems to describe various human diseases is of interest in recent years.

The complex network of proteins (gene products) and their biological processes mediating interactions among them in these diseases are of importance to understand. The application of protein interaction networks to available disease datasets in the public domain allows the identification of genes and their corresponding proteins. This helps the creation of sub-networks to study network properties for the classification of diseaseassociated genes in networks. It is found that several strategies have been employed to analyze gene networks using data for protein interactions in these conditions. However, this is a complex and a challenging task to pursue [9]. The information related to the disease mechanism gleaned using data for gene networks at a system level is critical yet it is highly convoluted. This is possible by collecting relevant data followed by cleaning such data by removing redundant information for useful yet specific knowledge establishments. This is helpful for improved data analysis followed by data integration to create a reliable model of the disease under study. Thus, gene network methods have been used to gain insights into disease mechanisms [10,11], co-morbidity (anomalous conditions) [12,13], protein target identification [14-16] and biomarker detection [17,18]. The gene network based study includes elucidation of a complex system by fragmenting them into finite components (nodes or vertices) and interactions (edges). This conceptual illustration helps in the understanding of complex molecular disturbances in diseased conditions. Therefore, it is of interest to use graph theory based pathway diagrams using pertinent co-localization information with shared domain data between MD and ARD by mediating protein-protein interaction networks to identify the genes in a common pathway among disease types, states and conditions.

Methodology

Disease associated gene data collection from known literature

We gathered disease associated proteins (gene products) and or their corresponding genes related data from publically (WWW – World Wide Web) available databases such as PubMed (http://www.ncbi.nlm.nih.gov/pubmed), PubMed Central (PMC - (http://www.ncbi.nlm.nih.gov/pmc) and other open access journals maintained by several publishers across Nations. This is done through disease specific manual keyword (metabolic disorder (MD), age related disorder (ARD), relevant genes) searching, article gathering, visual scanning, reading, studying, understating, cleaning, grouping, labeling, refining, storing in simple RDBMS, and subsequent data retrieval for value addition, information enrichment and knowledge creation on the subject of the study. It should be noted that PubMed and PMC are maintained at National Centre for Biotechnology (NCBI), National Institute of Health (NIH), USA.

Disease specific network creation

We used GeneMANIA (http://www.genemania.org/) [19] to collect data related to metabolic disorder (MD) for relevant information gathering and knowledge establishment with available graphical network diagrams retrieved from the server in this study. GeneMANIA provides data for protein-protein interactions, protein-DNA interactions and or protein-gene interactions, corresponding pathways, associated reactions, available phenotypic profiles and genes expression data with corresponding known yet characterized proteins in the network. The data thus described is reasonably representative if not comprehensive.

Common pathway identification

We used the pathway database (http://www. pathwaycommons.org) [20] to identify the common pathways amongst genes of metabolic disorder (MD). Pathway commons is a collection of pathway information from multiple organisms. A comprehensive collection of biological pathways from multiple sources is provided by pathway commons and represented in a common language for genes based metabolic pathway analysis.

Pathway analysis

We used Reactome (http://www.reactome.org/), a curated, peer reviewed and an open source pathway database for exploring the pathway knowledge for common genes [21].

Gene localization and domain analysis

Genes are often expressed in the same tissue and corresponding proteins can be found in the same location. Two or more genes are linked if they are expressed in the same tissue and their corresponding proteins are found in the same cellular location. Similarly, two proteins (gene products) are linked if they have the same defined (sequence and or structure) protein domain. This is completed using InterPro, SMART and Pfam facilities in the public domain. Gene localization and domain analysis are completed using the tools at GeneMANIA (http://www.gene mania.org/).

Results & Discussion

The hypothesis describing the mechanism leading the molecular disturbances in metabolic disorders (MD) and agerelated disorders (ARD) is usually non-trivial nature. Therefore, it is important to relate a disease condition with its know yet reasonably representative associated genes for the construction of its corresponding protein-protein interaction networks. It is also of critical importance to identify genes and or their protein products that share common pathways between different disease conditions (e.g. MD and ARD). Therefore, we used textmining (keyword searching) techniques to identify genes associated with MD as shown in Figure 1. The text-mining analysis searched for respective known genes associated with 36 metabolic diseases (Table 1). Thus, genes known to be associated with each described metabolic disorder is listed.

Figure 1

A schematic representation for the linkage hypothesis between MD and ARD is shown.

Table 1

Description of common genes in pathways associated with specific diseases.

Metabolic disorders		Responsible genes
Diabetes	Type 1	HLA DQA1, DQB1, DRB1
	Type 2	TCF7L2, PPARG, KCNJ11, NOTCH2, WFS1, IGF2BP2, SLC30A8, JAZF1, HHEX
Phenyl-keto-nuria		PAH
Glucose galactose mal-absorption		SGLT1
Tyrosine-mia		FAH, HPD, TAT
Alkaptonuria		HGD
Homo-cystinuria		CBS, MTHFR, MTR, MTRR, MMADHC
Maple syrup urine disease		BCKDHA, BCKDHB, DBT
Propionic acid-emia		PCCA, PCCB
Methyl-malonic acid-emia		MUT, MMAA, MMAB, MMADHC, MCEE
Hyper-cholesterol-emia		APOB, LDLR, LDLRAP1, PCSK9
Glycogen storage disease	Type 0	GYS1, GYS2
	Type 1	G6PC, SLC37A4
	Type 2	GAA (Pomp disease)
	Type 3	AGL
	Type 4	GBE1
	>Type 5	PYGM
	Type 6	PYGL
	Type 7	PFKM
	Type 9	PHKA1, PHKA2, PHKB, PHKG2
	Type 10	PGAM2 (Phospho-glycerate mutase deficiency)
Galactose-mia	GALE, GALK1, GALT
Congenital glycosylation disorder	ALG12
	Type Ic	ALG6
	Type Iii	COG5
Lesch-Nyhan syndrome		HPRT1
Gaucher disease		GBA
Tay-Sachs disease		HEXA
Fabry disease		GLA
Hurler syndrome		IDUA
Hunter syndrome		IDS
Sanfilippo syndrome		GNS, HGSNAT, NAGLU, SGSH
Maroteaux-Lamy syndrome		ARSB
Morquio syndrome		GALNS, GLB1
Refsum disease		PEX7, PHYH
Hemo-chromatosis		HAMP, HFE, HFE2, SLC40A1, TFR28

The listed data were further analyzed to identify key genes associated with various diseases for network analysis and evaluation (Figure 2). This exercise identified five common pathways with predicted genes for possible disease association from the network analysis (Figure 3). Subsequently, common pathway genes with specific diseases are thus summarized (Table 2). This data shows that APOB, LDLR, APOE, LIPC genes are found in common pathways for lipid digestion, mobilization, and transport. Results show that two genes APOE, LIPC were not in MD but they shared the common pathway with APOB, LDLR responsible for Glycogen storage Disease type 0 (GSD). These two predicted genes are known to cause age-related macular degeneration (AMD) [22,23]. The BCKDHA, DBT, DLD and BCKDHB genes were found in the pathways of branched-chain amino acid catabolism and metabolism of amino acids and its derivatives. However, the predicted DLD gene was known to be associated with skin photo aging [24]. The other three genes (BCKDHA, DBT, and BCKDHB) are the causative agent of MD described as Male Syrap Urine Disease (MSUD) as given in Table 1.

Figure 2

A network of genes associated with the metabolic disorder (MD) is illustrated in this diagram. It should be noted that genes that are already known to be associated with the disorder is shown in the figure. Black nodes indicate known MD genes and gray nodes indicate genes that are predicted to be associated with the disorder.

Figure 3

An illustration showing common pathways between MD and ARD is given. (A) Common pathways of MD with predicted disorder associated genes are shown. (B) Five common pathways between MD and predicted disorder genes are illustrated.

Table 2

Description of manually gleaned common pathway genes with known specific diseases.

ID = identifier; ATP = adenosine tri-phosphate; K+ = potassium; SLC = solute carrier; ABC = ATP-binding cassette; AMD = age-related macular degeneration; HC = Hemo-chromatosis; GSDV = Glycogen storage disease type 5; GSDIX = Glycogen storage disease type 9; MSUD = Maple Syrap Urine Disease; GSD = Glycogen storage disease type 0; SP = skin photo aging; glyco-genolysis = glycogen breakdown

Genes	Pathway ID R-HAS(prefix)	Pathways	Known disease genes			Common pathway (predicted)		Reference
			Disease	Genes	Genes	Disease
APOB; LDLR; APOE; LIPC	73923	Lipid digestion, mobilization and transport	GSD		APOB	APOE	AMD	[22]
					LDL	LIPC	AMD	[23]
BCKDHA; DBT; DLD; BCKDHB	70895	Branched chain amino acid catabolism	MSUD		BCKDHA; BCKDHB; DBT	DLD	SP	[24]
	71291	Metabolism of amino acids and its derivatives
PYGM; PHKB;PHKG1; PHKA1	70221	Glyco-genolysis	GSDV		PYGM	PHKG1	Adenoid cystic carcinoma	[25]
	70326	Carbohydrate metabolism	GSDIX		PHKB; PHKA1
	71387
	1430728
HEPH;CP;SLC40A1	425410	Metal ion SLC transporters	HC		SLC40A1	HEPH	AMD	[26]
	917937	Iron uptake and transport
	425366	Transport of sugars (glucose), bile salts, organic acids, metal ions and amine compounds
	425407	SLC-mediated trans-membrane transport				CP	AMD	[26]
	382551	Trans-membrane transport of small molecules
ABCC8;KCNJ11	1296025	ATP sensitive K+ channels	Type 2 diabetes		KCNJ11	ABCC8	Ageing	[27]
	1296065	Inwardly rectifying K+ channels
	382556	ABC proteins mediated transport
	422356	Insulin secretion regulation
	1296071	K+ channels
	163685	Energy metabolism
	112316	Neuronal system
	382551	Trans-membrane transport of small molecules
	1430728	Metabolism related

The four genes (PYGM, PHKB, PHKG1, PHKA1) were found in a common pathway from the network analysis sharing glycogenolysis (glycogen breakdown) and carbohydrate metabolism. The predicted gene PHKG1 is known to be associated with adenoid cystic carcinoma [25]. The other three genes PYGM, PHKB and PHKA1 involved in glycogen storage disease type 5 (GSDV) and glycogen storage disease type 9 (GSDIX) is also observed in the network analysis. The other set of genes (HEPH, CP and SLC40A1) were also found in common pathways of metal ion SLC transporters, iron uptake and transport, glucose transport, transport of bile salts and organic acids, metal ions and amine compounds, SLC-mediated transmembrane transport and trans-membrane transport of small molecules. It should be noted that HEPH and CP are not associated with MD but they are involved with age-related macular degeneration (AMD) [25].

Two genes (ABCC8 and KCNJ11) associated with Type 2 diabetes (Table 2) and aging share a common pathway and hence have a linkage [26]. ABCC8 and KCNJ11 shared the common pathways for ATP sensitive K+ channels, inwardly rectifying K+ channels, ABC-family proteins mediated transport, regulation of insulin secretion, integration of energy metabolism, neuronal system association, trans-membrane transport of small molecules and other metabolism as given in Table 2. Common genes associated with both MD and ARD are further processed for molecular interactions using network analysis as described in the methodology section. It is further showed that these genes shared the same domain in pathway regulation (Figure 4). They are also co-localized with each other in tissues when expressed.

Figure 4

A network between co-localized and domain-sharing genes is shown. Black nodes indicate the common pathway of genes between MD and ARD.

It is of important to understand the specific molecular pathways unique to a disease to elucidate the difference in these pathways. Therefore, it is essential to construct a ‘linkage network’ between diseases that are inter-linked by one or more genes found associated with the diseases using simplified network diagrams. We illustrated a linkage network based on pathway data, domain information and co-localization analyses between MD and ARD in this report. Thus, a model hypothesis is proposed using interaction networks that are linked between MD and ARD (Figure 5).

Figure 5

Venn diagram showing MD (dark green) and ARD (dark blue) disorders, and common genes of MD (light green) and ARD (light blue) is shown. Set of common pathway of genes for MD (light green) and intersection (Ash) for set of common genes between MD (light green) and ARD (light blue), (MD) ∩ (ARD) = {APOE, LIPC, HEPH, CP, DLD, ABCC8} are shown.

Conclusion

We report data for common pathway of genes responsible for metabolic disorders (MD) and age related disorders (ARD). Data shows the linkage of genes in these diseases by analyzing their co-localization and shared domains. Pathway analysis with gene regulatory network evaluation using gene circuits and module design for further analysis of gene product interactions is essential for understanding the mechanism of the disease in relation to molecular cellular biology events.