Network-based analysis of differentially expressed genes in cerebrospinal fluid (CSF) and blood reveals new candidate genes for multiple sclerosis

Abstract

Background: The involvement of multiple genes and missing heritability, which are dominant in complex diseases such as multiple sclerosis (MS), entail using network biology to better elucidate their molecular basis and genetic factors. We therefore aimed to integrate interactome (protein-protein interaction (PPI)) and transcriptomes data to construct and analyze PPI networks for MS disease.

Methods: Gene expression profiles in paired cerebrospinal fluid (CSF) and peripheral blood mononuclear cells (PBMCs) samples from MS patients, sampled in relapse or remission and controls, were analyzed. Differentially expressed genes which determined only in CSF (MS vs. control) and PBMCs (relapse vs. remission) separately integrated with PPI data to construct the Query-Query PPI (QQPPI) networks. The networks were further analyzed to investigate more central genes, functional modules and complexes involved in MS progression.

Results: The networks were analyzed and high centrality genes were identified. Exploration of functional modules and complexes showed that the majority of high centrality genes incorporated in biological pathways driving MS pathogenesis. Proteasome and spliceosome were also noticeable in enriched pathways in PBMCs (relapse vs. remission) which were identified by both modularity and clique analyses. Finally, STK4, RB1, CDKN1A, CDK1, RAC1, EZH2, SDCBP genes in CSF (MS vs. control) and CDC37, MAP3K3, MYC genes in PBMCs (relapse vs. remission) were identified as potential candidate genes for MS, which were the more central genes involved in biological pathways.

Discussion: This study showed that network-based analysis could explicate the complex interplay between biological processes underlying MS. Furthermore, an experimental validation of candidate genes can lead to identification of potential therapeutic targets.

Keywords: Clique analysis; Modularity; Multiple sclerosis; Protein–protein interaction network (PPIN); Topology; Transcriptome.

Figure 1. The degree distribution of nodes followed power law distribution.

(A) Degree distribution of differentially expressed genes in CSF QQPPI network. (B) PBMCs QQPPI network. The graph represents a decreasing trend of degree distribution with an increase in the number of links showing scale-free topology.

Figure 2. QQPPI networks generation by mapping of differentially expression genes on PPI data.

(A) CSF QQPPI network. (B) PBMCs QQPPI network. Nodes with high centrality measures are shown by bigger size than others. Green and red nodes represent proteins encoded by up- and down-regulated genes, respectively. Graphical representation of nodes was implemented by “Spring Embedded” layout in Cystoscape.

Figure 3. Functional categories of the networks were visualized using the Enrichment map plugin of the Cytoscape.

Significant biological processes are represented by one node in (A) CSF QQPPI network. (B) PBMCs QQPPI network. Nodes’ sizes indicate the significance of the enrichment (p-value). Edges show gene overlap between nodes and thickness indicates the number of overlapping enriched genes.

Figure 4. Nodes with high centrality measures which involved in significant biological pathways and their expression values.

More central nodes in (A) CSF QQPPI network. (B) PBMCs QQPPI network.

Figure 5. Candidate markers involved in functional modules and complexes.

The functional enrichment of candidate markers in (A) CSF QQPPI network. (B) PBMCs QQPPI network. Modules and complexes illustrated by brown and blue dotted circles, respectively.