Hub Gene Clusters Reveal Dysregulated Synaptic Neurotransmitter Signaling Pathways and Drug Repurposing Prospect in Brain Tumors

Simple Summary

The study analyzed transcriptomic data of different brain tumor types and revealed common hub genes. Hub genes revealed dysregulated synaptic neurotransmitter signaling pathways. The hub genes were used to assess drug–gene interactions for drug repurposing screening. Among the candidate drugs, only those capable of altering hub gene regulation were screened for blood–brain barrier permeability. It revealed that Valproic acid and Gabapentin are promising drugs against brain tumors.

Abstract

Background/Objectives: Brain tumors, particularly gliomas, have high mortality and are limited in treatment options, often complicated by severe conditions, which can be fatal. Given the increasing incidence and adverse effects of current drugs, an in silico drug repurposing approach using hub gene clusters to streamline and accelerate the search for new therapies. Methods: The GSE66354, GSE68848, GSE74195, and GSE43290 datasets were used to identify DEGs using GEO2R. A gene co-expression network was constructed using the STRING PPI database. Preserved clusters revealed hub genes, which were used for GO and KEGG pathway enrichment analyses. Drug repurposing screening was performed through drug–gene interactions in DGIdb. Suggestive drugs were then validated through GSEA-CMAP and BOILED-Egg. Results: The study identified three key gene clusters that serve a role in synaptic transmission and transmembrane transport, synaptic vesicle neurotransmission, and extracellular matrix formation. Five drugs passed the drug screening, which are Gabapentin, Pyrantel, Resveratrol, Trifluoperazine, and Valproic acid. Conclusions: Valproic acid and Gabapentin are highly suggestive as candidate repurposed drugs. This study enhances our understanding of brain tumor genetics and supports the development of new immunotherapeutic strategies.

1. Introduction

Gliomas are the most common primary brain tumors, originating from glial cells. They are further classified based on the cell type from which they arise: astrocytoma originates from astrocytes, oligodendroglioma from oligodendrocytes, and ependymoma from ependyma [1,2,3,4]. Gliomas diffusely infiltrate surrounding brain tissue, affecting brain function, and most commonly occur in adults [5]. Pilocytic astrocytoma is the least concerning benign glioma, while glioblastoma is the most malignant type [5]. In contrast, embryonal tumors are primary brain tumors that originate from embryonal cells that remain from fetal development and occur predominantly in children, affecting their neurocognitive function [6]. This includes medulloblastoma and atypical teratoid/rhabdoid tumor (ATRT), which occurs in the cerebellum, and pineoblastoma, which originates from the pineal gland. Moreover, primitive neuroectodermal tumors (PNETs) were classified as embryonal tumors; however, according to the 2007 WHO CNS classification, the term PNET was removed, and the tumors were reclassified as embryonal tumors [7]. The fifth edition of the World Health Organization Classification of Tumors of the Central Nervous System (WHO CNS5) recently introduced twenty-two new tumor types and subtypes [8]. Brain tumor types can be graded as WHO Grade I, II, III, and IV depending on their clinicopathology, histology, and molecular features [8].

Current brain tumor treatment includes surgery, radiotherapy, and temozolomide chemotherapy [9,10,11,12,13]. Temozolomide (TMZ) is an alkylating agent that induces DNA methylation, ultimately leading to senescence, autophagy, and apoptosis in tumor cells. It has been the lead option for chemotherapy of brain tumors [9]. TMZ, coupled with other clinical therapies for brain tumors, such as radiotherapy, is the primary treatment option for high-grade gliomas. Still, challenges with the use of TMZ were observed, such as a high toxicity limit, subjecting the patient to more harm during long-term use [9]. Existing brain tumor drugs have several adverse side effects and limit their therapeutic chances due to difficulty in passing the blood–brain barrier (BBB). The oral bioavailability of drug compounds presents limited challenges in drug discovery and development for brain tumors to pass the BBB [14]. In addition, the pressure to search for interesting new drugs is rapidly increasing, attributable to the continuous increase in global cases and harmful side effects of current brain tumor drugs. To reduce requisite resources and accelerate time on drug development and search for new brain tumor drugs, drug repurposing provides a suitable opportunity to identify probable and readily available drugs for brain tumors.

Drug repurposing utilizes known drugs for new therapeutic purposes, speeding up the drug validation and approval process [15]. Therefore, drug repurposing is a promising approach in drug search, providing readily available drug options for brain tumors [16]. Drug repurposing may provide a rapid, cost-efficient approach with a lower risk of side effects, as this technique focuses on existing drugs and primarily examines mechanisms similar to those of conventional drugs used in brain tumor chemotherapy [17,18]. Lately, computational approaches to drug–gene interactions have provided avenues for drug repurposing [19,20]. Identifying deregulated genes in brain tumors may provide targets for drug repurposing. Differentially expressed genes (DEGs) help identify tumor-promoting or tumor-suppressing genes, which is important for reconstructing gene co-expression networks to reveal hub genes that may serve as drug target genes [21,22,23]. Co-expression networks are bridges among biological molecules, represented as nodes linked by edges, forming a web of interactions [24,25,26]. Co-expression network analysis provides insights into interactions among biological molecules and the underlying molecular mechanisms, based on conserved clusters or hub genes. In the case of brain tumors, it provides insight into the molecular signaling pathways that lead to tumor development. Hub genes are key contributors to tumor development and may serve as a suitable target for drug repurposing [19,20]. These provide avenues to discover new therapeutic targets and opportunities for drug repurposing via drug–gene interactions [27,28]. Computational studies, such as gene co-expression network analysis, reveal clusters of genes and their relationships to biological pathways. Tumor gene drivers play an important role in improving cancer therapeutics; they serve as biomarkers for further classifying tumors, as prognostic markers, and as specific target genes for drug discovery and other cancer therapeutic options.

In this study, differential gene expression analysis of the GSE66354, GSE68848, GSE74195, and GSE43290 datasets, which contain gliomas, embryonal tumors, and meningiomas, is necessary for co-expression network analysis. The gene co-expression network revealed preserved gene clusters and hub genes. Hub gene clusters were associated with dysregulated synaptic neurotransmitter signaling pathways. Drug–gene interactions revealed candidate repurposed drugs. Our findings revealed a list of repurposed drugs that may serve as general treatment options for brain tumors regardless of the type of brain tumor. Our study may serve as a suggestive strategy for the future of drug validation and development against brain tumors.

2. Materials and Methods

2.1. Acquiring and Preprocessing Microarray Datasets

2.1.1. Microarray Expression Profiles Acquisition

Microarray datasets used in the study were GSE66354 [29], GSE68848 [30,31], GSE74195 [32], and GSE43290 [33], which were DNA microarray expression profiles containing different brain tumor samples retrieved from an international public repository of expression profiles, the National Center for Biotechnology Information-Gene Expression Omnibus (NCBI-GEO) [https://www.ncbi.nlm.nih.gov/geo/, accessed on 3 March 2025] [34,35,36]. These datasets were chosen based on (1) the DNA source, which was surgically removed brain tumors, (2) the dataset contained control samples for differential gene expression analysis, and (3) the platform similarity; datasets used the Affymetrix platform HG-U133A and HG-U133 Plus 2.0, where probe IDs are comparable [37,38,39,40].

Table 1. Summary of central nervous system tumor datasets used and analyzed in this study.

NCBI GEO Accession	GSE66354 a	GSE68848 b	GSE74195 c	GSE43290 d
Published	27 February 2015	14 May 2015	21 October 2015	5 January 2013
Type	Expression profiling by array
Conditions	2 ACM 17 ATRT 29 EPN-PFA 26 EPN-PFB 9 EPN-ST 19 GBM 4 MED-G3 7 MED-G4 8 MED-SHH 15 PA 13 normal brain tissue	148 ACM 1 228 GBM 2 67 ODG 67 unknowns 11 mixed 1 unclassified 30 tumor cell lines 28 nontumor brain tissue	13 EPN 1 EPN-BM 5 PNET 27 MED 5 normal cerebellum tissue	47 MEN 4 normal meninges
Platform	GPL570 (HG-U133_Plus_2) Affymetrix Human Genome U133 Plus 2.0 Array			GPL96 (HG-U133A) Affymetrix Human Genome U133A Array
RNA Source	Surgically removed brain tumors and normal brain tissue.
No. of Samples	149	580	51	51

^a High-grade Astrocytoma (ACM) ¹, Atypical Teratoid/Rhabdoid Tumor (ATRT), Posterior Fossa Group A Ependymoma (EPN-PFA), Posterior Fossa Group B Ependymoma (EPN-PFB), Supratentorial Ependymoma (EPN-ST), Glioblastoma (GBM) ², Group 3 Medulloblastoma (MED-G3), Group 4 Medulloblastoma (MED-G4), Sonic Hedgehog Group Medulloblastoma (MED-SHH), and Pilocytic Astrocytoma (PA). ^b Oligodendrogliomas (ODG). ^c Ependymomas (EPN), Ependymoblastoma (EPN-BM), Primitive Neuroectodermal Tumor (PNET), and Medulloblastoma (MED). ^d Meningioma (MEN).

displays the summary of the dataset used in this study.

2.1.2. Differential Gene Expression Analysis

The GEO2R [https://www.ncbi.nlm.nih.gov/geo/geo2r/, accessed on 6 March 2025] was utilized for differential gene expression analysis of each dataset [34,35,36]. Samples were divided into two groups: the brain tumor group, which includes all brain tumor samples, and the normal group, which includes all normal or nontumor brain tissue samples. Benjamini & Hochberg’s false discovery rate was applied to adjust p-values to reduce false positives [41]. The limma precision weights (vooma) were applied for the mean-variance relationship [42,43]. Force normalization was applied to achieve identical value distributions, and the samples were log-transformed. The adjusted p-value cutoff was 0.05, and the log₂fold threshold (FC) was ≥1 for upregulated genes and ≤−1 for downregulated genes [41,44]. DEGs were filtered for duplicates, and upregulated and downregulated DEGs were intersected separately in this study.

2.2. Gene Co-Expression Network Analysis

2.2.1. Protein-Protein Interaction Network

Shared DEGs (upregulated and downregulated) were submitted to Search Tool for Recurring Instances of Neighboring Genes (STRING) [https://string-db.org/, accessed on 9 March 2025] and produced a Protein–Protein Interaction (PPI) network [24]. A medium confidence interaction score of 0.400 was set for the PPI network. The PPI network was then reconstructed using Cytoscape [45]. Disconnected nodes were disregarded, and the gene co-expression network was finally constructed. A fold-change bar plot of co-expressed genes was generated using the average log2FC from each dataset obtained via GEO2R, revealing the most upregulated and downregulated genes.

2.2.2. Preserved Clusters and Hub Genes

Cytoscape software version 3.10.1 was used to reconstruct the STRING PPI network [45,46], which revealed the gene co-expression network. The Molecular Complex Detection (MCODE) algorithm in the ClusterViz APP was used to identify preserved clusters within the gene co-expression network [21,47]. MCODE parameters were set as follows: Degree Threshold = 2, Node Score Threshold = 0.2, K-Core Threshold = 2, and Max Depth = 100. Then, the Maximal Clique Centrality (MCC) algorithm in the cytoHubba plugin was used to rank hub differentially expressed genes (HDEGs) within each preserved cluster [48].

2.3. Pathway Enrichment Analysis

The Database of Annotation, Visualization, and Integrated Discovery (DAVID) Knowledgebase v2023q2 [https://davidbioinformatics.nih.gov/home.jsp, accessed on 16 March 2025] was utilized for the pathway enrichment analysis for Gene Ontology (GO) terms through Biological Processes (BP), Cellular Components (CC), and Molecular Functions (MF), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of each preserved cluster [49,50,51]. In the DAVID pathway enrichment analysis, p-values ≤ 0.05 were considered, and Homo sapiens was specified.

2.4. Screening of Candidate Repurposed Drugs

2.4.1. Drug–Gene Interactions

Drug Gene Interaction Database (DGIdb) beta version 4.0 [https://beta.dgidb.org/, accessed on 23 March 2025] was used to perform drug–gene interaction searches for each cluster [52]. Classified drugs and their corresponding targets were reconstructed in Cytoscape to generate drug–gene networks.

2.4.2. Drug Validation by Gene Set Enrichment Analysis and Connectivity Mapping

Candidate drugs were counter-validated through Gene Set Enrichment Analysis (GSEA) and Connectivity Map (CMAP) analysis through the Enrichr database [https://maayanlab.cloud/Enrichr/, accessed on 26 March 2025] [53,54]. Drugs with p-values ≤ 0.05 were considered, and their gene regulatory effects were predicted using the CMAP-up and CMAP-down datasets [54]. Drugs that alter gene regulation, which switch upregulated genes to downregulated and downregulated genes to upregulated, were considered candidate repurposed drugs.

2.4.3. Brain or IntestinaL EstimateD (BOILED) Permeation Predictive Model of Drugs

Pharmacokinetic properties were assessed using the Brain Or IntestinaL EstimateD permeation predictive model (BOILED-Egg) in SwissADME [http://www.swissadme.ch/, accessed on 30 March 2025]. To determine the bioavailability of drugs on the Blood–brain barrier (BBB) and Human Intestinal Absorption (HIA) based on the BOILED-Egg plot [55,56]. The drugs with high bioavailability and that permeate the BBB were proposed drugs for repurposing.

3. Results

3.1. Differentially Expressed Genes

In each of the four datasets, two groups were compared: the tumor samples and the normal brain tissue samples. The volcano plot (see Figure 1) displays the DEGs: blue dots represent downregulated genes, red dots represent upregulated genes, and black dots represent genes that fail to meet the significance cutoff. This analysis represents initial information to uncover shared tumor genes across brain tumors.

In each dataset, the DEGs after filtering are as follows (upregulated/downregulated): GSE66354: 2666/2434; GSE68848: 1836/2109; GSE74195: 903/982; and GSE43290: 598/889. The DEGs from the four datasets were intersected and revealed seventeen (17) upregulated and seventy-three (73) downregulated shared DEGs (see Figure 2).

3.2. Gene Co-Expression Network

The ninety (90) shared DEGs were used for the STRING PPI network. After removing disconnected nodes, fifty-seven (57) genes formed the gene co-expression network, mainly composed of downregulated genes, 84.21% (48/57), and 15.79% (9/57) upregulated genes (see Figure 3). SNAP25 was the hub gene of the network with twenty-six (26) edges, the highest degree of interaction or edges in the network, followed by MAPT and GRIA2, both of which have an edge score of 21, then SYT1 with 20. Genes with a single degree of interaction were TRPM3, SYT13, MAP7, EHD3, NSG1, and RUNDC3A for downregulated, and ZFP36L2 and RUVBL1 for upregulated.

Three preserved clusters in the gene co-expression network were identified. Two clusters comprised downregulated genes (clusters 1 and 2), while cluster 3 comprised upregulated genes. As shown in Figure 4, cluster 1, with eight preserved genes, GRIA2 was the hub gene, interacting with all its cluster genes and having an MCC score of 30, followed by STXBP1 (24), CACNA1A and DNM1 (18), GABRA1 (15), and the lowest were MAPT, SLC12A5, and SNAP91 (6). Cluster 2, the largest cluster with 15 preserved genes, has SNAP25 as the hub gene, with 10 interactions and an MCC score of 48. The cluster 2 hub ranking was as follows SNAP25 (48), SNCA (40), SLC17A7 and SH3GL2 (30), SYT1 (29), KIF5C (16), NEFH (12), DYNC1I1 (10), TUBB2A (7), TUBB4A (6), DNAJC6 and STMN2 (4), and the least were DNM3, CAMK2B, and MOBP (2). In cluster 3, all genes (FN1, COL6A2, FBN2, and COL1A1) were considered hub genes because they had similar interaction scores of 6. The three preserved clusters revealed 27 cluster genes, 23 of which were downregulated and 4 upregulated. To determine the most upregulated and downregulated HDEGs, a fold-change bar plot was generated (see Figure 5). Among the four upregulated HDEGs in cluster 3, COL6A2 is the most upregulated, while SLC12A5 is the most downregulated in cluster 1. Overall, the downregulated HDEGs are followed by SYT1, the most downregulated in cluster 2.

3.3. Cluster Genes Pathway Enrichment

In cluster 1 (Figure 6a), GO BP, chemical synaptic transmission was the most significant and had the highest gene counts (CACNA1A, GRIA2, GABRA1, and SLC12A5). In GO, CC showed that the neuronal cell body was the most significant. Protein kinase binding was the most significant term, and the top genes (SLC12A5, STXBP1, MAPT, SNAP91, and DNM1) were in GO MF. In the KEGG pathway, nicotine addiction (GABRA1, GRIA2, and CACNA1A) was highly significant. To sum up cluster 1, most GO BP terms were involved in synaptic transmission and transmembrane transport; GO MF was enriched in synapse and neuron-to-membrane transmission; GO CC was involved in the binding of important molecules for neuronal signaling; and the KEGG pathway revealed mental-related disorders.

For cluster 2 (Figure 6b), the most significant term was GO BP negative regulation of microtubule polymerization. In GO CC, the perinuclear region of cytoplasm was most significant. Most of the GO MF have a poor −log₁₀(p-value) score, but microtubule binding was the most significant. The synaptic vesicle cycle was highly enriched and significant in KEGG pathways, while pathways of neurodegeneration-multiple diseases had the most gene count (CAMK2B, TUBB2A, KIF5C, NEFH, TUBB4A, and SNCA). In summary of the cluster 2 enrichment analysis, GO BP mainly focused on synaptic vesicle and microtubule processes; neuron vesicles and membranes were revealed in GO CC; and GO MF, protein binding for neurotransmitters, was observed. In KEGG pathways, a pathological process of neurodegeneration was determined, such as Alzheimer’s and Parkinson’s disease.

Cluster 3 enrichment analysis (Figure 6c), the bone trabecula formation (GO BP) and ECM−receptor interaction (KEGG) were also the most significant terms. At the same time, the extracellular matrix and extracellular matrix structural constituents were significant for GO CC and GO MF. In summary, for cluster 3, GO BP was mostly related to regeneration; GO CC showed that collagen and extracellular support were mainly involved; and GO MF displayed local tissue growth. KEGG pathways revealed that Cluster 3 was associated with cancer pathways and was involved in diabetes, amoebiasis, and human papillomavirus infection.

3.4. Drug-Repurposing Candidates

Figure 7 shows the approved drugs and the cluster gene network. Among the candidate repurposed drugs, Curcumin has the highest number of downregulated genes (MAPT, TUBB2A, and TUBB4A), and Ocriplasmin has the highest number of upregulated genes (COL1A1, COL6A2, and FN1); both were the leading drugs in the three (3) gene interactions. Among downregulated genes, GABRA1 had the most drug interactions, while among upregulated genes, COL1A1 had the most. Eleven target genes were found to have no drug match: DNAJC6, DNM1, DNM3, DYNC1I1, FBN2, KIF5C, NEFH, SH3GL2, SLC17A7, STMN2, and STXBP1. Genes with specific drug interactions were MOBP (Creatine), SYT1 (Cocaine), and SLC12A5 (Bumetanide). Drugs with single-gene interactions have the potential to be explored for specific gene-targeting treatment. Target genes with several drug matches may serve as primary or general targets in discovering new drug compounds for brain tumors. Approved drugs still require further validation before proceeding as repurposed drugs for brain tumors. Further drug validation was applied based on their regulatory effect on the target genes.

3.5. Gene Set Enrichment Analysis and Connectivity Mapping for Repurposed Drug Validation

Two hundred forty-two (242) gene-interacting drugs were identified. Approved drugs were validated for their regulatory effects and CMAPs based on GSEA; see

Table 2. Regulatory effect of drugs on target genes based on GSEA and CMAP of Enrichr database.

Validated Drugs	Genes	Regulation
Dacarbazine	FN1 * and CACNA1A	Upregulate
Doxorubicin	COL1A1 * and SNCA	Upregulate
Gabapentin	CAMK2B, COL1A1 *, and SLC17A7	Upregulate
Nifedipine	COL1A1 * and KIF5C	Upregulate
Paclitaxel	CAMK2B and DNAJC6	Upregulate
Pyrantel	CAMK2B, COL1A1 *, DNM3, and SNAP25	Upregulate
Reserpine	FN1 * and SNCA	Upregulate
Resveratrol	COL1A1 * and MAPT	Upregulate
Sulfasalazine	CAMK2B and SNCA	Upregulate
Trifluoperazine	COL1A1 * and FN1 *	Upregulate
Valproic acid	COL1A1 *, DNAJC6, FN1 *, KIF5C, MAPT, MOBP, SLC12A5, SLC17A7, SNCA, STXBP1, SYT1, and TUBB2A	Upregulate
Vorinostat	DNAJC6, DNM3, KIF5C, SLC17A7, STXBP1, SYT1, and TUBB2A	Upregulate
Valproic acid	DNM1 $ and DYNC1I1 $	Downregulate

* HDEGs that are upregulated in tumor samples and exhibit further upregulation in the corresponding drug. ^$ HDEGs that are downregulated in tumor samples and exhibit further upregulation to the corresponding drug.

. CMAP-up dataset validated twelve (12) drugs that upregulate target genes. In contrast, the CMAP-down dataset validated one (1) drug that downregulates target genes; overall, twelve (12) drugs were validated to have regulatory effects on target genes. Selecting suitable drugs based on CMAP validation depends on their regulatory effect on target genes. Drugs that switch the regulation of target genes are considered suitable repurposed drugs for brain tumors. Certain drugs can further downregulate or upregulate target genes. These drugs were Dacarbazine, Doxorubicin, Gabapentin, Nifedipine, Pyrantel, Reserpine, Resveratrol, Trifluoperazine, and Valproic acid. Suitably validated candidate repurposed drugs for brain tumors that switch the regulation of target genes were Paclitaxel, Sulfasalazine, and Vorinostat. In the case of Gabapentin, Pyrantel, and Valproic acid, they switch more target genes than they further regulate, implying that these drugs remain suitable for repurposing. Among the validated drugs, Doxorubicin, Nifedipine, Reserpine, Resveratrol, and Trifluoperazine regulate one target gene while further modulating another. Further drug selection was performed based on the drug’s capability to permeate the BBB and HIA.

3.6. Pharmacokinetics Based on Brain or IntestinaL EstimateD (BOILED) Permeation Predictive Model

The twelve validated drugs—Dacarbazine, Doxorubicin, Gabapentin, Nifedipine, Paclitaxel, Pyrantel, Reserpine, Resveratrol, Sulfasalazine, Trifluoperazine, Valproic acid, and Vorinostat—were used to predict their pharmacokinetics to determine whether they permeate the BBB and HIA, as observed in Figure 8. The lead drugs that readily permeate the BBB and HIA were Gabapentin, Pyrantel, Resveratrol, Trifluoperazine, and Valproic acid. Vorinostat was nearly permeable to the BBB and was passively absorbed in the gastrointestinal tract, along with Dacarbazine, Nifedipine, and Reserpine. Among the twelve drugs, Doxorubicin and Paclitaxel were out of range, while Sulfasalazine had low absorption and bioavailability. Reserpine and Trifluoperazine were observed to be effluated from the central nervous system through P-glycoprotein.

4. Discussion

4.1. Hub Differentially Expressed Genes

Different brain tumor samples, based on DNA microarray expression profiles from GSE66354, GSE68848, GSE74195, and GSE43290, were used for differential gene expression analysis. Herein, ninety (90) shared DEGs were reported, seventy-three (73) downregulated (potential tumor-suppressing) and seventeen (17) upregulated (potential tumor-promoting), and probable prognostic biomarkers [5,32,57,58,59,60,61,62,63,64,65,66,67]. Diverse brain tumor types were included and treated as a single state of disease, which identified shared brain tumor drivers. Primary brain tumors, including gliomas (astrocytoma, oligodendroglioma, ependymoma, and glioblastoma), meningioma, and embryonal tumors (medulloblastoma, ATRT, and PNET), were covered in this study [29,30,31,32,33]. Some brain tumor samples may be classified according to the former WHO CNS classification, such as PNET. This term was removed and reclassified under embryonal tumors in the 2007 WHO CNS classification of tumors [7,8,68].

The study aimed to present shared tumor genes across brain tumors, providing a general gene target for drug–gene interaction search. The shared DEGs were used for gene co-expression network analysis, revealing a common link among brain tumors and joint gene targets [25]. Genetic markers revolutionize therapeutics and treatment options for patients with tumors, whether specific to individual cases or broadly applicable [5,22,66,69]. Gene co-expression network analysis has been used to investigate individual cases of gene associations with tumor development [22,70,71,72]. Brain tumors are well recognized to be highly heterogeneous, with even single tumor types (e.g., medulloblastoma) comprising multiple molecular subtypes characterized by distinct signaling pathways. In this context, our analysis does not assume biological equivalence across tumor types. Instead, the aim was to identify convergent molecular signatures that may be shared despite this heterogeneity, thereby highlighting potentially common biological processes and therapeutic vulnerabilities. Observed in

Table 3. The twenty-seven (27) target genes and their molecular importance.

	Gene Symbol	Description	Molecular Function	References
Cluster 1	CACNA1A	Calcium voltage-gated channel subunit α1 A	Neuronal P/Q type voltage-dependent Ca2+ channels, the central synapse neuromuscular junction, and neurotransmitter release	[73,74,75,76,77,78]
	DNM1	Dynamin 1	Mechanochemical GTPase for clathrin-mediated endocytosis, microtubule bundle formation, and mitochondrial fission machinery	[79,80,81,82,83,84]
	GABRA1	γ-Aminobutyric acid type A receptor subunit α1	GABAergic neurotransmission and transmembrane ligand-gated Cl− channel	[85,86,87,88,89]
	GRIA2	Glutamate ionotropic receptor α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid type subunit 2	Glutamatergic synaptic transmission, Q/R receptor, CP−-AMPARs differ from CI−-AMPARs, Ca2+ Flux, and excitatory postsynaptic currents	[90,91,92,93]
	MAPT	Microtubule-associated protein τ	Neuronal plasticity, microtubule assembly, and stability	[65,94,95,96,97]
	SLC12A5	Solute carrier family 12-member 5	Neuronal K+-Cl− co-transporter, activation of GABAA and glycine receptors	[98,99,100]
	SNAP91	Synaptosome-associated protein 91	Synaptic vesicle reformation, neuronal vesicle trafficking, and clathrin-associated protein sorting adaptors AP180	[57,97,101]
	STXBP1	Syntaxin binding protein 1	Synaptic vesicle fusion machinery, syntaxin-1 trafficking, SNARE complex formation, and presynaptic protein Munc18-1	[102,103,104,105,106,107]
Cluster 2	CAMK2B	Calcium/Calmodulin-dependent protein kinase II β	Serine/threonine kinase, neuronal migration, neuronal excitability, synaptic plasticity, and memory formation	[108,109,110,111,112]
	DNAJC6	DNAJ heat shock protein family (Hsp40) member C6	Auxilin chaperone engages HSPA8/HSC70 for the disassembly of clathrin-coated vesicles and endocytosis, and synaptic vesicle post-endocytotic recycling	[113,114,115]
	DNM3	Dynamin 3	Mechanochemical GTPases for clathrin-mediated endocytosis, microtubule bundle formation, and mitochondrial fission machinery	[84,116,117,118,119,120]
	DYNC1I1	Dynein cytoplasmic 1 intermediate chain 1	Retrograde-directed transport motor complex, subcellular localization of sphingosine kinase 2, RAS-RAF-MEK signaling pathway, and dendritic length	[121,122,123]
	KIF5C	Kinesin family member 5C	Cortical neuronal migration, dendritic branching, synaptic plasticity, and excitatory, and local translation of RNA substrates long-distance transport	[124,125,126,127,128]
	MOBP	Myelin-associated oligodendrocyte basic protein	Myelin sheath stabilization, myelin to a membrane-associated signaling complex, and oligodendrocyte differentiation	[129,130,131,132]
	NEFH	Neurofilament heavy chain	Neurofilament formation, neuronal caliber, and axonal sensorimotor	[133,134]
	SH3GL2	Src homology 3 domains containing growth factor receptor-bound protein 2-like 2	Endophilin A1 protein, early endosomes, BDNF-NTRK2, early endocytic signal trafficking, synaptic vesicle endocytosis, EGFR endocytosis, and STAT3/MMP2 signaling	[57,135,136]
	SLC17A7	Solute carrier family 17-member 7	Vesicular L-glutamate transporters, γ + LAT1 light chain, Na+-dependent inorganic phosphate (Pi) transport, and vesicular K+/H+ antiport activity	[137,138]
	SNAP25	Synaptosome-associated protein 25	Synaptic connectivity, plasticity, communication, and t-SNARE neurotransmitter release	[139,140]
	SNCA	Synuclein α	Synaptic vesicle trafficking, SNARE-complex assembly chaperone, and dopamine neurotransmission	[129,131,141,142,143]
	STMN2	Stathmin-2	Microtubules dissemble, Smad2/3 translocation, MAPK8 phosphorylation neurite extension and cell motility	[144,145]
	SYT1	Synaptotagmin 1	Neuronal growth-associated protein, microtubule stability, Ca (2+)-dependent interaction to putative receptors, and neurite length	[146]
	TUBB2A	Tubulin β 2A class IIA	GTP-tubulin dimers, α-tubulin GTPase activity, and neuronal migration	[67,147,148]
	TUBB4A	Tubulin β 4A class IVA	GTP-tubulin dimers, α-tubulin GTPase activity, and neuronal migration	[148,149]
Cluster 3	COL1A1	Collagen type I α1 chain	Type I collagen, fibrillar-forming collagen, and ECM formation	[70,150,151,152,153]
	COL6A2	Collagen type VI α2 chain	Collagen VI non-fibrillar heterotrimeric cell-binding protein, ECM disorganization and organization	[61,154,155]
	FBN2	Fibrillin 2	ECM organization, elastin fibers, extracellular calcium-binding microfibrils, and elastic fiber assembly	[156,157]
	FN1	Fibronectin 1	Cytoskeletal organization, ECM glycoprotein, cell surface binding of collagen, fibrin, heparin, DNA, and actin, WNT/β-catenin signaling, and FCGR1A/CD64-mediated monocyte activation	[58,158,159,160,161]

is the molecular function of the genes. Most of the suppressed genes were involved in neurotransmitter release, neuronal signaling, and cell-to-neuron communication. In contrast, expressed genes were associated with microfibrils, microtubules, and, ultimately, the extracellular matrix, which supports cell and tissue structure.

4.2. Cluster Genes and Enriched Signaling Pathways

The cluster 1 genes play a role in synaptic transmission and transmembrane transport, similar to cluster 2 genes, with the addition of synaptic vesicle endocytosis and exocytosis. In contrast, cluster 3 genes are involved in regeneration processes. The association of genes from clusters 1 and 2 with neurotransmission elucidates their dysregulation across a wide range of neurocognitive impairments and neurodegenerative diseases [19,20,61,74,93,146,150]. Recently, glutamatergic neuron-to-brain tumor synaptic communication (NBTSC) was shown to be hijacked by malignant cells, leading to alterations in neurotransmission and promoting tumor cell proliferation [88,162]. In the view of brain tumors, the downregulation of cluster 1 and cluster 2 genes may contribute to early tumor development, as it promotes the binding of several proteins and neuronal transmission [92,93]. In contrast, the upregulation of cluster 3 genes initiates tumor progression by promoting protease and collagen binding, which are responsible for the extracellular matrix, leading to an abnormal extracellular matrix in tumor growth [163].

The overexpression and underexpression of the different hub genes determined previously have a direct effect on tumor growth, proliferation, differentiation, migration, invasion, survival, resistance to therapy, and even the regulation of the extracellular matrix of the tumor microenvironment. Upon close investigation, most of the genes, when underexpressed, affect the PI3K-AKT and MAPK signaling pathways, which are primary drivers of tumor growth and cell proliferation. Additionally, the PI3K-AKT pathway contributes to chemotherapeutic drug resistance by promoting anti-apoptotic genes in the tumor [164]. On the other hand, the MAPK signaling pathway can upregulate oncogenes and exert tumor-suppressor activities depending on the tissue-specific tumor microenvironment [165]. These signaling pathways have been shown to exhibit pro-oncogenic activity, which is hard to overlook, thus presenting a new avenue for drug targets to cure cancer. Figure 9 shows the overexpressed genes (COL1A1 and FN1) and underexpressed genes, along with the associated signaling pathways they affect, which are directly related to tumor progression. Additionally, the overexpressed genes show overactivation of their associated pathways. Regarding the under-expressed genes, these genes have an inverse effect on the signaling pathways. For example, the SNCA gene is underexpressed yet overactivates the pathways it regulates, such as PI3K-AKT and MAPK signaling, thereby promoting cell proliferation and survival in cancer cells. Signaling pathways beyond the PI3K-AKT and MAPK pathways influence tumor progression; however, their effects are less direct than those of these pathways. Other signaling pathways modulate the PI3K-AKT and MAPK pathways by producing the proteins required for their activation. Proteins produced by other signaling pathways overactivated the PI3K-AKT and MAPK pathways, aiding tumor progression. To visualize, the CACNA1A gene is a gene that encodes for the CaV2.1 channel, which is crucial for cell cycle progression and proliferation. Changes in these genes can initiate progression, proliferation, and invasion [166]. As observed, the MAPK and PI3K-AKT pathways are the most affected and are associated with cell proliferation and survival. In addition, there is a need to investigate further the numerous indirect relationships among genes and signaling pathways.

The over- or underexpression of a hub gene could be a cause of the dysregulation of synaptic neurotransmitters. The binding of specific neurotransmitters can lead to the underexpression or overexpression of a hub gene, resulting in the production of proteins involved in the signaling pathway and ultimately in its underactivation or overactivation. Hence, the levels of neurotransmitters in synapses can greatly contribute to the progression of brain tumors. As evidenced by several studies, neurotransmitters are key players in tumor-related processes, including proliferation, differentiation, metastasis, and inactivity. Studies have shown that cells beyond the autonomic nervous system, such as immunocytes and tumor cells, release neurotransmitters. As a result, more neurotransmitters are present in the tumor microenvironment that non-neuronal cells can respond to, leading to over- or underexpression of hub genes [167,168]. Serotonin is one of the neurotransmitters linked to overactivation of the MAPK signaling pathway in several cancer types. Tumor progression occurs by inhibiting apoptosis in cancer cells and stimulating their proliferative activity. On the other hand, Dopamine decreases the activity of PI3K-AKT and MAPK pathways. PI3K-AKT reduces cancer activity, promotes cell cycle arrest, and decreases cell viability. As for the MAPK signaling pathway, it inhibits phosphorylation, rendering the pathway inactive [169]. Additionally, other signaling pathways may contribute to tumor progression when overactivated by several neurotransmitters, thereby indirectly affecting the PI3K-AKT and MAPK pathways. Thus, neurotransmitter levels should be an important criterion for assessing tumor activity, since tumor therapies can also target neurotransmitters that promote tumor growth by inhibiting them. This opens new avenues for investigating tumor activity, the tumor microenvironment, and novel drug targets not only in brain tumors but also in other cancer types [170,171,172]. Dysregulation of hub genes leads to sustained dysregulation of synaptic neurotransmitters, which influence brain activity and favor tumor development.

An important limitation of the present study is the lack of integration with clinical outcome data. Linking gene expression patterns to patient prognosis would strengthen the biological and translational relevance of the identified HDEGs. However, given the scope of this study and the absence of uniformly annotated clinical datasets compatible with our analysis framework, a systematic evaluation of outcome associations was not conducted. Future studies incorporating well-annotated clinical cohorts will be necessary to assess the prognostic and therapeutic significance of these genes.

4.3. Candidate Repurposed Drugs

The study revealed two hundred forty-two (242) approved drugs that interact with target genes, focused on the approved drugs rather than non-approved drugs to provide readily available option drugs in the market, offering fast and sustainable drugs that can be used for brain tumors. It is important to note that mRNA expression does not always directly correlate with protein abundance due to post-transcriptional and post-translational regulatory mechanisms. In the present study, transcriptomic data were used for initial screening to identify candidate genes and pathways. Accordingly, gene expression levels were interpreted as proxies for potentially dysregulated biological processes rather than direct measures of protein activity.

The identified approved drugs were further validated using GSEA and CMAP to determine their effects on target genes. Herein, twelve (12) drugs were identified and validated to exert regulatory effects on target genes: Dacarbazine, Doxorubicin, Gabapentin, Nifedipine, Paclitaxel, Pyrantel, Reserpine, Resveratrol, Sulfasalazine, Trifluoperazine, Valproic acid, and Vorinostat. Validated drugs were further selected based on their regulatory effect: they must reverse the regulation of the target genes, shifting from upregulated to downregulated and vice versa. Among these drugs, three that switched the regulation of the target gene were Paclitaxel, Sulfasalazine, and Vorinostat.

One of the obstacles to developing medications for brain tumors was their inability to cross the blood–brain barrier [14,173,174]. The twelve drugs were further tested for their pharmacokinetic properties, including their ability to cross the BBB and the gastrointestinal tract. The results of the pharmacokinetic study showed that only Gabapentin, Pyrantel, Resveratrol, Trifluoperazine, and Valproic acid passively permeate the BBB and gastrointestinal tract. At the same time, Dacarbazine, Nifedipine, Reserpine, and Vorinostat were passively absorbed in the gastrointestinal tract. Vorinostat was close to passing the BBB, while Reserpine and Trifluoperazine were predicted to be effluated from the central nervous system through P-glycoprotein. Sulfasalazine displayed low absorption, while Doxorubicin and Paclitaxel were out of range.

Reserpine and Trifluoperazine, as predicted to be effluated in the central nervous system by the P-glycoprotein, may not be suitable for brain tumors, as they may contribute to neuronal damage [175]. Although Trifluoperazine could freely pass the BBB, its possible neuronal impairment makes it not an ideal drug for brain tumors.

Valproic acid, an HDAC inhibitor, is considered an ideal treatment for malignant brain tumors, according to a recent review [176]. Valproic acid inhibits HDAC1, 2, and 3, melatonin 1 (MT1) receptor, methyl CpG binding protein 2 (MeCP2), and vascular endothelial growth factor (VEGF), which inhibits Akt/mTOR, Wnt pathway, and 5-HT2A signaling pathways, leading to inhibition of glioma proliferation, migration, and invasion [177]. In Taiwan, patients with high-grade glioma receiving temozolomide showed improved overall survival when valproic acid was added [177]. Remarkably, the combinatorial potential of Valproic acid with temozolomide has been demonstrated to prolong survival in patients with brain tumors [177,178].

Gabapentin is another drug that can be used as an antitumor drug for gliomas [179]. This drug can reduce glutamate synthesis by inhibiting the BCAT-1 gene and the thrombospondin-1 receptor α2δ-1 (THBS1). The protein from this gene is a subunit of the disulfide-linked homotrimeric protein, which acts as an adhesive by binding to fibrinogen, fibronectin, laminin, type V collagen, and integrins. THBS1 inhibition reduced functional connectivity, thereby inhibiting synaptogenesis and tumor cell proliferation [180].

Resveratrol could cross the BBB; however, it could further upregulate COL1A1, which is overexpressed in brain tumors. This natural polyphenol could inhibit Bcl-2, Survivin, and XIAP, thereby reducing anti-apoptotic protein levels and inducing apoptosis in cancer cells [176]. PDL1 expression in tumor cells was inhibited by Resveratrol, demonstrating its efficacy as an anti-cancer agent across several cancer types, including glioblastoma, myeloma, chondrosarcoma, colorectal, ovarian, breast, pancreatic, lung, prostate, cervical, and thyroid cancer [181], commonly hijacking the PI3K/MAPK pathway and NF-κB/Akt-STAT3 pathways [182].

Lastly, Pyrantel in the form of Pyrantel pamoate is an anthelmintic against E. vermicularis. The WNT/β-catenin signaling pathway is strongly associated with early cancer development and progression. Pyrantel inhibits WNT signaling pathways in certain cancer types by activating Casein kinase 1α (CK1α), leading to degradation of β-catenin, which promotes gene transcription [183]. There are no studies that explore Pyrantel as an anti-cancer. However, Pyrvinium, a common alternative to Pyrantel, was proven to have anti-pancreatic cancer in combination with Gemcitabine [184]. Through inhibition of STAT3 due to inhibition of mitochondrial electron-transport chain complex I and activating AMPK to inhibit STAT3-Tyr⁷⁰⁵ phosphorylation, thereby inhibiting apoptosis and inducing cell cycle progression of myeloma/erythroleukemia cells [185].

Due to the limitations of studies on Pyrantel, Resveratrol further upregulates COL1A1, and Trifluoperazine may cause neuronal damage within the BBB. This leaves two (2) suggestive candidate repurposed drugs, the Valproic acid and Gabapentin, based on their gene-deregulation activity and bioavailability to cross the BBB. Although Gabapentin and Valproic acid could also further upregulate COL1A1, this effect is offset by other genes that could switch regulation, such as CAMK2B and SLC17A7 in Gabapentin, and DNAJC6, KIF5C, MAPT, MOBP, SLC12A5, SLC17A7, SNCA, STXBP1, SYT1, and TUBB2A in Valproic acid.

These suggestive candidate repurposed drugs might require further validation before being considered for brain tumor treatment, including in vitro and in vivo studies. Furthermore, conducting in vitro and in vivo drug validation can confirm the proposed drug candidate’s effects and provide a comprehensive profile of its benefits and side effects. BOILED-Egg predicts only the passive diffusion in the BBB; however, it fails to evaluate the absorption, distribution, metabolism, excretion, and toxicity. Findings suggested a list of drugs suitable as a general treatment for brain tumors and may inspire the development of new drugs based on compound structures and derivatives. Combinatory treatment and cocktail drug formulation based on the list of repurposed drugs unlocks a promising future study investigation. Consistent bioactivity of the proposed drugs, depending on the preferred route of administration and their ability to cross the blood–brain barrier, should be considered in future drug repurposing. This parameter is greatly emphasized and poses an absolute challenge in discovering drugs for targeting brain-associated diseases such as brain tumors, Alzheimer’s disease, and Parkinson’s disease. Additionally, drugs that can easily pass through the BBB, including alcohols, anesthetics, nootropics, NSAIDs, and sedatives, can be interesting for drug repurposing studies to eliminate the difficulties in determining whether the drugs can pass through the BBB. Additionally, research on natural products for treating brain tumors is ongoing, making it a topic of interest for many researchers. Here, an established co-expression network approach for drug repurposing may serve as a beneficial platform for drug validation and development. Likewise, further experimental validation may be required before the use of the drugs in brain tumor treatment.

Our analyses integrated four independent datasets to identify consistent therapeutic hypotheses in brain tumors. However, external validation with a larger-scale cohort dataset would further strengthen the generalizability and robustness of the findings. Due to scope limitations and time constraints, this validation was not performed in the study. We value and recognize this as a future research validation to further support the translational relevance of our results.

5. Conclusions

Our study developed a platform to identify hub genes that could transform drug development and treatment for brain tumors. Using DNA microarray datasets (GSE66354, GSE68848, GSE74195, GSE43290) from various gliomas, embryonal tumors, and meningiomas, revealed 90 shared differentially expressed genes, 17 upregulated (tumor-promoting) and 73 downregulated (tumor-suppressing). Gene co-expression analysis revealed three key clusters: cluster 1 (transmembrane transport and synaptic transmission), cluster 2 (synaptic vesicle transport), and cluster 3 (extracellular matrix formation). These clusters highlighted 27 genes as potential targets for drug repurposing. Validated Drugs for repurposing suggest Valproic acid and Gabapentin as probable candidates for repurposing for brain tumors. However, in vivo and in vitro validation of the candidate drugs was not explored, limiting this study to predictive suggestion. Future work should focus on validating our results in larger cohorts to support translational relevance.