Aberrations in the Cross-Talks Among Redox, Nuclear Factor-κB and Wnt/Catenin Pathway Signaling Underpin Myalgic Encephalomyelitis andChronic Fatigue Syndrome: A Review and New Hypothesis Based on Results of Network, Enrichment and Annotation Analyses

There is evidence that chronic fatigue spectrum disorders (CFAS-D) including Myalgic Encephalomyelitis (ME), chronic fatigue syndrome (CFS) and chronic fatigue with physiosomatic symptoms including when due to comorbid medical disease are characterized by neuroimmune and neuro-oxidative biomarkers. The present study was performed to delineate the protein-protein interaction (PPI) network of CFAS-D and to discover the pathways, molecular patterns and domains enriched in their PPI network. We performed network, enrichment and annotation analysis using differentially expressed proteins and metabolics, which we established in CFAS-D patients. PPI network analysis revealed that the backbone of the highly connective CFAS-D network comprises NFKB1, CTNNB1, ALB, peroxides, NOS2, TNF, and IL6, and that the network comprises interconnected immune-oxidative-nitrosative and Wnt/catenin subnetworks. MultiOmics enrichment analysis shows that the CFAS-D network is highly significantly associated with cellular (antioxidant) detoxification, hydrogen peroxide metabolic process, peroxidase and oxidoreductase activity, IL10 anti-inflammatory signaling, and neurodegenerative, canonical Wnt, the catenin complex, cadherin domains, cell-cell junctions and TLR2/4 pathways; and the transcription factors NF-κB and RELA. The top-10 DOID annotations of the CFAS-D network include four intestinal, three immune system disorders, cancer and infectious disease. Custom GO term annotation analysis revealed that the CFAS-D network is associated with a response to a toxic substance, lipopolysaccharides, bacterium or virus. In conclusion, CFAS-D may be triggered by a variety of stimuli and their effects are mediated by aberrations in the cross-talks between redox, NF-κB, and Wnt/catenin signaling pathways leading to dysfunctions in multicellular organismal homeostatic processes.


Contemporary research of chronic fatigue spectrum disorders (CFAS-D) including
Myalgic Encephalomyelitis (ME), chronic fatigue syndrome (ME/CFS) and chronic fatigue (CF) and associated psychosomatic symptoms is plagued by a cacophony of controversies. These different approaches, sometimes even competing, comprise folk psychology (the culprit of CFAS-D is a psychological problem) and the medical approach (the culprit is one specific virus or different viruses or bacteria). The dominant view, especially in Europe, is that of the cognitivebehavioral and the biopsychosocial schools [1]. This view entails that CFAS-D, even when due to medical disease (e.g. cancer), are the consequence of psychosocial and biological factors and negative cognitions [1-4]. The Wessely model, for example, conceptualizes that the effects of a trigger factor, which may be a virus, are mediated by boom and bust activity and bedrest and the Vercoulen model considers that CFAS-D symptoms are aggravated by causal attributions and reduced physical activity [1]. Nevertheless, it appears that the label "biopsychosocial" is more window dressing than the actual approach because, in fact, folk psychology statements abounds in their publications as for example "it is in the mind", "they think themselves ill" and "it is a disorder of perception whereby patients think those symptoms are the consequence of a virus" [review: 2].
Nevertheless, this folk psychology approach is embraced by the NHS, the Lancet and national health case system all over Europe (e.g.. UK, France, Sweden, Benelux). In addition, we also discovered that ME/CFS is characterized by increased IgA/IgM responses to Gram-negative bacteria indicating increased bacterial or lipopolysaccharide translocation, and IgM-mediated autoimmune responses to a number of neoantigens including malondialdehyde (MDA), azelaic acid and nitrosylated proteins, indicating increased nistrosylation [18,19].
All in all, these findings show that multiple differentially expressed proteins (DEPs) and metabolic pathways are involved in the pathophysiology of CFAS-D. However, no research has delineated the protein-protein interaction (PPI) and metabolic-protein interactions (MPI) networks of CFAS-D and the biological processes, molecular functions and complexes, cellular components, pathways, transcriptional regulatory relationships, protein domains and human disease annotations which are associated with the PPI and MPI networks of CFAS-D.
Hence, we have conducted network, enrichment and annotation analyses in order to delineate the hotspots in the CFAS-D network and the top functions and paths enriched in the networks. This is important because the most influential genes, metabolics, and pathways may constitute new drug targets to treat CFAS-D. Moreover, such analyses may disclose the putative trigger factors of the CFAS-D interactome and its associations with comorbid medical disorders.
As such, these enrichment and annotation analyses may help to explain the strong comorbidity of CFAS-D with immune, infectious, and neuro-psychiatric disorders and the possible shared pathophysiological core which may underpin CFAS-D.

Selection of seed proteins and metabolic markers
"This study is a secondary data analysis on existing data using open, deidentified and noncoded data sets and, therefore, this is non-human subjects research which is not subject to Institutional Review Board approval" [20]. In case-control studies, we previously have identified metabolic pathways and differentially expressed proteins (DEPs) in CFAS-D including when due to comorbid medical disease. Almost all biomarkers included in this study were extracted from our studies on ME/CFS and CF-like symptoms with a duration > 6 months in comorbid disorders including major chronic kidney disease with hemodialysis, depression, schizophrenia, and rheumatoid arthritis (Electronic Supplementary File (ESF), 1, References). One study was performed on patients with CF-like symptoms due to acute COVID-19 infection. We were able to

PPI network construction, and enrichment and annotation analyses.
The network, enrichment and annotation analysis were conducted as reviewed previously [20]. In brief, we constructed two network subtypes, the first was constructed using the abovementioned DEPs whereby the physical interactions between the DEPs were visualized using STRING version 11.0 (https://string-db.org) and Cytoscape (https://cytoscape.org). The second was constructed using OmicsNet (www.omicsnet.ca) using the abovementioned metabolites and examining the MPI based on KEGG reactions. Consequently, the genes interacting with the metabolics in the different subnetworks were used in STRING to build a giant network based on our seed genes enlarged with the MPI-derived DEPs, which was, consequently, analyzed in OmicsNet to construct a composite network consisting of metabolites and DEPs first entering MPIs and then the PPIs (analyzed with IntAct Molecular Interaction Database: https://www.ebi.ac.uk/intact/) an with a targeted incorporation of TF-protein interactions (TFPI) (analyzed using TTRUST (www.grnpedia.org/trrust).
Network features were computed using STRING and the Cytoscape plugin Network Analyzer and comprise number of nodes, number of edges and expected number of edges, average node degree, network diameter and radius, characteristic path length, and network density and heterogeneity. The top hubs (high degree) and bottlenecks (high betweenness centrality) were computed and used to delineate the backbone of the network, i.e. the top7 hubs and the top-2 nonhub bottlenecks. We used Markov Clustering (MCL) employing STRING to discover communalities of interconnected nodes, which display similar attributes and/or functions.
We examined the different networks and MCL subnetworks for their enrichment scores and annotated terms and these analyses were also performed on the downregulate seed genes and the hotspots of the enlarged network. Enrichment/annotation analyses were performed using seed nodes. In the first-order non-seed genes we found that EGFR was connected with 3 subnetwork 2 seed genes (GRN, CTNNB1, and DKK1) and with 19 seed genes in subnetwork 1.

MultiOmics Analysis including the oxidative stress-associated metabolites.
In order to construct a second giant network including proteins interconnecting with the metabolites, we entered the latter in OmicsNet analysis and examined the MPIs (using KEGG and IntAct and using only the first order MPIs). We found 5 subnetworks: one centered around hydroperoxides (46 nodes), another around EPA (6 nodes), 6 around 3OHK, 4 around NO, and 3 around DHEA (albeit some overlapping). The 59 nodes coupled with the nodes from the first network (see Figure 2) were consequently examined using STRING and Network Analyzer.  Figure 2 shows the results of MCL cluster analysis (inflation parameter of 1.7) displaying two significant protein subnetworks, a first comprising immune and nitro-oxidative stress genes (see Figure 6; this cluster is now re-named the immune-inflammatory, oxidative and nitrosative cluster, IO&NS), and a second Wnt/catenin cluster. There were also some communities with only few nodes, for example one centered around KYNA (kynureninase) and KMO (kynurenine 3monooxygenase).    and molecular, REACTOME and PANTHER performed on the IO&NS/Wnt genes. We found five significant molecular complexes, the first represents cytokine/IL10 signaling and a response to LPS; the second comprises cellular oxidant detoxification and response to a toxic substance; the third amine oxidase reactions, hydrogen peroxide metabolic process and degradation of beta catenin; the fourth peroxisomal protein import and a carboxylic acid catabolic response; and the fifth was the same as MCODE2 in Table 2.   Figure 5 shows a heatmap (made using Enrichr and Appyter) with the top KEGG pathways that were overrepresented in the IO&NS/Wnt network, namely pathways of neurodegeneration, Wnt and NF-κB signaling and peroxisome. Figure 6 shows a heatmap with the PANTHER 2016 pathways which were over-represented in the network, namely Wnt, apoptosis, TLR and cadherin pathways and the Alzheimer disease presenilin pathway. TTRUST analysis showed that NF-κB (pFDR=6.58E-23) and RELA (pFDR=8.691E-21) were the most important transcriptional factors of the network followed at a large distance by SP1 (pFDR=7.079E-13). annotation analysis performed on the DEPs (or selected DEPs) presented in Figure 1.

The networks and subnetworks of CFAS-D
The first major finding of this study is that the PPI network of the DEPs and metabolics of CFAS-D show high connectivity and comprises two subnetworks, a first centered around IO&NS genes and a second around Wnt/catenin genes. The backbone of the master network comprises DEPs/metabolics including NFKB1, CTNNB1, ALB, peroxides, NOS2, TNF, and IL6, while in the giant network many redox-related enzymes were predicted to be hubs, including NOS3 (endothelial NO synthase producing NO relaxing smooth muscle relaxation), SOD2 (mitochondrial superoxide dismutase 2) and PRDX6 (peroxiredoxin 6; catalyzes the reduction of hydrogen peroxides). It is important to note that without the delineation of the MPIs, one would have concluded that especially pro-inflammatory cytokine genes are the dominant forces in this network, whereas, in fact, the IO&NS subnetwork is more dominated by redox genes, NF-κB and IL-10.
CTNNB1, which belongs to the Wnt subnetwork was another hotspot and additionally showed many interactions with genes in both subnetworks indicating that this gene is a relevant switch linking both subnetworks. Other relevant switches belonging to the immune subnetwork were ALB and IL6 which showed many interactions with cluster 1 genes, but also with CTNNB1, DKK1 and AGRN. Hotspots and switches are considered to be new drugs targets because they govern and control the network and/or link the subnetworks [20]. All in all, we may conclude that dysfunctions in the IO&NS and Wnt subnetworks underpin the pathophysiology of CFAS-D.

Terms over-represented in the IO&NS MultiOmics subnetwork
The second major finding of this study is that the top relevant functions and pathways in the first network revolve primarily around redox mechanisms, namely cellular oxidant detoxification (and thus cellular detoxification and a cellular response to toxic substance), the hydrogen peroxide metabolic process, antioxidant activity, amine oxidase reactions, oxidoreductase activity, peroxidase activity, and haem and glutathione peroxidase. These findings indicate that increased oxidative stress (hydrogen peroxides) probably as a response to a toxic The results of our network and enrichment analyses show that NFKB1 was not only one of the most important hotspots in the CFAS-D network, but also that NF-κB (p50-p52 unit) and RELA (NF-κB p65 unit or transcription factor p65) were the most important transcription factors controlling the network and that the NF-κB signaling pathway was one of the most important paths enriched in the network. NF-κB is a major transcriptional factor involved in the response to a vast array of stimuli including bacterial or viral antigens, cytokines, free radicals, oxidized epitopes, and glutamate, and is a transcriptional inducer of many genes including cellular adhesion molecules, pro-inflammatory cytokines, chemokines, and growth, apoptosis and coagulation factors, and antioxidants and pro-oxidants [27][28][29]. RELA is involved in the activation of NF-κB and stimulates NF-κB translocation to the nucleoplasm and by forming a RELA-NF-κB complex activates target gene expression [30]. NF-κB p50, which is associated with RELA, and NF-κB p52 are major components of the canonical and noncanonical NF-κB signaling pathways which both lead to target gene activation [29]. Both transcription factors are known to regulate different pathways that we observed to participate in the network including the MAPK pathway [31] and ROS and Wnt/catenin pathways.
There are many intersections between reactive oxygen species (ROS) and NF-κB signaling.
Indeed, ROS may modulate the NF-κB response leading to transcriptional activation of NF-κBtarget antioxidant genes (e.g. SOD, HO1, GPX1, TRX1 and TRX2), which reduce ROS production, thereby promoting cell survival [29], and NF-κB-target pro-ROS genes including NOX2, COX2, and NOS2 [29,32]. It should be stressed that the effects of ROS on NF-κB are more than complex with stimulatory effects in the cytoplasm and inhibitory effects in the nucleus [33], while ROS may also oxidize NF-κB p50 leading to decreased DNA binding capacity [34].
All in all, the results of our network and enrichment analysis and Maes et al. [35,36] indicate that the complex cross-talks between NF-κB and ROS signaling are involved in the pathophysiology of CFAS-D. Moreover, recent studies show that increased NF-κB expression may be associated with central fatigue by modulating central nervous system genes and regulating immune-inflammatory processes, synaptic plasticity, and memory and exerting neurotoxic effects Our MCODE analysis revealed that the IL-10 anti-inflammatory signaling pathway, regulation of cytokine production, signaling by interleukins and a response to LPS was a relevant molecular complex in the IO&NS subnetwork of CFAS-D. Finally, our network and enrichment analyses showed that the TRYCAT pathway may be involved in CFAS-D, although it is not a key component but rather a spin-off of the IO&NS response. Previously, it was reported that this pathway is highly strongly associated with somatization disorder, a psychiatric disease accompanied by physiosomatic symptoms but not necessarily by fatigue [46]. The TRYCAT pathway acts as a redox-regulator and is one of the major antioxidant systems, although some TRYCATs have pro-oxidant and neurotoxic activities [47].
Thus 3OHK, one of the metabolics in our MPI network, is one of the neurotoxic TRYCATS produced during activation of this pathway as a consequence of IO&NS activation. KYNU (kynurenine hydroxylase) is as IDO an oxygen-consuming enzyme which catabolizes kynurenine into 3OHK (STRING).

Terms over-represented in the Wnt MultiOmics subnetwork
Our enrichment analyses revealed that the CFAS-D network was highly significantly associated with two major interrelated functions/pathways/domains namely the canonical Wnt/βcatenin pathway, T-cell factor (TCF) dependent signaling in response to Wnt, degradation of βcatenin by the destruction complex, the catenin complex and DIX domain, and adherens junctions, cadherin prodomain, and cell-cell junctions. The DIX domains and axin, GSK-3 and Dishevelled (Dvl) are key players in the β-catenin destruction complex thereby determining the interaction of β-catenin with the transcription factor TCF and, thus, the expression of the Wnt target genes [48]. . H2O2 not only inhibits Wnt/b-catenin signaling but may also increase Wnt/catenin signaling [58]. In fact, peroxides have a biphasic effect on Wnt signaling with an increase 20 minutes after activation and reduced signaling some hours later [57]. Finally, β-catenin is a key regulator of the homeostatic cell response, which helps to repair the damage due to nitro-oxidative stress [59].
Furthermore, the NF-κB and Wnt pathways shows multiple cross-talks and negatively or positively regulate each other, thereby forming a mutual regulatory network [60]. Wnt modulates the production of inflammatory cytokines and NF-κK signaling and bridges innate and adaptive immune pathways [61]. On the other hand, increased levels of IL-6 and TNF-α may maintain The Wnt/catenin pathway is also involved in a) pain and neuropathic pain with Wnt inhibition improving pain [67,70]; b) skeletal muscle dynamics, the neuromuscular synapse, and musculoskeletal functions including the electrophysiologic properties of muscle cells [71,72]; c) intestinal functions including epithelial homeostasis and integrity, the physiological proliferation of the transit-amplifying cells and differentiation of Paneth, goblet, and enteroendocrine cells, and the maintenance of mucosa and barrier functions [73][74][75][76]. The Wnt/catenin pathway also plays a key role in autoimmunity as observed in rheumatoid arthritis [77,78] and in the response to bacterial infections and inflammation [79,80]. Figure 8 summarizes the pathways and molecular complexes or functions that accompany CFAS-D. Aberration in the cross-talks among redox, NF-κB, and Wnt signaling may be key pathways, which are associated with dysfunctions in multicellular organismal homeostatic processes including in cell-junction organization. Disorders in the intertwined interactions between these systems may explain the broad spectrum of organs and dysfunctions that participate in CFAS-D (brain, musculoskeletal system, immune system, gastro-intestinal system). An important spin-off is increased IL10 production, which may contribute to immunosuppression and recurrent or protracted infections, and also increased TRYCAT production may aggravate the neurotoxic effects of oxidative stress. Increased translocation of Gram-negative bacteria with increased LPS load is probably a major trigger factor but also other bacterial infections, toxoplasmosis, viral infections (e.g. cytomegalovirus), cancer, and gastro-intestinal, autoimmune, immuneinflammatory, neuroinflammatory and neurodegenerative disorders appear to be associated with those pathways via activation of TLR/LTF/TWEAK signaling.

Conclusions
Future research should scrutinize the specific role of the NF-κB, ROS and Wnt axis in CFAS-D. The cross-talks between these three pathways may also constitute new drug targets to treat CFAS-D. Given that the Wnt pathway shows many complex, negative and positive feedback loops interacting with redox systems and NK-κB signaling, manipulations of Wnt signaling and β-catenin (despite being a hub and master switch) appears to be very challenging. It may be more promising to simultaneously target the crosstalk among redox and NF-κB pathways. New knowledge on the precise aberrations in the Wnt pathway in association with ROS and NF-κB signaling may lead to even more effective treatments by specifically targeting β-catenin, the βcatenin destruction complex or the TCF transcription factor.
Our new model shows that disorders in cross-talks among these three key pathways mediate the effects of a variety of trigger factors in the onset of CFAS-D. These findings also support the theory that once CFAS-D is present the abovementioned pathways may increase morbidity and even mortality of IO&NS-associated medical disorders through the detrimental effects of disorders in redox, NF-κB and Wnt axis [81,82]. The model also explains that the acute phase of inflammatory conditions may lead to CFAS-D via disorders in this axis, oxidative damage and the neurotoxic effects of LPS, pro-inflammatory cytokines and TRYCATs. Since most biomarkers included in this study were extracted from studies on ME/CFS and CF-like symptoms in comorbid disorders with duration > 6 months, we may conclude that after resolution of acute inflammation, CAFS-D symptoms are maintained by continued aberrations in the redox, NF-κB, and Wnt axis, increased IL-10 production and increasing oxidative damage including secondary autoimmune responses and nitrosylation.

Funding statement
There was no specific support for this specific study.

Conflict of interest
The authors have no conflict of interest with any commercial or other association in connection with the submitted article.
All the contributing authors have participated in the manuscript. MM designed the study and performed the network, enrichment and annotation analyses. All authors contributed to interpretation of the data and writing of the manuscript.
Compliance with ethical standards.
The study was conducted according to international ethics and privacy laws.

IRB statement.
This study is a secondary data analysis on existing data using open, deidentified and non-coded data sets and, therefore, this is non-human subjects research, which is not subject to IRB approval.
References.    ) was centered around NFKB1, TNF, IL6, IL10, IL1, TLR4, etc., and 2) a second Wnt/catenin subnetwork (green nodes) centered around CTNNB1 Figure 2. First order protein network of chronic fatigue spectrum disorders. MCL cluster analysis found two major subnetworks: 1) a first immune, oxidative and nitrosative subnetwork (IO&NS; red color) was centered around NOS2, NFKB1, IL10, etc, and 2) a second Wnt/catenin subnetwork (yellow nodes) was centered around CTNNB1. The tryptophan catabolite pathway (KYNU and KMO) and opioid (OPRK1 and OPRM1) genes appear to be spin-offs of the IO&NS subnetworks   ESF 1, Table 1 displays the results of MCODE analysis employing KEGG, WikiPaths, GO biological and molecular, REACTOME and PANTHER performed on the first-order network built using all seed genes. We observed three significant molecular complexes: the first represents IL10 signaling and signaling by interleukins; the second reflects canonical Wnt signaling and cell-cell signaling by Wnt; and the third represents TCF-dependent signaling in response to Wnt and Wnt signaling.
ESF 2, Figure 1 shows the enriched ontology term clusters in the first order network build using all genes (the network is constructed and visualized using MetaScape and Cytoscape, v3.1.2). This figure shows that the immune and Wnt subnetworks are strongly intertwined as well as the MAPK cascade and that these pathways are interconnected with multicellular organismal homeostasis.
ESF 2, Figure 2 shows the top-20 terms which were over-represented in the first order network built using all genes. This bar graph shows that besides cytokine (especially IL10) and Wnt signaling also positive regulation of the MAPK cascade is a significant path.
ESF 1, Table 2  ESF 2, Figure 3 shows the enriched ontology term clusters in the immune subnetwork of CFAS-D and that cytokine (in particular IL10, but also IL4 and IL13) signaling, a response to LPS or an external stimulus were strongly interacting pathways.
ESF 2, Figure 4 shows the top WikiPaths which were over-represented in the immune subnetwork, namely the LTF danger signal response pathway and the TLR pathway. ESF 2, Figure 5 shows a bar graph with the top-10 BioCarta terms which are overrepresented in the immune subnetwork, namely NFKB pathway, anti-inflammatory IL10, IL1R, TNFR1 and ceramide signaling. ESF 2, Figure 6 shows the top-10 KEGG annotations including Chagas disease, tuberculosis, CMV infection, toxoplasmosis, and viral protein interactions with cytokine and cytokine receptor.

Enrichment analysis on the Wnt subnetwork genes of CFAS-D.
ESF 2, Figure 7 and 8 shows the enriched ontology term clusters in the Wnt subnetwork of CFAS-D. Besides Wnt/cateninassociated paths, these also include diseases of signal transduction, signaling by Wnt in cancer, PID PS1 pathway and cell-junction organization. ESF 2, Figure 9 shows the WikiPathways which were statistically over-represented in the Wnt/catenin subnetwork, including involvement of this pathway in colorectal cancer, leukemia, endometrial and breast cancer, ect. The bar graph shown in ESF 2, Figure 10 shows the top-10 InterPro domains which were enriched in the Wnt subnetwork, including the cadherin prodomain and cytoplasmic domain, and the DIX domain. ESF 2, Figure 11 shows a bar graph with top-10 enriched GO cellular components, including the catenin complex and adherens and cell-cell junctions. TTRUST enrichment analysis showed that NFKB1 (pFDR=2.911E-25) and RELA (pFDR=1.014E-22) were the two most important transcriptional factors in this network.

Annotation analyses and functional categorization of the PPI network and selected genes
ESF 2, Figures 12 and 13 depict the hierarchical structure of GO terms and the hubs and master regulatory transcription factor (ESF 2, Figure 12), downregulated seed genes (ESF 2, Figure 13A), seed genes of the Wnt/catenin pathway subnetwork (ESF 2, Figure   13B), and the major hotspots in the STRING enlarged networks (ESF 2, Figure 13C). Thus, ESF 2, Figure 12 shows that the GO functions which are regulated by NFKB1, CTNNB1, TNF and IL6. ESF 2, Figure 13A shows that the downregulated genes are associated with cellular oxidant detoxification and detoxification in general. ESF 2, Figure 13B shows that the seed genes of the Wnt subnetwork are associated with synapse organization and cell-cell signaling, whereas the major hotspots of the enlarged network (ESF 2, Figure 13C) are associated with a variety of processes including smooth muscle cell proliferation, regulation of DNA metabolic and apoptotic processes, and response to lipids, steroid hormones. ESF 2, Figure 1. Enriched ontology term clusters in the chronic fatigue spectrum disorders network. Terms are represented by a circle node, with the colors representing cluster identity and their size reflecting the number of input genes. One term is used to describe the clusters (colored labels). The thickness of the (bundled) edges linking the terms reflects the similarity score (threshold is > 0,3). The network is constructed and visualized using MetaScape and Cytoscape (v3.1.2) ESF 2, Figure 3. Enriched ontology term clusters in the immune subnetwork of chronic fatigue spectrum disorders. Terms are represented by a circle node, with the colors representing cluster identity and their size reflecting the number of input genes. One term is used to describe the clusters (colored labels). The thickness of the (bundled) edges linking the terms reflects the similarity score (threshold is > 0.3). The network is constructed and visualized using MetaScape and Cytoscape (v3.1.2) ESF 2, Figure 7. Enriched ontology term clusters in the Wnt/catenin subnetwork of chronic fatigue spectrum disorders. Terms are represented by a circle node, with the colors representing cluster identity and their size reflecting the number of input genes. One term is used to describe the clusters (colored labels). The thickness of the (bundled) edges linking the terms reflects the similarity score (threshold is > 0.3). The network is constructed and visualized using MetaScape and Cytoscape (v3.1.2).
ESF 2, Figure 12. Results of GOnet annotation visualization in chronic fatigue spectrum disorders depicting the hierarchical structure of GO terms and the hubs and master regulatory transcription factor ESF 2, Figure 13. Results of GOnet annotation visualization in chronic fatigue spectrum disorders depicting the hierarchical structure of GO terms and A: downregulated seed genes; B: seed genes of the Wnt/catenin pathways; and C: the hotspots in STRING enlarged networks ESF 2, Figure 14. An extended network constructed with inBio Discover showing the top Disease Ontology (DOID) annotations of chronic fatigue spectrum disorders