Sevoflurane-medicated the pathway of chemokine receptors bind chemokines in patients undergoing CABG

Bing Kong; Yu-Wu Ma; De-Xue Li; Xi-Jiang Liu; Yong-Guang Xu

doi:10.1515/biol-2017-0052

Article Open Access

Sevoflurane-medicated the pathway of chemokine receptors bind chemokines in patients undergoing CABG

Bing Kong , Yu-Wu Ma , De-Xue Li , Xi-Jiang Liu and Yong-Guang Xu

Published/Copyright: December 29, 2017

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information Explore this Subject

From the journal Open Life Sciences Volume 12 Issue 1

Abstract

Background

We aim to identify sevoflurane-induced modules and pathways in patients following coronary artery bypass graft (CABG) surgery, and to further elucidate the molecular mechanisms of the cardioprotective effects of sevoflurane.

Methods

Differential co-expression network (DCN) was constructed. Candidate modules were identified via three steps: selection of seed genes, search of modules using snowball sampling, and refinement of modules. Afterwards, the significance of the candidate modules was assessed. Ultimately, pathway analyses for genes in differential modules were implemented to illuminate the biological processes.

Results

Overall, 122 genes were identified to serve as seed genes. From every seed gene, we extracted 122 modules and the mean node size in a module was 3. By setting the classification accuracy cutoff at 0.9 and the number of nodes in a module at 5, 7 candidate modules were identified, including module 80, 82, 82, 84, 85, 86 and 89. Based on the random permutation test, we found that these 7 candidate modules were all differential ones. Moreover, pathway analysis showed that genes in the differential modules 80, 82, and 85 were all enriched in the pathway of chemokine receptors bind chemokines.

Conclusion

Sevoflurane might exert cardioprotective functions in patients following CABG, partially through regulating the pathway of chemokine receptors bind chemokines.

Keywords: coronary artery bypass graft; sevoflurane; module; pathway enrichment analysis

1 Introduction

In recent years, off-pump coronary artery bypass grafting (CABG) surgery has emerged as a common method in treating cardiac disease [1]. However, this less invasive technique can result in the myocardial injury because of a transient period of local ischemia with temporary occlusion of the target vessel, especially in patients with poor cardiac contractile function [2,3]. Thus, improving the perioperative management by reducing myocardial ischemia–reperfusion damage is of paramount importance.

Recent work from clinical trials provides strong evidence that certain anesthetic agents have the potential to decrease myocardial ischemic damage, albeit to a certain degree [4,5]. Sevoflurane, as a kind of inhalational anaesthetic agent, has been indicated to exert a protective function in the preconditioning of cardiac events in patients treated by CABG [6]. Another study has demonstrated that sevoflurane remarkably reduces the size of infarcts as well as Ca²⁺ loading to protect the myocardium against reperfusion injury [5]. In addition, a former study has indicated that sevoflurane provides the myocardial protection in CABG surgery via downregulating the expression of troponin I [7]. Nevertheless, the underlying molecular mechanisms of sevoflurane for cardioprotection remain to be elucidated.

Currently, microarray analysis is broadly employed to elucidate the underlying mechanisms of various diseases, and is helpful for determining the potential biomarkers of diseases, because of the lower expense and advancements in this technique [8]. Thus, microarray technology might be useful to determine the effects of sevoflurane on gene expression in patients treated by CABG. In an attempt to determine the influence of sevoflurane on postoperative recovery in patients undergoing CABG, a novel method named as EgoNet [9] on the basis of egocentric network-analysis technique was developed in this study. This method searches significant sub-networks which are functionally related with diseases. The sub-networks identified by our method are defined as modules, which are well-defined in the study of social networks [10]. Firstly, a dataset from the EMBL-EBI database (E-GEOD-4386) was downloaded for subsequent analysis. Then, the differential co-expression networks (DCNs) were constructed. Next, snowball sampling was used to collect all modules. Thereafter, the significance of the identified candidate modules was assessed using random permutation test. Finally, functional analysis for differential modules was carried out to illuminate the biological processes.

2 Materials and methods

2.1 Gene expression profile

The gene expression data under the series number of E-GEOD-4386 [11] were recruited from the EMBL-EBI database from the platform of A-AFFY-44 -Affymetrix GeneChip Human Genome U133 Plus 2.0 [HG-U133_ Plus_2]. There were 40 samples in the data profile E-GEOD-4386, including data from patients treated by CABG plus propofol treatment (n = 10), sevoflurane treatment (n = 10) and control samples (n = 20). The control samples were composed of the same patients before CABG surgery. In the present analysis, to evaluate the influence of sevoflurane on postoperative recovery in patients undergoing CABG, we only selected atrial samples prior to and following CABG with sevoflurane treatment. The average age of patients was 65.2 years. All the patients were male. Moreover, patients with hemodynamic instability were not included. We implemented microarray analysis according to E-GEOD-4386, where total RNA was extracted from the frozen cardiac tissue using an RNeasy Mini kit.

2.2 Pretreatment of raw data

The annotation file in TXT format was utilized to map the association between probes and gene symbols. Then, background adjustment was performed using RMA, and quantile normalization was carried out through quantiles. Simultaneously, PM/MM correction was implemented via MAS, followed by summarization of the expression measures using median polish. Next, the probes were mapped to gene symbols, and finally, we obtained 20545 genes.

2.3 Protein-protein interaction network (PPIN)

Since complicated biological processes were usually regulated by tightly connected proteins, and the PPIN was further analyzed based on the public database STRING (Search Tool for the Retrieval of Interacting Genes/ Proteins, version 9.1, http://string-db.org/) [12].

To begin with, the ensemble PPIN including 787896 interactions and 16730 genes were downloaded from the STRING database. Then, the intersection of interactions between the ensemble PPIN and microarray data were extracted. Finally, a new PPIN including 48778 interactions and 7953 genes was obtained.

2.4 DCN construction

For the patients undergoing CABG with sevoflurane treatment, there were two steps in the process of DCN construction.

Firstly, the construction of a binary co-expression network was implemented. To construct the binary gene co-expression network, edges were picked out according to the absolute value of the Pearson correlation coefficient (PCC) of the expression profiles of two genes. In order to take out the indirect correlation because of a third gene, the 1st order partial PCC was used [13]. Only edges whose correlations were greater than the pre-defined threshold δ were chosen. In our study, the δ value was set at 0.8, and thus maximal number of genes was connected in the DCN to be constructed.

Subsequently, weight values were assigned to the edges of the binary gene co-expression network on the basis of the P values of differential gene expression between the sevoflurane and control groups. In our analysis, we utilized one tail t-test to identify differential gene expression. The weight w_{i, j} on edge (i, j) in the DCN was computed as:

wi,j=log⁡pi+log⁡Pj1/22∗maxl∈v|log⁡PI|1/20,,ifcor(i,j)≥δ,if cor(i,yj)<δ,

In this formula, P_i and P_j were the respective P values of differential expression for gene i and gene j. V was the node set in the DCN. cor(i, j) represented the absolute value of PCC between gene i and j according to their expression profiles. Under the weighting scheme, genes were assigned higher weight values when the genes were co-expressed and remarkably differentially expressed, which met our hypothesis that these genes were likely involved in a biological pathway that showed differential activities between the two groups.

2.5 Overview of the EgoNet algorithm for identification of human disease modules

In order to identify significant sub-networks that were functionally related with diseases, a novel method called EgoNet was proposed to exhaustively search and prioritize sub-networks and bio-markers from a large-scale biological network. The type of sub-network identified by our method was defined as module. Specifically, a module was a part of a network that involved a special node we were focusing on, which we called seed gene. Except seed gene, the network was made of a neighborhood including all nodes to which the seed gene was connected to at a certain path length. The hypothesis of our model was that if the majority of neighbors of a central disease gene were disease genes, and the other neighbors were likely to be involved in the disease pathway. Alternatively, if most neighbors of the seed gene were related with a disease, the seed gene itself was regarded highly likely to exert a function in the disease. We planned to identify the hidden genes which exhibited no significance by themselves but were clustered in a sub-network in which genes collectively were highly predictive of disease state.

This algorithm took the PPIN and microarray data as input, and then EgoNet repeatedly scanned through all genes with two or more neighbors in the network. With each initial node (seed gene), it found the score of the level-one module based on how well the genes as a collection predicted the classification accuracy. Next, it spread outward from the seed gene progressively to involve more genes in the predictive model. The procedure stopped when the classification accuracy decreased. The process of growing module was known as snowball sampling. After getting the score of a module, the statistical significance was assessed by permutation test.

2.6 Identification of candidate modules

The objective of EgoNet is to extract significant modules from microarray profile and biological network data. The EgoNet algorithm consisted of three main steps: selection of seed genes, search of modules using snowball sampling, and refinement of modules.

2.6.1 Selection of seed genes

In the stage of seed genes selection, genes in the DCN were ranked using the topological measurement (degree) analysis. In brief, for DCN, Gk = (V, E_k) (1 ≤ k ≤ M) with an adjacency matrix A_k = (a_ijk)_nxn, we constructed a function to denote the importance of vertex i:

g(i)=∑j∈Nk(i)A'ijkg(j)

In this function, N_k(i) stood for the set of neighbors of gene i in G_k; g(i) denoted the importance of vertex i in the DCN, Ak′ was the degree normalized weighted adjacency matrix which was computed based on Ak′=D−1/2AkD1/2 where D was diagonal matrix.

For each gene, after acquiring its ranks in DCN, marked as g = [g⁽¹⁾,…, g^(M)], a z-score was computed for each rank g^(l). Subsequently, the rank was obtained for that gene in the DCN via averaging the z-scores. The top 5% genes in the DCN were extracted as the seed genes.

2.6.2 Module search

This stage exhaustively searched all genes v whose addition caused the classification accuracy increase until there was no increase. For a given seed gene v ∈ V, it was treated as a module C={v}. For each vertex u in its neighborhood in the DCN, N(v) was defined as N(v) = U_iN_i(v) where N_i(v) was the neighbor set in G_i as the candidate for module C. For each u ∈ N(v), the classification accuracy increase was calculated between the new module C’ = CU{v} and C, that was to say, ΔS (C’, C) = S(C) - S(C’). if ΔS (C’, C) was more than 0, which meant the addition of gene u improved the classification accuracy of the module C. The gene u whose addition maximizes ΔS was added to the module C. If there were more than one node which were contained at each stage, we randomly selected one. The spreading stage stopped when the classification accuracy dropped. The above process of growing module was known as snowball sampling [14].

2.6.3 Refinement of candidate modules

In the refinement stage, the modules with classification accuracy smaller than 0.9 were removed. Moreover, the modules with gene size less than 5 were eliminated.

2.7 Statistical significance of the identified modules

A test of significance was implemented for the identified candidate modules to obtain the statistical significance. The null distribution of classification accuracy was obtained by randomly permuting 1000 times and computing the score from the same module each time. Then, the actual score of the module was indexed on the null distribution to get the P value which was calculated as the probability of the module having smaller classification accuracy by chance based on the following formula:

Pvalue=sum(S(HR)>S(HC))/count(HR)

Where count (H_R) was the number of random permutation test, S (H_R) denoted the classification accuracy value of modules produced by random permutation test. S (H_C) stood for the classification accuracy value of candidate modules.

After obtaining P-values, the Benjamini & Hochberg [15] method was used to calculate the false discovery rate (FDR) to further adjust the P-values. In our study, FDR ≤ 0.05 was considered as significant.

2.8 Pathway enrichment analysis

The development of diseases is frequently caused by altered pathways which participated in the biological process. For this, pathway enrichment analyses for differential module genes were performed. In the present analysis, all reference pathways in human were downloaded from Reactome database. Subsequently, the intersection of the genes in each reference pathway with PPIN genes was extracted. Then, a set of informative pathways were generated for subsequent analysis after removing pathways including less than 5 genes or more than 100 genes. Pathways with too few genes might not have sufficient biological information, and pathways with too many genes may be too generic. Next, genes in differential modules were aligned to each reference pathways to extract the pathways enriched by each differential module. Fisher’s exact test was used to evaluate the enrichment effects. Afterwards, we used Benjamini & Hochberg method to calculate the FDR to further adjust the P-values. In our study, pathways with FDR less than 0.05 were considered as the pathways enriched by a given module. Then, we ranked these pathways based on FDR values, and the pathway with the smallest FDR was selected as the significant pathway of a given module.

2.9 Power analysis to calculate the sample size

In the microarray profile of E-GEOD-4386, there were 4 groups (sevoflurane treatment group, baseline sevoflurane group, propofol treatment group, baseline propofol) which included 40 atrial samples collected at the beginning and at the end of the offpump CABG surgery. Among these 40 samples, 20 atrial samples were obtained from 10 patients receiving the anesthetic gas sevoflurane and 20 atrial samples were derived from 10 patients receiving the intravenous anesthetic propofol. In the current study, we utilized G*Power 3.1.9.2 software to make power analysis to calculate the sample size using α = 0.05 and power(1-β) = 0.80 based on the results of NT-proBNP and the expression level of single seed genes.

Ethical approval

The conducted research is not related to either human or animals use.

3 Results

3.1 Construction of DCN and identification of candidate modules

To construct gene co-expression network, edges were selected according to the absolute value of the PCC of the expression profiles of two genes. A total of 2440 nodes were selected to construct the DCN based on the δ greater than 0.8.

Based on the z-score distribution of 2440 nodes in DCN, we selected the genes with more than one connection and with the top 5% z-score value to serve as seed genes. Totally, 122 seed genes were identified. The top 20 seed genes are listed in Table 1. Subsequently, from every seed gene, we progressively grew the modules by levels. From every module, this procedure stopped when the classification accuracy dropped with the growth. After that, we extracted 122 modules and the mean node size in a module was 3. By setting the classification accuracy cutoff at 0.9 and the number of nodes in a module at 5, a total of 7 modules were identified, including module 80, 82, 83, 84, 85, 86 and 89.

Table 1

The top 20 seed genes identified in the differential co-expression network (DCN) and the distribution of their z-scores

Node	Z-score	Node	Z-score
RAD51AP1	46.72203	CDC6	26.38525
TRIP13	37.74864	NMUR2	26.05800
RPS6	33.30842	EIF3E	25.09675
RPL8	29.70821	TTK	24.58506
DLGAP5	29.35711	CDC45	24.54506
RPL3	29.21577	RPS3	23.60314
RPL6	28.73787	RPLP0	23.59052
RPL37	28.23891	LPAR1	23.57449
CEP55	26.68925	MCHR2	23.50679
SSR1	26.45610	P2RY10	23.46214

3.2 Significance of candidate modules

After random permutation test with 1000 times, we obtained the FDR distribution of the identified modules, and we found that the FDR values of these 7 candidate modules were less than 0.05, which meant these 7 modules were significant (Table 2). The composition of these differential modules is shown in Figure 1.

Figure 1

The composition of the differential modules. A, B, C, D, E, F, and G are modules 80, 82, 83, 84, 85, 86 and 89 respectively. Yellow nodes are the seed genes.

Table 2

A total of 7 differential modules based on FDR ≤ 0.05

Modules	Accuracy	Count_members	Seed genes
Module 80	1	6	MCHR2
Module 82	1	5	PNOC
Module 83	1	6	NMUR1
Module 84	1	5	PMCH
Module 85	1	6	NPY5R
Module 86	1	5	LPAR1
Module 89	1	5	AURKB

Note: FDR, false discovery rate.

3.3 Pathway enrichment analysis

After removing pathways with less than 5 genes or more than 100 genes, we ended up with 1136 pathways for enrichment analysis. After multiple test, we found that the genes of the module 80, 82, and 85 were all enriched in the pathway of chemokine receptors bind chemokines (FDR = 4.30E-06, 1.73E-06, 3.47E-04, respectively), the genes in the module 83 were enriched in the pathway of ERK2 activation (FDR = 5.04E-03). The specific information is shown in Table 3.

Table 3

Pathway enrichment analysis of genes in differential modules

Modules	Pathways	FDR
Module 80	Chemokine receptors bind chemokines	4.30E-06
Module 82	Chemokine receptors bind chemokines	1.73E-06
Module 83	ERK2 activation	5.04E-03
Module 84	Regulation of Complement cascade	6.44E-03
Module 85	Chemokine receptors bind chemokines	3.47E-04
Module 86	TRAF6 Mediated Induction of proinflammatory cytokines	1.07E-03
Module 89	Resolution of Sister Chromatid Cohesion	2.62E-05

3.4 Power analysis

In the current study, we utilized G^*Power 3.1.9.2 software to make power analysis to calculate the sample size using α=0.05 and power(1-β) = 0.80 based on the results of NT-proBNP and the expression level of single seed gene (IL6, CCL20, and CXCL1).

We assumed that the ratio of sample size in the sevoflurane treatment group and control group was 1:1. Statistical power values suggested that the total samples were 36 based on the results of NT-proBNP. Moreover, statistical power values indicated that the total samples were respective 26, 30 and 26 based on the expression level of seed genes IL6, CCL20, and CXCL1. Significantly, in our study, the patients in sevoflurane treatment group and control groups were 10 and 10, respectively. According to these results, we thought the overall power value was relatively good, and the sample size was sufficient for our study.

4 Discussion

Former studies have demonstrated that sevoflurane, to a certain degree, plays cardioprotective role in CABG surgery [16,17]. However, the underlying mechanisms of the effects of sevoflurane are still unclear. In the current study, in order to determine the protective molecular mechanisms of sevoflurane, the dataset E-GEOD-4386 was selected for further analysis. After snowball sampling, we extracted 122 candidate modules and the mean nodes in a module was 3. By setting the classification accuracy cutoff at 0.9 and the number of nodes in a module at 5, a total of 7 modules were identified, including module 80, 82, 82, 84, 85, 86 and 89. Significantly, these 7 modules were differential. Moreover, pathway analysis showed that genes in the module 80, 82, and 85 were all enriched in the pathway of chemokine receptors bind chemokines.

In the past decade, more effort has been done to extract differentially expressed genes (DEG) among different disease conditions, which can be employed as diagnostic signatures for classifying disease conditions or predicting clinical outcomes [18,19]. Nevertheless, bio-markers based on gene expression data alone are not reliable [20]. To solve this problem, network-based approaches are rapidly developed to obtain a comprehensive understanding of the disease process [21,22]. Network-based methods have many potential biological and clinical applications, including a better understanding of the effects of interconnection of disease genes and disease pathways, which, in turn, may offer better targets for drug development. So far, many methods have been created to integrate microarray profile with PPIN or pathway databases to identify key sub-network markers for predicting clinical outcomes [23,24]. For example, network-based Support Vector Machine (SVM) and Network-Guided Forests (NGF) have been utilized to select a set of genes which maximize the prediction performance [25,26]. However, the methods mentioned above are largely heuristic, and the definition of the sub-networks is ambiguous without a formal topological feature. As consequence, identified network modules tend to include only crucial genes based on their expression level, yet exclude non-DEGs despite that they are functionally linked to differentially expressed disease genes. Thus, a novel method called EgoNet was created to extract key sub-networks, and select the hidden genes which showed no significance by themselves, but were clustered in a module. More importantly, the ego-networks have been indicated to play central role in the identification of novel disease genes [27]. A former study has indicated that there is the interaction between improved glucose and CCL20 [28]. Significantly, sevoflurane has been indicated to suppress glucose-induced insulin secretion and then result in insulin resistance [29]. More importantly, insulin resistance has been demonstrated to be a risk factor for coronary heart disease [30]. Another gene CXCL6 has been reported to be associated with coronary artery disease [31]. Moreover, CXCL1 has also been exhibited to impair endothelium-dependent vasorelaxation in coronary arteries [32]. We explored these 3 genes in the identified modules. Interestingly, we observed that these 3 genes were clustered in one module (Figure 1 A and B). This finding was consistent with the property of disease networks-proteins involved in the same disease had an increased tendency to interact with each other [21]. Accordingly, our observations indicate that EgoNet can fully identify the novel signatures in a more accurate way and provide a deeper understanding of their roles in complex diseases.

As reported, inflammatory mediators released as a consequence of intraoperative events, for instance, ischemic trauma as well as reperfusion, may activate endothelial cells in remote organs [33]. This response to ischemia and reperfusion can induce leukocyte-dependent microvascular injury which is the feature of multiple organ dysfunction syndrome [34]. Moreover, inflammation exerts an important role in vascular repair [35]. Inflammatory cell extravasation depends on chemoattractants including chemokines, cytokines which stimulate inflammatory leukocyte chemotaxis, proliferation, and apoptosis [36]. Chemokines interact with chemokine receptors in a complex [36]. Significantly, chemokines CCL2, CCL11, and CXCL8, as well as chemokine receptors like CCR2 and CCR3 have been demonstrated to be related with vascular injury [37,38]. In addition, the expression of chemokines, for example, CCL2, CCL4, CXCL2, and CXCL3, has been reported to be increased in patients following CABG [39]. Furthermore, a previous study has demonstrated that sevoflurane shows direct antiinflammatory effects on cultured human kidney cells following ischemiareperfusion injury [40]. Several studies have also demonstrated that sevoflurane regulates the interaction of polymorphonuclear neutrophils and the microvascular endothelium, and this may play a key role in the initiation of reperfusion injury, even when cardiopulmonary bypass is performed [41,42]. Notably, in the current study, it was identified that the genes of the module 80, 82, and 85 were significantly enriched in the pathway of chemokine receptors bind chemokines. Thus, understanding the inflammation state of patients before CABG procedure is crucially important to decrease the post-operative complications.

Nevertheless, the sub-network markers involved in the molecular mechanisms underlying sevoflurane cardioprotective effects needs further confirmation at the gene and protein levels by means of reverse transcription quantitative polymerase chain reaction or western blot analyses. Hence, subsequent investigations are warranted, aiming to include a cell culture system or animal models to verify the results of our study. Taken together, sevoflurane might exert important functions in reducing myocardial reperfusion injury in patients following CABG, partially through regulating the pathway of chemokine receptors bind chemokines. However, the potential mechanisms for the effects of sevoflurane in cardioprotection need further confirmation.

Conflict of interest: Authors state no conflict of interest.

Reference

[1] Møller C., Penninga L., Wetterslev J., Steinbrüchel D., Gluud C., Off-pump versus on-pump coronary artery by-pass grafting for ischaemic heart disease, Cochrane Libr., 2012, 410.1002/14651858.CD007224.pub2Search in Google Scholar

[2] Selvanayagam J.B., Petersen S.E., Francis J.M., Robson M.D., Kardos A., Neubauer S., et al., Effects of Off-Pump Versus On-Pump Coronary Surgery on Reversible and Irreversible Myocardial Injury A Randomized Trial Using Cardiovascular Magnetic Resonance Imaging and Biochemical Markers, Circulation, 2004, 109, 345-35010.1016/j.accreview.2004.04.064Search in Google Scholar

[3] Orhan G., Sargin M., Senay S., Yuksel M., Kurc E., Tasdemir M., et al., Systemic and myocardial inflammation in traditional and off-pump cardiac surgery, Texas Heart Institute Journal, 2007, 34, 160Search in Google Scholar

[4] Frässdorf J., De Hert S., Schlack W., Anaesthesia and myocardial ischaemia/reperfusion injury, British journal of anaesthesia, 2009, aep14110.1093/bja/aep141Search in Google Scholar

[5] Varadarajan S.G., An J., Novalija E., Stowe D.F., Sevoflurane before or after ischemia improves contractile and metabolic function while reducing myoplasmic Ca2+ loading in intact hearts, The Journal of the American Society of Anesthesiologists, 2002, 96, 125-13310.1097/00000542-200201000-00025Search in Google Scholar

[6] Lucchinetti E., Da S.R., Pasch T., Schaub M.C., Zaugg M., Anaesthetic preconditioning but not postconditioning prevents early activation of the deleterious cardiac remodelling programme: evidence of opposing genomic responses in cardioprotection by pre- and postconditioning, Bja British Journal of Anaesthesia, 2005, 95, 140-15210.1093/bja/aei155Search in Google Scholar

[7] Yao Y.T., Li L.H., Sevoflurane Versus Propofol for Myocardial Protection in Patients Undergoing Coronary Artery Bypass Grafting Surgery: a Meta-analysis of Randomized Controlled Trials, Chinese Medical Sciences Journal, 2009, 24, 133-14110.1016/S1001-9294(09)60077-8Search in Google Scholar

[8] Nakanishi T., Oka T., Akagi T., Recent advances in DNA microarrays, Acta Medica Okayama, 2002, 55, 319-328Search in Google Scholar

[9] Yang R., Bai Y., Qin Z., Yu T., EgoNet: identification of human disease ego-network modules, BMC genomics, 2014, 15, 110.1186/1471-2164-15-314Search in Google Scholar PubMed PubMed Central

[10] Borgatti S.P., Mehra A., Brass D.J., Labianca G., Network analysis in the social sciences, Science, 2009, 323, 892-89510.1126/science.1165821Search in Google Scholar PubMed

[11] Lucchinetti E., Hofer C., Bestmann L., Hersberger M., Feng J., Zhu M., et al., Gene regulatory control of myocardial energy metabolism predicts postoperative cardiac function in patients undergoing off-pump coronary artery bypass graft surgery: inhalational versus intravenous anesthetics, Anesthesiology, 2007, 106, 444-45710.1097/00000542-200703000-00008Search in Google Scholar PubMed

[12] Franceschini A., Szklarczyk D., Frankild S., Kuhn M., Simonovic M., Roth A., et al., STRING v9. 1: protein-protein interaction networks, with increased coverage and integration, Nucleic acids research, 2013, 41, D808-D81510.1093/nar/gks1094Search in Google Scholar PubMed PubMed Central

[13] Watson-Haigh N.S., Kadarmideen H.N., Reverter A., PCIT: an R package for weighted gene co-expression networks based on partial correlation and information theory approaches, Bioinformatics, 2010, 26, 411-41310.1093/bioinformatics/btp674Search in Google Scholar

[14] Goodman L.A., Snowball sampling, The annals of mathematical statistics, 1961, 148-17010.1214/aoms/1177705148Search in Google Scholar

[15] Benjamini Y., Drai D., Elmer G., Kafkafi N., Golani I., Controlling the false discovery rate in behavior genetics research, Behavioural brain research, 2001, 125, 279-28410.1016/S0166-4328(01)00297-2Search in Google Scholar

[16] Soro M., Gallego L., Silva V., Ballester M.T., Lloréns J., Alvariño A., et al., Cardioprotective effect of sevoflurane and propofol during anaesthesia and the postoperative period in coronary bypass graft surgery: a double-blind randomised study, European Journal of Anaesthesiology (EJA), 2012, 29, 561-56910.1097/EJA.0b013e3283560aeaSearch in Google Scholar PubMed

[17] He W., Zhang F.-j., Wang S.-p., Chen G., Chen C.-c., Yan M., Postconditioning of sevoflurane and propofol is associated with mitochondrial permeability transition pore, Journal of Zhejiang University Science B, 2008, 9, 100-10810.1631/jzus.B0710586Search in Google Scholar PubMed PubMed Central

[18] Stratford J.K., Bentrem D.J., Anderson J.M., Fan C., Volmar K.A., Marron J., et al., A six-gene signature predicts survival of patients with localized pancreatic ductal adenocarcinoma, PLoS Med, 2010, 7, e100030710.1371/journal.pmed.1000307Search in Google Scholar PubMed PubMed Central

[19] Boutros P.C., Lau S.K., Pintilie M., Liu N., Shepherd F.A., Der S.D., et al., Prognostic gene signatures for non-small-cell lung cancer, Proceedings of the National Academy of Sciences, 2009, 106, 2824-282810.1073/pnas.0809444106Search in Google Scholar PubMed PubMed Central

[20] Venet D., Dumont J.E., Detours V., Most random gene expression signatures are significantly associated with breast cancer outcome, PLoS Comput Biol., 2011, 7, e100224010.1371/journal.pcbi.1002240Search in Google Scholar PubMed PubMed Central

[21] Barabási A.-L., Gulbahce N., Loscalzo J., Network medicine: a network-based approach to human disease, Nature Reviews Genetics, 2011, 12, 56-6810.1038/nrg2918Search in Google Scholar PubMed PubMed Central

[22] Chan S.Y., Loscalzo J., The emerging paradigm of network medicine in the study of human disease, Circulation research, 2012, 111, 359-37410.1161/CIRCRESAHA.111.258541Search in Google Scholar PubMed PubMed Central

[23] Winter C., Kristiansen G., Kersting S., Roy J., Aust D., Knösel T., et al., Google goes cancer: improving outcome prediction for cancer patients by network-based ranking of marker genes, PLoS Comput Biol., 2012, 8, e100251110.1371/journal.pcbi.1002511Search in Google Scholar

[24] Nie Y., Yu J., Mining breast cancer genes with a network based noise-tolerant approach, BMC systems biology, 2013, 7, 110.1186/1752-0509-7-49Search in Google Scholar

[25] Zhu Y., Shen X., Pan W., Network-based support vector machine for classification of microarray samples, BMC bioinformatics, 2009, 10, 110.1186/1471-2105-10-S1-S21Search in Google Scholar

[26] Dutkowski J., Ideker T., Protein networks as logic functions in development and cancer, PLoS Comput Biol., 2011, 7, e100218010.1371/journal.pcbi.1002180Search in Google Scholar

[27] Jordán F., Nguyen T.-P., Liu W.-c., Studying protein–protein interaction networks: a systems view on diseases, Briefings in functional genomics, 2012, 11, 497-50410.1093/bfgp/els035Search in Google Scholar

[28] Papadimos T., Pappada S., Lather J., Bazalitski V., Stawicki S., Cameron B., et al., Correlation between CCL20 and serum glucose in postoperative coronary bypass patient: A call for further investigation, OPUS, 2010, 12, 1-210.4103/IJAM.IJAM_101_16Search in Google Scholar

[29] Saho S., Kadota Y., Sameshima T., Miyao J., Tsurumaru T., Yoshimura N., The effects of sevoflurane anesthesia on insulin secretion and glucose metabolism in pigs, Anesthesia & Analgesia, 1997, 84, 1359-136510.1213/00000539-199706000-00034Search in Google Scholar

[30] Ai M., Otokozawa S., Asztalos B.F., White C.C., Cupples L.A., Nakajima K., et al., Adiponectin: an independent risk factor for coronary heart disease in men in the Framingham offspring study, Atherosclerosis, 2011, 217, 543-54810.1016/j.atherosclerosis.2011.05.035Search in Google Scholar

[31] Lundberg G.A., IN CELL RECRUITMENT, ACTIVATION, Chemokine Research Trends, 2007, 171Search in Google Scholar

[32] Laing K.J., Secombes C.J., Chemokines, Developmental & Comparative Immunology, 2004, 28, 443-46010.1016/j.dci.2003.09.006Search in Google Scholar

[33] Carden D.L., Granger D.N., Pathophysiology of ischaemia–reperfusion injury, The Journal of pathology, 2000, 190, 255-26610.1002/(SICI)1096-9896(200002)190:3<255::AID-PATH526>3.0.CO;2-6Search in Google Scholar

[34] Cremer J., Martin M., Redl H., Bahrami S., Abraham C., Graeter T., et al., Systemic inflammatory response syndrome after cardiac operations, The Annals of Thoracic Surgery, 1996, 61, 1714-172010.1016/0003-4975(96)00055-0Search in Google Scholar

[35] Farb A., Weber D.K., Kolodgie F.D., Burke A.P., Virmani R., Morphological predictors of restenosis after coronary stenting in humans, Circulation, 2002, 105, 2974-298010.1161/01.CIR.0000019071.72887.BDSearch in Google Scholar PubMed

[36] Gerard C., Rollins B.J., Chemokines and disease, Nature immunology, 2001, 210.1038/84209Search in Google Scholar PubMed

[37] Cipollone F., Marini M., Fazia M., Pini B., Iezzi A., Reale M., et al., Elevated circulating levels of monocyte chemoattractant protein-1 in patients with restenosis after coronary angioplasty, Arteriosclerosis, thrombosis, and vascular biology, 2001, 21, 327-33410.1161/01.ATV.21.3.327Search in Google Scholar

[38] Charo I.F., Taubman M.B., Chemokines in the pathogenesis of vascular disease, Circulation research, 2004, 95, 858-86610.1161/01.RES.0000146672.10582.17Search in Google Scholar PubMed

[39] Voisine P., Ruel M., Khan T.A., Bianchi C., Xu S.-H., Kohane I., et al., Differences in gene expression profiles of diabetic and nondiabetic patients undergoing cardiopulmonary bypass and cardioplegic arrest, Circulation, 2004, 110, II-280-II-28610.1161/01.CIR.0000138974.18839.02Search in Google Scholar PubMed

[40] Lee H.T., Kim M., Jan M., Emala C.W., Anti-inflammatory and antinecrotic effects of the volatile anesthetic sevoflurane in kidney proximal tubule cells, American Journal of Physiology-Renal Physiology, 2006, 291, F67-F7810.1152/ajprenal.00412.2005Search in Google Scholar PubMed

[41] Heindl B., Reichle F.M., Zahler S., Conzen P.F., Becker B.F., Sevoflurane and isoflurane protect the reperfused guinea pig heart by reducing postischemic adhesion of polymorpho-nuclear neutrophils, The Journal of the American Society of Anesthesiologists, 1999, 91, 521-53010.1097/00000542-199908000-00027Search in Google Scholar PubMed

[42] Mobert J., Zahler S., Becker B.F., Conzen P.F., Inhibition of neutrophil activation by volatile anesthetics decreases adhesion to cultured human endothelial cells, The Journal of the American Society of Anesthesiologists, 1999, 90, 1372-138110.1097/00000542-199905000-00022Search in Google Scholar PubMed