Pathway reconstruction of airway remodeling in chronic lung diseases: a systems biology approach

Abstract

Airway remodeling is a pathophysiologic process at the clinical, cellular, and molecular level relating to chronic obstructive airway diseases such as chronic obstructive pulmonary disease (COPD), asthma and mustard lung. These diseases are associated with the dysregulation of multiple molecular pathways in the airway cells. Little progress has so far been made in discovering the molecular causes of complex disease in a holistic systems manner. Therefore, pathway and network reconstruction is an essential part of a systems biology approach to solve this challenging problem. In this paper, multiple data sources were used to construct the molecular process of airway remodeling pathway in mustard lung as a model of airway disease. We first compiled a master list of genes that change with airway remodeling in the mustard lung disease and then reconstructed the pathway by generating and merging the protein-protein interaction and the gene regulatory networks. Experimental observations and literature mining were used to identify and validate the master list. The outcome of this paper can provide valuable information about closely related chronic obstructive airway diseases which are of great importance for biologists and their future research. Reconstructing the airway remodeling interactome provides a starting point and reference for the future experimental study of mustard lung, and further analysis and development of these maps will be critical to understanding airway diseases in patients.

Figure 1. A schematic picture demonstrating the pathway reconstruction workflow.

The workflow shows that the master gene list is provided from two sources (microarray gene expression and literature mining). Protein-Protein interaction network (PPi) and Gene Regulatory Network (GRN) are compiled using Cytoscape software and its plugins. Also, signaling pathway reconstruction is accomplished in CellDesigner software based on biological knowledge accumulated in related literatures and public databases (e.g. Gene Ontology, Uniprot, and NCBI geneRIFs) and related signaling pathways deposited in the signaling pathway databases such as Biocarta, Reactome, and NCI-PID. The final mustard signaling pathway was drawn manually using CellDesigner.

Figure 2. Major disease-risk modules in mustard lung network.

Colored nodes are selected based on fold change ranking in microarray gene list and previous reports. They have a central role in these modules as driver nodes. The A, B, and C sub-networks extract from the mustard lung network using Hubba, a plugin in Cytoscape software.

Figure 3. The proposed pathway of airway remodeling in mustard lung.

Pathway nodes are indicated by color coding; Green: single protein; White: protein complex; Yellow box: gene; Bisque: receptor protein; Purple: biological process. The pathway illustrates several paths such as ERK/MAPK, ERBB1/ERBB2 complex activation via TFF3, and EPAS1/ARNT transcription factor activation. Different paths were extracted from pathway databases (table 1) and then were curated and reconstructed manually in the CellDesigner graphical interface.

Figure 4. Analysis of mustard lung network with 172 nodes and 1169 edges.

These graphs were generated using NetworkAnalyzer plugin of Cytoscape. The scatter plot of betweenness centrality vs. number of neighbors (left) indicates that a limited number of nodes control the information flow between other nodes within the biological network. This means that a limited number of nodes with high interactions (hubs) control other nodes with lower interactions. The node degree distribution (right) shows that the network is scale-free considering the power-law degree distribution P(k)∼k^γ (fitting result is γ = 0. 790 and R-squared = 0.667). This means that the mustard lung network is a biological network which differs from random networks.