Brain cancers are considered high-value research targets because of the complex dysregulation of multiple signaling pathways that leads to tumor growth, survival, and resistance. Commonly known pathways include those of receptor tyrosine kinase (RTK), epidermal growth factor receptors (EGFRs), platelet-derived growth factor receptor alpha (PDGFRA) (Li et al., 2023; Stachyra and Grzybowska-Szatkowska, 2025), MAPK with BRAF mutations (Talloa et al., 2022), reorganization of cell-cycle regulation of cyclin-dependent kinases (CDK1/2/3/4) along with cell-cycle checkpoints (Li et al., 2023), DNA repair and apoptotic systems (TP53 and O6-methylguanine-DNA methyltransferase (MGMT)) (Yu et al., 2020; Donson et al., 2007), telomere maintenance mechanisms such as telomerase reverse transcriptase (TERT) (Patel et al., 2020; Killela et al., 2013), and inflammatory mediators and angiogenesis-related signaling (Li et al., 2023). Of these, researchers have explored the specific roles of cyclooxygenases (COX-1/COX-2) and CYP2D6 in brain tumors, including their inflammatory and metabolic processes, which are well known in cancer biology. The direct involvement of these dysregulated pathways in the treatment of brain cancers makes them attractive candidates for drug repurposing. There has been an increased interest in drug repurposing in recent times owing to the time and cost savings that can be achieved compared to developing an entirely new drug. Among these, molecules approved by the Food and Drug Administration (FDA) are at the forefront of drug repurposing because of their established safety profiles, which can accelerate clinical translation (Xia et al., 2024). Brain-cancer-related drug repurposing requires additional attention to the ability of a drug to penetrate the blood–brain barrier (BBB) so that it can be delivered effectively to the central nervous system (CNS) (Ntafoulis et al., 2022).

The strategy for drug repurposing requires observation of the multidimensional interplay between molecular profiles and network-driven methodologies. Networks can be derived from computational models, may be real biological networks, or may be defined based on systems biology approaches (Gonzalez et al., 2025). However, despite the integration of multiple approaches, drug repurposing is scientifically and clinically limited by the need to adopt advanced methodologies against barriers, such as pharmacological boundaries, clinical trial parameters, and molecular stratification (Xia et al., 2024). In this context, we hypothesize that FDA-approved drugs can be used to observe dysregulated signaling pathways and can enable the exploration of therapeutically meaningful strategies for drug repurposing.

Network graphs are often used to capture the complexity of a biological system with isolated components relevant to the purpose of the study. These studies aim to understand the perturbations existing in a network of targets, genes, expressions, etc. The graph approach has also been adopted in different areas, such as network medicine and network pharmacology, in support of rational drug discovery (Joshi et al., 2025). Network analysis is also known for efficient drug repurposing as it integrates paired node data (Selvaraj et al., 2022). Beyond these uses, network analysis can be employed to reveal hidden parameter combinations from multiple biological pathways (Li et al., 2016).

Several therapeutic agents targeting brain cancers have received FDA approval for neurooncological purposes. Some of the well-known drugs include temozolomide for standard chemoradiation therapy, lomustine and carmustine for high-grade gliomas, and bevacizumab as a monoclonal antibody targeting vascular endothelial growth factors (Buzatu et al., 2025). Temozolomide is also used in combination therapies (Fisher and Adamson, 2021). Several mitochondrial inhibitors, such as trifluoperazine, mitoxantrone, and pyrvinium pamoate, have been repurposed for the treatment of brain cancers. Multitarget antiangiogenesis drugs such as axitinib, sorafenib, lenvatinib, sunitinib, cabozantinib, and regorafenib have also been approved for brain cancer therapy (Buzatu et al., 2025).

Because of the frequent chaotic interactions among altered targets, such as EGFRs, BRAF, PDGFRA, TP53, MGMT, CDKs, COX-1/2, vascular endothelial growth factor receptor 2 (VEGFR2), TERT, and CYP2D6, drug repurposing poses a high degree of uncertainty for therapeutic decision-making. However, the ensemble packaging of molecular signatures (or molecular profiles) in reference to these chaotic interactions reveals patterns for drug repurposing. This motivated us to develop two applications using molecular signatures, namely, “in-mac” and “ReBrain.” These applications were designed to reduce the complexity of the processes required for drug repurposing. Accordingly, scientifically simple observations were documented by generating structured molecular profiles that mirror disrupted pathways and were compared with FDA-approved drugs in an unbiased and data-driven manner. Together, these applications provide a systematic framework that supports more confident reasoning and discovery in drug repurposing. The developed applications were grounded in the need to analyze small molecules against the key dysregulated components of brain cancer and to organize the resulting data in a systematic platform for repurposing therapeutics. Thus, a computational profiling system was created to capture characteristic drug patterns in relation to different pathway components, and a database was developed to catalog these profiles for all FDA-approved drugs. Together, these resources enable large-scale comparisons of drug behaviors and the construction of a flexible network to identify repurposing opportunities. Finally, existing drugs for the treatment of brain cancer were processed through in-mac and ReBrain to discover the new repurposed drugs.

The molecular profile capabilities can be evaluated and validated through the consistent accuracy obtained using multiple machine-learning models, which are considered rigorous methods. Obtaining consistent accuracies for multiple machine-learning models also proves that the corresponding molecular profiles can be used for data featurization. The machine learning methods at the forefront of these efforts include logistic classification, decision trees, random forests, artificial neural networks, and support vector machines. These methods have already demonstrated strong performances for clinical predictions (Wu et al., 2023; Nwanosike et al., 2022). Similarly, machine-learning and deep-learning methods are well known in the fields of oncology and precision medicine for combining the molecular and phenotypic profiles for various data-driven validation frameworks applicable across multiple domains (Gawri et al., 2024; Balamurugan et al., 2024a; Jayaraman et al., 2024; Balamurugan et al., 2024b; Balamurugan et al., 2024c).

The proposed integrative approach allows prioritization of candidate repurposed drugs in reference to any chosen molecule to facilitate detailed exploration of their pathway contributions. The top-priority repurposed agents can be further reevaluated through in-mac-based analysis. Collectively, the proposed pathway-dysregulation-profile-based network-driven framework provides a powerful strategy to accelerate the discovery of repurposed, FDA-approved drugs.

The methodological rationale of this study is that brain cancers exhibit high molecular heterogeneity involving multiple signaling pathways, such that single targets or linear models are insufficient to capture the resulting biological interactions; therefore, a network-driven computational framework needs to be adopted. Molecular-signature-based profiling is significant for capturing patterns from multidimensional pathway interactions. The repurposed drugs were chosen from FDA-approved compounds because their established safety profiles could reduce the time and costs associated with drug development. Additionally, emphasis was placed on considering the BBB to ensure the clinical relevance of deliverability to the CNS. Therefore, existing brain cancer drugs were used in the analysis. Collectively, this integrative systems biology approach provides an informed method for drug repurposing. Therefore, an integrated computational workflow was developed by combining pathway-level dysregulation profiling and network-based searching of repurposed drugs based on reference drugs to systematically identify FDA-approved drugs with repurposing potential for brain cancers. Central to this workflow, we developed two new web-based applications called in-mac and ReBrain. The following sections describe the datasets used, application descriptions, and implementation for drug repurposing, which enable the identification and prioritization of candidate drugs based on their mechanistic relevance and pathway-level perturbation profiles (Figure 1).

Using the integrated computational pipeline and the two newly developed platforms, a systematic profiling was performed on the FDA-approved drug molecules. This combined framework enabled the generation of high-resolution molecular profiles, identification of recurrent pathway perturbations, and construction of a comprehensive drug–drug similarity network for drug repurposing. The results of these efforts are highlighted in three stages as follows: pathway-level alterations revealed by in-mac, network architecture and drug–drug functional similarities identified through ReBrain, and prioritization of candidate FDA-approved molecules for drug repurposing.

The present work integrates drug profiling with network-based analysis to explore repurposing opportunities for FDA-approved small molecules for treating brain cancer. Accordingly, a drug-repurposing study was conducted in two steps by referencing known molecules targeted at brain cancers against FDA-approved drugs. A profile generator in-mac was defined in the first step, while a multidimensional landscape was generated based on FDA-approved drug molecules in the second step. The landscape was implemented in a database called ReBrain, where a drug similarity network was used to identify drug repurposing potential. Sufficient literature is available on pathway-specific drug repurposing efforts in the recent past (Macak et al., 2025; Jia et al., 2025; Liu and Luo, 2024; Sinha et al., 2023). Drug repurposing is also evident with FDA- and European Medicines Agency (EMA)-approved molecules (Ma et al., 2025; Huang et al., 2025). The involvement of pathway dysregulation in drug repurposing is a known process (Wang et al., 2025; Zheng et al., 2025). In the case of brain cancer, there is available evidence on repurposing neurological drugs for brain cancer (Sarker and Newland, 2025). Therefore, in the present study, the pathway components were considered for drug repurposing. Here, the pathway components form a theoretical representation of the biological system defined by SAR-based molecular profiles, and these profiles were generated using the in-mac application. The utility of in-mac can be observed in terms of the authenticity of the quantitative SAR models for predicting molecular activities. In-mac generates 1,000 SAR models for each query molecule and uses only these prediction models, which enables an accuracy of >80%, provided the query molecule lies within the applicability domain of each SAR model. For SAR model development, the Monte–Carlo method was adopted to capture the natural behaviors for prediction. Therefore, there may be slight deviations in the predicted values because of the randomized process. However, beyond these deviations, the application is highly accurate for population data prediction. In-mac creates bioassay models for activity predictions, where these bioassay models contain patterns regarding the links between structures and their activities. Fifteen specific assay components were chosen to quantify the molecules using approximately all possible pharmacodynamic and pharmacokinetic methods. This represents significant diversity coverage in defining therapeutic concerns for brain cancer. The combination of these model patterns is used to create a profile indicating the functional behaviors of the molecules in a system. These profile components are separable at the elemental level, so this method can be used for profile matching. Profile-based drug–drug functional similarity is the direct representation of drug repurposing. Here, the proposed workflow is strong as it considers global properties for decision-making, while limitations may arise with the extent of data availability in public databases. Regression entails vector elements for establishing the relationships between vectors and is therefore significant for evaluating the profile match. Profile similarities may be negative or positive; therefore, there was a need to consider the squared differences between the molecules to calculate neutral values for the similarity matrix. Since Euclidean distance computation involves this neutralization process, it was used in this study for network creation. Theoretically, the repurposing workflow used here is significant for making predictions. The present workflow can be considered as a theoretical representation of a biological system, but the results from ReBrain were proven to be authentic since they were found to have common pathways of the repurposed drugs.

The R ² values for cefaclor vs. belzutifan (R ² = 0.04), glyceryl 1-monooctanoate vs. eflornithine (R ² = 0.17), and neryl acetate vs. temozolomide (R ² = 0.22) prove that the corresponding molecules are not functionally similar at the systemic level based on the 15 profile components. The genotypic–phenotypic relationships were found to be significant when R ² was >0.6, as in the cases of clofibric acid vs. carmustine (R ² = 0.90) and vorasidenib citrate vs. mefloquine (R ² = 0.91, weight = 0.88). In the case of armillarisin A vs. lomustine (R ² = 0.34, weight = 0.97), a high weight was observed with a low R ² value that indicated slight deviations at the anti-inflammatory activity level but good alignment in the functionalities of the other pathway components.

In terms of the biological interpretation and mechanistic insight of the prioritized repurposed drugs, the following observations are noted. (i) Mefloquine shows high potential for treating brain cancer based on various in vitro and in vivo studies, including those investigating cytotoxicity and inhibition of angiogenesis. Mefloquine is a known antimalarial molecule that targets lysosomal integrity; it induces oxidative and endoplasmic reticulum stress and can also trigger apoptosis and autophagy inhibition in tumor cells. It shows good BBB penetration and can sensitize tumor cells via inhibition of drug efflux pumps. Mefloquine is reported to act by interfering with lysosomal functions and autophagy flux; it has been shown to activate different stress pathways (JNK, ERK, and AMPK) and inhibit PI3K/Akt/mTOR signaling in several tumor models. In glioblastomas, it was shown to inhibit angiogenesis by targeting glioblastoma microvascular endothelial cells and impairing tumor vasculature (Wan et al., 2021). (ii) Clofibric acid shows low potential for direct brain cancer therapy; it acts as a PPARα agonist and is involved in fatty-acid-β oxidation, peroxisome proliferation, and xenobiotic metabolism. In this manner, it influences tumor cell lipid metabolism and oxidative stress. It also engages in PPARα signaling (Wang et al., 2010). (iii) Armillarisin A is not a plausible treatment for brain cancer at present; it is a natural product (from fungal sources) and shows antimicrobial and anti-inflammatory activities. It is known to be involved in general pathways, such as oxidative stress response, inflammatory signaling, and possibly microbial defense. No citations were found relating it to brain cancer.

The main limitations of the proposed methods here are that they have bound profile constraints for defining biological systems and lack sufficient experimental data for authentic empirical modeling of individual system components. These limitations should be addressed over time through the availability of diverse experimental datasets and the discovery of novel drug targets.

The proposed applications are suitable for observing all possible permutations and combinations of therapeutic molecules for the repurposing of not only existing but also future drugs. Furthermore, new profile components may need to be added to enhance the accuracy of deep searching of repurposed drugs. In the future, additional assay components can also be defined and included in the applications to enhance their accuracy. Although we have used machine-learning processes in this work, other options for pathway selection and phenotype predictions may be added in the future. Moreover, clinical data and adverse effect profiles can also be integrated to refine the repurposed drugs. Researchers can also supplement relevant data in terms of community-level inputs to enhance the utility of the database.

The present work entails rational drug repurposing by incorporating the molecular heterogeneity of brain cancers into two new computational frameworks, namely, the in-mac tool and the ReBrain database. Here, dysregulated pathway components were utilized to develop the molecular profiles of FDA-approved drugs using in-mac. Then, 2,809 FDA-approved small molecules with molecular weights ≤500 Da were profiled to create a multidimensional network-based database. The proposed profile-network-based method showed 70%–95% accuracy for drug repurposing across different disease categories related to the brain. By analyzing known drug molecules against brain cancers, we identified three repurposed drug molecules as high-priority candidates: mefloquine, clofibric acid, and armillarisin A. Of these, mefloquine was found to meet all the criteria established in this study. These findings demonstrate that system-level network analyses support drug repurposing and synergistic opportunities for the treatment of different types of brain cancers. Furthermore, additional profile components can be identified to ensure the authenticity of the profiles in drug repurposing and the synergistic utilization of the identified drugs.