Construction of Parkinson's disease marker-based weighted protein-protein interaction network for prioritization of co-expressed genes.

Background: Parkinson's disease (PD) is a complex neurodegenerative movement disorder that primarily results due to the loss of dopaminergic neurons in the substantia nigra region. Studying gene expression in the substantia nigra region would potentially unravel disease-relevant protein-protein interactions.

Methods: In this study we have used network science approach to prioritize candidate genes by focussing on differentially expressed genes (DEGs) that interact with established PD associated-genes (PDAG). Prioritizing genes that interact with already established PDAG would reduce the probability of spurious protein-protein associations. The dataset GSE54282 with Parkinson's disease affected substantia nigra samples was extracted from Gene Expression Omnibus (GEO) database. Protein-Protein Interaction Network (PPIN) was constructed by retrieving all possible interactions between DEGs from high-throughput experiments and literature data using Bisogenet. This complex PPIN was decomposed to construct a subnetwork of Parkinson's Disease-Protein Interaction Map (PD-PIM) by including PDAG and following well-established concepts of network biology such as degree and betweenness centrality. We then implemented a "two-way analysis" where we selected genes belonging to PDPIM subnetwork with their primary interacting partners and highly coexpressed genes on the basis of Pearson score.

Results: A complex PPIN comprised of 5340 nodes (genes) and 39,199 edges (interactions) was obtained. A list of 205 genes (123 PDAGs, 69 hub genes and 13 bottleneck genes) with their respective first level interacting partners were extracted from PPIN interactome to build a PD-specific subnetwork, PD-PIM. This subnetwork PD-PIM comprised of 5078 nodes and 38,357 edges. We then employed a "two-way" gene prioritization method that delineated 267 genes of which 16 genes were found to intersect in the two networks of the "two-way analysis". Of the 16 genes, we narrowed down to 7 novel candidate genes (VCAM1, BACH1, CALM3, EGR1, IKBKE, MYC and YWHAG) displaying significant changes in their network interactions between control and disease samples. Interestingly, these genes were associated with neuroinflammation signaling pathway, MAPK signaling apoptosis pathway, movement disorders and development of neurons that are linked with development of PD.

Conclusion: We propose that VCAM1, BACH1, CALM3, EGR1, IKBKE, MYC and YWHAG genes might play important roles in PD pathogenesis, as well as facilitate the development of effective targeted therapies. Our integrative and network based approach for finding therapeutic targets in genomic data could accelerate the identification of novel drug targets for Parkinson's disease.