Network Biology Analysis of the Human Abdominal Aortic Aneurysm Plasma Proteome

Research Article

Austin J Proteomics Bioinform & Genomics. 2014;1(2): 10.

Network Biology Analysis of the Human Abdominal Aortic Aneurysm Plasma Proteome

Alessandra Modesti1*, Tiziana Alberio2, Tania Gamberi1, Mauro Fasano2, Francesca Magherini1, Tania Fiaschi1, Merry L. Lindsey3 and Pietro A. Modesti4*

1Dipartimento di Scienze Biomediche, Sperimentali e Cliniche, Università degli Studi di, Firenze, Italy

2Dipartimento di Scienze Teoriche e Applicate, Biomedical Research Division, Università, dell'Insubria, Italy

3Mississippi Center for Heart Research, Department of Physiology and Biophysics, University of Mississippi Medical Center, and Research Service, G.V. (Sonny) Montgomery,Veterans Affairs Medical Center, Jackson, MS

4Dipartimento di Medicina Sperimentale e Clinica, Università degli Studi di Firenze, Italy

*Corresponding author: Alessandra Modesti - Dipartimento di Scienze Biomediche, Sperimentali e Cliniche,Viale G. Morgagni 50-50134. Università degli Studi di Firenze, Italy

*Corresponding author: Pietro Amedeo Modesti - Dipartimento di Medicina Sperimentale e Clinica Largo Brambilla, 3-50134, Università degli Studi di Firenze, Italy

Received: September 30, 2014; Accepted: November 24, 2014; Published: November 26, 2014

Abstract

Objective: Abdominal aortic aneurysm (AAA) affects up to 9% of adults, and the incidence of asymptomatic and ruptured AAA currently accounts for 1-2% of male deaths. A number of studies have applied classical proteomics methodology to identify plasma biomarkers of AAA.

The use of a systems biology protein-protein interaction network analysis on proteomics results can reveal novel mechanisms.

Methods: We performed a bioinformatics analysis on a compiled set of proteins previously identified in multiple proteomics studies. Results from a total of nine papers were identified and included in the analysis. The 64 proteins related to AAA were first analyzed by over-representation analysis (ORA) using Webgestalt. Afterwards, the same input list was submitted to BioProfiling. This portal used statistical methodology for the network based interpretation of the protein list (Rspider, PPIspider). The networks obtained by both Reactome and PPI analysis were further analyzed by ORA using the JEPETTO application in the Cytoscape environment.

Results: This analysis revealed a strong over-representation for proteins related not only to blood clotting but also to cellular mediated response, such as cell adhesion and cytokine activation of platelets and white blood cells.

Conclusions: We used a network strategy to generate statistically valid novel hypotheses about biological mechanisms related to AAA, which may be useful in providing new insights into the understanding of the pathogenesis of AAA.

Keywords: Abdominal aortic aneurysm; Proteomics; Network analysis

Introduction

Abdominal aortic aneurysm (AAA) affects up to 9% of adults older than 65 years of age [1]. The mortality rate for patients with ruptured AAA approaches 90%, with 40-50% of deaths occurring before the patient reaches the hospital. Asymptomatic AAA is easily detectable with abdominal ultrasound, and screening programs are now considered a reasonable practice option since surgical repair is restricted to AAA >5.5 cm at centers with acceptable operative mortality [2-6]. Identifying markers other than AAA size that would predict a risk of rupture is the focus of current investigation. For AAA, one significant limitation relates to the fact that many biomarkers are not disease specific; indeed most of them are also markers for atherosclerosis [7]. Serum Elastin Peptide holds promise to be an improved biomarker to predict AAA expansion and rupture [8-10] because it positively correlates between its plasma level and the AAA growth rate [10]. The extent of aortic dilation and vessel tortuosity is associated with blood levels of several fibrinolytic factors, including D-dimer, fibrinogen/fibrin degradation products, and plasmin inhibitor-plasmin complexes [11]. Specifically, there is a positive correlation between AAA size and fibrinogen concentration in the plasma [12]. For AAA many of the identified differences in the serum proteome often represented proteins highly abundant in serum rather than true candidate biomarkers [13]. Using a proteomic approach, our team has confirmed several plasma biomarkers that others have previously identified as being associated with AAA [7-14]. New approaches to quantitatively identify low-abundance proteins (such as cytokines) and their functional analysis into specific protein networks are of crucial importance in the post-genomic era [15-17]. A classical proteomics approach has been used to identify in AAA subjects four serum proteins that show altered expression profiles specifically linked to AAA pathology [18,19]. Proteome analysis has been generally limited in complex samples such as serum and plasma that have a dramatically wide dynamic range of protein abundance. The use of pre-fractionated plasma samples revealed that patients with small AAA showed an increased level of the enzyme glycosyl phosphatidyl inositol-specific phospholipase D compared to controls without aneurysm [20]. However, smaller proteins may be part of larger protein complexes and hence the removal of proteins involved in complexes with high-abundance proteins such as albumin may inhibit the search for disease biomarkers [7]. In recent years, bioinformatics approaches have been applied to provide systems biology interpretations of proteomic data [21]. The focus of the present study was the application of bioinformatics to analyze all the proteins lists obtained by proteomics experiments performed on human AAA serum or plasma samples. A common approach to the visualization and examination of "omics" data involves the generation of a network of all the individual components of a given setof experiments [22,23]. We used a network strategy to generate statistically valid hypotheses about biological mechanisms related to AAA and to provide new insight into the understanding of pathogenesis of such a complex disease [24]. To proceed we carried out a detailed bioinformatics analysis (using open source software for analyses/interpretation of biological data and visualization of complex networks obtained) on proteins identified in previous proteomics. Such an approach is being used with increased frequency, and such applications increase our understanding of complex human diseases including cardiovascular diseases [25].

Methods

Literature search to obtain the protein input list

Publications with the medical subject heading "abdominal aortic aneurysms" and keywords "proteomics" and "plasma" and "progression" or "growth" or "expansion rate" or "rupture" were searched in the MEDLINE/PubMed and EMBASE databases. The search was restricted until May 2013. Only human studies that reported aneurysm size, expansion rates and plasma markers were selected.

Bioinformatics analysis

We performed over-representation analysis (ORA) of the protein list obtained by the literature search (the input list) using the Webgestalt online tool (https://bioinfo.vanderbilt.edu/webgestalt/) against Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), WikiPathway and Pathway Commons databases [26]. The analysis of functional genomics data often involves the assessment of potential functional associations between a protein set of interest. To identify these putative associations several enrichment analysis tools have been developed including over-representation analysis (ORA) techniques. This method assess the statistical over-representation of a user-defined, pre-selected gene/protein list of interest in a reference list of known gene/protein sets using a statistical test, e.g. the one-sided Fisher's exact test or the hypergeometric distribution.

Afterwards the web portal BioProfiling.de (https://www. bioprofiling.de/) was used to build enriched networks. This portal covered most of the available information regarding signaling and metabolic pathways (database: Reactome) and physical protein interactions (database: IntAct) [27]. Accordingly, the input list was analyzed by Rspider [28] and PPI spider [29,30]. Both tools employed advanced enrichment or network-based statistical frameworks. The p-value provided was computed by the Monte Carlo simulation (https://www.bioprofiling.de/statistical_frameworks.html) and referred to the probability of obtaining a model of the same quality for a random gene list of the same size [31]. The significant networks, using a cut-off of p < 0.01, were further considered for proteomics data interpretation and discussion. The enriched network was exported as an .xgmml file and visualized and modified by Cytoscape 3.1.0 (https://www.cytoscape.org/) [32]. Moreover, the Cytoscape Jepetto (Java Enrichment of Pathways Extended to Topology) application was employed to perform ORA of the enriched networks [33] (https://apps.cytoscape.org/apps/jepetto). For the enrichment analysis by JEPETTO, EnrichNet server is used. EnrichNet maps the input gene set onto a molecular interaction network and, using a random walk, scores distances between the genes and pathways/ processes in a reference database. This network-based association score (XD-score) is relative to the average distance to all pathways and represents a deviation (positive or negative) from the average distance. Significantly over-represented pathways were selected if theirXd score was greater than that calculated as the intercept of the linear regression of Xd scores vs the Fisher's exact test p-value after Benjamini-Hochberg correction (q-value; Figure 1) [34].