Assessment of complementarity of WGCNA and NERI results for identification of modules associated to schizophrenia spectrum disorders.

Arthur Sant’Anna Feltrin,David Corrêa Martins,Helena Brentani,Sérgio Nery Simões,Ana Carolina Tahira,Francisco J Esteban

doi:10.1371/journal.pone.0210431

Abstract

Psychiatric disorders involve both changes in multiple genes as well different types of variations. As such, gene co-expression networks allowed the comparison of different stages and parts of the brain contributing to an integrated view of genetic variation. Two methods based on co-expression networks presents appealing results: Weighted Gene Correlation Network Analysis (WGCNA) and Network-Medicine Relative Importance (NERI). By selecting two different gene expression databases related to schizophrenia, we evaluated the biological modules selected by both WGCNA and NERI along these databases as well combining both WGCNA and NERI results (WGCNA-NERI). Also we conducted a enrichment analysis for the identification of partial biological function of each result (as well a replication analysis). To appraise the accuracy of whether both algorithms (as well our approach, WGCNA-NERI) were pointing to genes related to schizophrenia and its complex genetic architecture we conducted the MSET analysis, based on a reference gene list of schizophrenia database (SZDB) related to DNA Methylation, Exome, GWAS as well as copy number variation mutation studies. The WGCNA results were more associated with inflammatory pathways and immune system response; NERI obtained genes related with cellular regulation, embryological pathways e cellular growth factors. Only NERI were able to provide a statistical meaningful results to the MSET analysis (for Methylation and de novo mutations data). However, combining WGCNA and NERI provided a much more larger overlap in these two categories and additionally on Transcriptome database. Our study suggests that using both methods in combination is better for establishing a group of modules and pathways related to a complex disease than using each method individually. NERI is available at: https://bitbucket.org/sergionery/neri.

Highlights

In linear systems, the performance of the whole is the superposition of the effects of each of its forming parts
The Weighted Gene Correlation Network Analysis (WGCNA) uses the values of topological overlap measure (TOM) from both groups for a module preservation analysis, where it indicates the preservation of the modules across the control and case networks
Using the overlap between SZGB’s differentially expressed genes (DEG) and WGCNA-Network-Medicine Relative Importance (NERI) hub genes (37 hub genes for BAHN and 19 hub genes for kME correlation with sample traits (KATO)), we identified genes related to neuroimmunity and neuroinflammation such as as CHI3L1 (Chitinase 3 Like 1) and CHI3L2 (Chitinase 3 Like 2), both genes related to the degeneration of motor neurons in amyotrophic lateral sclerosis (ALS) as well to neuroinflammatory processes in general [47,48,49]; IL6 (Interleukin 6), a mediator of TH2 (T helpers cells), involved as well into embryogenesis [50]; CXCL9 (C-X-C Motif Chemokine Ligand 9), protein coding gene related to T cell trafficking [51]

Summary

Introduction

The performance of the whole is the superposition of the effects of each of its forming parts. In complex systems, such as gene expression, a global view of the system is different from that obtained by considering only the sum of its parts, or by ignoring the reciprocal influence of its constituent elements. As studies in multiples species, and different tissues have shown that co-expressed genes tend to be functionally related [5, 6], co-expression gene network has been widely used [7, 8]. Psychiatric disorders involve changes in multiple genes, and different types of variations as common variation of small effects, very rare or the de novo variants intolerant to gene mutation and by uncommon, highly penetrant variants of larger effect as CNVs. All potentially converge in deregulated biological functions. Gene co-expression networks allowed the comparison of different tissues, stages [10] and parts of the brain contributing to an integrated view of genetic variation [11]

Objectives

Methods

Results

Discussion

Conclusion