Network analysis identifies protein clusters of functional importance in juvenile idiopathic arthritis

Abstract

Introduction: Our objective was to utilise network analysis to identify protein clusters of greatest potential functional relevance in the pathogenesis of oligoarticular and rheumatoid factor negative (RF-ve) polyarticular juvenile idiopathic arthritis (JIA).

Methods: JIA genetic association data were used to build an interactome network model in BioGRID 3.2.99. The top 10% of this protein:protein JIA Interactome was used to generate a minimal essential network (MEN). Reactome FI Cytoscape 2.83 Plugin and the Disease Association Protein-Protein Link Evaluator (Dapple) algorithm were used to assess the functionality of the biological pathways within the MEN and to statistically rank the proteins. JIA gene expression data were integrated with the MEN and clusters of functionally important proteins derived using MCODE.

Results: A JIA interactome of 2,479 proteins was built from 348 JIA associated genes. The MEN, representing the most functionally related components of the network, comprised of seven clusters, with distinct functional characteristics. Four gene expression datasets from peripheral blood mononuclear cells (PBMC), neutrophils and synovial fluid monocytes, were mapped onto the MEN and a list of genes enriched for functional significance identified. This analysis revealed the genes of greatest potential functional importance to be PTPN2 and STAT1 for oligoarticular JIA and KSR1 for RF-ve polyarticular JIA. Clusters of 23 and 14 related proteins were derived for oligoarticular and RF-ve polyarticular JIA respectively.

Conclusions: This first report of the application of network biology to JIA, integrating genetic association findings and gene expression data, has prioritised protein clusters for functional validation and identified new pathways for targeted pharmacological intervention.

Figure 1

Overview of the workflow used to identify protein clusters of functional importance in rheumatoid factor-negative (RF-ve) polyarticular and oligoarticular juvenile idiopathic arthritis (JIA). This figure outlines each of the steps taken and gives cross references to the specific tables and figures that provide further details of that particular step in the process. (A) JIA genetic association data of replicated single nucleotide polymorphisms (SNPs) associated with RF-ve polyarticular JIA or oligoarticular JIA and JIA genome-wide association data findings were ascertained. SNPs were mapped to genes using Gene Relationships Across Implicated Loci (GRAIL) software (n = 348) (Additional file 1: Table S1A and B). (B) A JIA interactome network model (JIA Interactome) was inferred from JIA-associated genes using the BioGRID database (Figure 2A). (Bi) The top 10% of nodes, ranked by connectivity and bottlenecks, were used to derive a minimal essential network (MEN) of 248 genes (Figure 2B and Additional file 3: Table S3A). The Reactome database and spectral partition clustering was used to determine clusters of genes within the MEN (Figure 2C). (Bii) The Disease Association Protein-Protein Link Evaluator (Dapple) algorithm was used to identify Highly Connected Genes (Figure 3 and Additional file 3: Table S3B). The overlap of genes between Bi and Bii was determined (n = 26 genes) (Additional file 3: Table S3B). (C) JIA gene expression data from published sources, collated from the Gene Expression Omnibus database (GEO) (Additional file 2: Table S2 and Additional file 4: Table S4, Additional file 5: Table S5, Additional file 6: Table S6), was integrated with the overlap of the genes from Bi and Bii. (D) A prioritised list of JIA-associated genes with functionally associated network properties was identified from the integration of JIA genetic association and gene expression data (n = 3 genes) (Table 2). (E) MCODE was used to determine RF-ve polyarticular and oligoarticular JI-specific network clusters, protein:protein interactions of functional relevance (Figure 4A and B). GWAS, genome-wide association studies.

Figure 2

Network analysis of juvenile idiopathic arthritis (JIA)-associated genes. (A) A collated list of 348 JIA-associated genes was used to derive an interaction network inferred from the BioGRID model of the human interactome (3.2.99); red = JIA-associated gene, blue = inferred interaction (Figure 1B). (B) Minimal essential network of top 10% of the genes from the JIA interactome, ranked by connectivity and bottleneck, colours represent clusters of related genes calculated by spectral partition clustering [41] (Figure 1Bi). (C) Biological pathways associated with minimal essential network (MEN) clusters (colour of cluster relates to Figure 2B), false discovery rate (FDR) P-value of hypergeometric test (P ≤0.05) (Figure 1Bi). MHC Major histocompatibility complex, FGF Fibroblast growth factor, PDGF Platelet-derived growth factor IGF-1 Insulin-like growth factor 1, TGF beta Transforming growth factor beta, IL-2 Interleukin 2, IL-8 Interleukin 8 PI3K Phosphoinositide 3-kinase, TNF Tumour necrosis factor, CXCR1 chemokine (C-X-C motif) receptor 1, CXCR2 chemokine (C-X-C motif) receptor 2.

Figure 3

Changes in connectivity of juvenile idiopathic arthritis (JIA)-associated genes. The Disease Association Protein-Protein Link Evaluator (Dapple) algorithm was used to generate an inferred interactome network from the 348 JIA-associated genes, an iterative process was then used to generate random networks and significant changes in seed gene connectivity were calculated. The JIA-associated seed genes are shown coloured by significance of deviation of observed network connectivity from expected (red = highly significant to green = not significant) (Figure 1Bii).

Figure 4

Identification of clusters of highly connected nodes within the juvenile idiopathic arthritis (JIA) interactome. Two clusters were identified using the MCODE clustering algorithm. (A) The oligoarticular JIA cluster includes PTPN2 and STAT1; (B) the RF-ve polyarticular related cluster contains KSR1. JIA-associated genes = red hexagons, blue circles = inferred genes within the JIA interactome (Figure 1E).