Unveiling the molecular mechanisms of clozapine-induced seizures: a network pharmacology and molecular docking-based investigation

Abstract

Clozapine is the preferred treatment for refractory schizophrenia. However, clozapine use is constrained by the risk of clozapine‑induced seizures (CISs). The molecular mechanisms behind CISs are complex and poorly understood. The objective of this study was to identify the core targets and signaling pathways contributing to CISs using network pharmacology and molecular docking. Potential targets for clozapine and seizures were screened using multiple databases, and overlapping targets were identified. Protein-protein interaction (PPI) networks were constructed using the STRING database and visualized in Cytoscape to identify core targets. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of the core targets were performed, and a disease-target-pathway-drug network was constructed. The docking of clozapine with the core targets was evaluated to determine the stability of their interactions. Of the 656 potential targets for clozapine and 1260 targets for seizure, 104 targets overlapped. PPI analysis of the overlapping targets yielded 12 core targets. KEGG pathway enrichment analysis demonstrated that glutamatergic, dopaminergic, and GABAergic synapses may be key pathways mediating CIS. The molecular docking results showed that clozapine had strong binding affinities for the core targets, BDNF, GRIN2B, GRIN2A, and GAD1. Thus, clozapine may interfere with the glutamatergic, dopaminergic, and GABAergic synaptic pathways by modulating BDNF, GRIN2B, GRIN2A, and GAD1. Modulation of these core targets may contribute to excitatory–inhibitory imbalance in the central nervous system, thereby potentially increasing seizure susceptibility. This study provides an integrative perspective on the potential pathophysiology of CIS and suggests directions for future experimental validation and risk intervention strategies.

Introduction

Schizophrenia is a chronic psychiatric disorder, affecting tens of millions of people worldwide and imposing a substantial public-health burden. Over the past 30 years, its global incidence has increased by approximately 37%¹. Clozapine is the highly effective first-line treatment for refractory schizophrenia². However, the clinical utility of clozapine is limited by adverse effects, including seizures^3–5. The incidence of clozapine-induced seizures (CIS) is up to 3.5%⁶. The risk of CIS positively correlates with clozapine dose, plasma concentration, and rate of dose escalation, which can severely impact patient quality of life and compromise treatment adherence; these effects can ultimately lead to poor clinical outcomes⁷. Electroencephalographic (EEG) studies revealed that clozapine slows background rhythms and induces epileptiform discharges, even in patients without clinically apparent seizures⁸. Unfortunately, the molecular mechanisms underlying CISs are poorly understood.

Most current studies concerning the mechanism of CIS have focused on the broad and complex pharmacology of clozapine. Clozapine interacts with numerous neurotransmitter receptors, including dopamine, serotonin, and histamine receptors, which are critical for maintaining excitability in neural networks^9–12. For example, serotonin receptors are widely distributed in brain circuits implicated in seizure generation, and clozapine acts as a potent serotonin receptor antagonist. Reduced central serotonin can increase neuronal excitability, lower the seizure threshold, and elevate seizure susceptibility^13–15. Clozapine acts on numerous targets via diverse mechanisms, and single‑target approaches are insufficient to comprehensively explain the epileptogenic effects of clozapine. Therefore, seizure is treated as an integrated gene network in this study, and the key nodes and pathways through which clozapine perturbs that network are identified.

Network pharmacology provides a systems-level approach to characterize complex drug–disease interactions. Biomolecular networks are constructed using network pharmacology to identify synergistic, multi‑target, and multi‑pathway mechanisms of drug action rather than following a single‑target paradigm¹⁶. Molecular docking simulates drug–target binding at the molecular level, providing a molecular basis for drug design and precision therapeutic strategies¹⁷. Therefore, this study combines network pharmacology and molecular docking to systematically identify the potential targets and molecular mechanisms of clozapine-induced seizures. The goal of this study is to provide an integrative, systems-level perspective on the potential neurotoxic mechanisms of clozapine that may inform future experimental validation and the development of safer clinical dosing strategies. The overall study workflow is shown in Fig. 1.

Fig. 1.

Technological roadmap of the study.

Methods

Identification of potential target genes of clozapine

The chemical structure and canonical Simplified Molecular Input Line Entry System for clozapine were obtained from PubChem¹⁸ (https://pubchem.ncbi.nlm.nih.gov/). To identify putative targets of clozapine, the following four databases were searched using the keyword “clozapine”: SwissTargetPrediction¹⁹ (http://swisstargetprediction.ch/), PharmMapper database²⁰, GeneCards²¹ (https://www.genecards.org/), and DrugBank²² (https://go.drugbank.com/). The retrieved targets were mapped to gene symbols using the UniProt database²³ (https://www.uniprot.org/). The target genes from the four databases were combined, and overlapping genes were eliminated.

Identification of seizure-related target genes

The GeneCards²¹ (https://www.genecards.org/) and Online Mendelian Inheritance in Man²⁴ (OMIM, https://www.omim.org/) databases were searched for Homo sapiens genes related to seizures. The retrieved genes from the two databases were combined, and overlapping genes were eliminated.

Common targets of clozapine and seizures

The potential targets of clozapine and seizures were uploaded to the online bioinformatics platform²⁵ (http://www.bioinformatics.com.cn/), and a Venn diagram was constructed to identify common targets.

Protein–protein interaction network construction and selection of core targets

A protein–protein interaction (PPI) network of the common targets was generated using the STRING database with the organism restricted to Homo sapiens. The minimum required interaction score was set to 0.4, corresponding to the default “medium confidence” threshold recommended by STRING, which has been widely adopted in network pharmacology studies to balance the retention of biologically meaningful interactions with the exclusion of low-confidence associations²⁶. Disconnected nodes were excluded from subsequent topological analysis. The PPI network was visualized and analyzed in Cytoscape²⁷ (v3.10.1). Core targets were identified based on the degree centrality, betweenness centrality, and closeness centrality using the CytoNCA plugin. Targets that exceeded the median value for all three centrality measures were retained as core targets.

Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses

To identify the biological functions of the core targets, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using the DAVID online tool²⁸ (https://david.ncifcrf.gov/). The GO analysis covered three domains: biological processes (BPs), cellular components (CCs), and molecular functions (MFs). Enriched terms were defined as terms with a significance threshold of p < 0.05. The top 10 significantly enriched terms from each GO category and the top 10 KEGG pathways were visualized as bubble diagrams using the Bioinformatics platform (http://www.bioinformatics.com.cn). Cytoscape²⁷ (v3.10.3) was used to construct a disease-pathway-target-drug network to intuitively illustrate the relationship between clozapine and seizure.

Molecular docking

Molecular docking was performed using AutoDock Vina (https://autodocksuite.scripps.edu/adt/) to determine the binding affinities between clozapine and the core targets. To generate the 3D ligand structure of clozapine, the 2D structure was retrieved from PubChem¹⁸, the energy was minimized using ChemBio3D (https://library.bath.ac.uk/chemistry-software/chem3d), and the optimized structure was converted to the PDBQT format using AutoDockTools (v1.5.6). Receptor structures were obtained from the Protein Data Bank (PDB)²⁹, and water molecules and co-crystallized ligands were removed in PyMOL 2.4.1 (https://pymol.org/2/). The crystal structures used for docking were selected according to the following criteria: Homo sapiens origin, appropriate structural resolution, inclusion of a complete or functionally relevant domain suitable for ligand docking, and absence of extensive missing residues in the predicted binding region. AutoDockTools was used to add polar hydrogens and assign Gasteiger charges before saving the receptors as PDBQT files. A docking grid box was defined for each target, encompassing the active site. Docking calculations were performed using AutoDock Vina. The lowest-energy binding models were selected for visualization and interaction analysis using PyMOL 2.4.1 and Discovery Studio 2025.

Results

Screening of potential targets of clozapine and seizure

Screening for core targets revealed 656 potential targets of clozapine, after removing duplicates, and 1260 potential targets of seizure. A comparison of the two screening results revealed 104 overlapping targets (Fig. 2A).

Fig. 2. Target prediction of clozapine and seizure.

A Venn diagram of potential overlapping targets. B Construction of a PPI network of intersection targets. C Core targets for seizure treatment were identified using Cytoscape. The color intensity, node size, and number of connecting lines in the diagram indicate the strength of the connections among targets. Twelve core targets were identified according to topological parameters, including degree centrality (DC), betweenness centrality (BC), and closeness centrality (CC).

PPI network construction and identification of key targets

The 104 overlapping targets were submitted to the STRING database to construct the PPI network (Fig. 2B). After removing all disconnected nodes and edges, the network consisted of 104 nodes and 774 edges. The PPI network was reconstructed and visualized using Cytoscape (v3.10.3), as shown in Figure 2C1. Targets with degree centrality, betweenness centrality, and closeness centrality above the median for each metric were selected, yielding the key targets (Figure 2C2 and 2C3). The twelve key targets included brain-derived neurotrophic factor (BDNF), amyloid precursor protein (APP), insulin (INS), glutamate ionotropic receptor NMDA type subunit 2B (GRIN2B), and catenin beta 1 (CTNNB1), glial fibrillary acidic protein (GFAP), tyrosine hydroxylase (TH), glutamate decarboxylase 1 (GAD1), glutamate ionotropic receptor NMDA type subunit 2 A (GRIN2A), pro-opiomelanocortin (POMC), synaptosomal-associated protein 25 (SNAP25), and metabotropic glutamate receptor 1 (GRM1).

GO enrichment and KEGG pathway analysis

The GO and KEGG enrichment analyses revealed 342 BPs, 83 CCs, 98 MFs, and 104 KEGG pathways. The top 10 enriched BPs included chemical synaptic transmission, negative regulation of gene expression, neurotransmitter transport, gamma-aminobutyric acid signaling pathway, regulation of presynaptic membrane potential, positive regulation of gene expression, response to xenobiotic stimulus, locomotor behavior, neuron apoptotic process, and regulation of synaptic plasticity. The most enriched MF terms included transmitter-gated monoatomic ion channel activity involved in regulation of postsynaptic membrane potential, ligand-gated monoatomic ion channel activity involved in regulation of presynaptic membrane potential, glutamate-gated calcium ion channel activity, ligand-gated monoatomic ion channel activity, glutamate binding, GABA-gated chloride ion channel activity, GABA-A receptor activity, identical protein binding, protein binding, and signaling receptor binding. The most enriched CC terms included dendrite, plasma membrane, postsynaptic membrane, presynaptic membrane, axon, neuron projection, synapse, cell surface, neuronal cell body, and GABAergic synapse. The top 10 enriched terms from each GO category are presented in a bubble chart (Fig. 3A).

Fig. 3. Enrichment analysis of clozapine‑induced seizure.

A The top 10 functions from the GO enrichment analysis of clozapine‑induced seizure. B The top 10 KEGG signaling pathways associated with clozapine‑induced seizure. C Sankey diagram based on KEGG enrichment analysis. The left rectangular nodes of the Sankey diagram represent the therapeutic targets, the right nodes represent the KEGG pathways, and the lines represent the associations between the targets and pathways.

The top 10 KEGG enrichment pathways are detailed in Table 1. Their connections to target genes are illustrated in the bubble plots in Fig. 3B. As shown in the Sankey diagram in Fig. 3C, the glutamatergic synapse, dopaminergic synapse, and GABAergic synapse pathways are likely contributors to clozapine-associated seizure induction.

Table 1.

Basic information about the KEGG enrichment results.

ID	Term	Count	Gene	P-value
hsa05030	Cocaine addiction	12	GRIN2A/DDC/CDK5/TH/MAOA/BDNF/PDYN/GRIN2B/GNAI1/SLC6A3/SLC18A2/GRIN1	2.91E-12
hsa04727	GABAergic synapse	14	GABRA2/GABBR2/GABRA1/GABRB1/KCNJ6/GABBR1/SLC32A1/GABRA5/GAD1/CACNA1C/SLC6A1/GNAI1/GLS/GNAO1	9.24E-12
hsa04724	Glutamatergic synapse	14	SLC1A2/SLC1A3/CACNA1C/GRIK2/GRIN2B/GNAI1/GRM1/GRIN1/GLS/GNAO1/PPP3CA/GRIN2A/GRM7/GNAQ	2.85E-10
hsa04728	Dopaminergic synapse	14	DDC/KCNJ6/MAOA/CACNA1C/GRIN2B/SLC6A3/GNAI1/MAPK10/GNAO1/PPP3CA/GRIN2A/TH/GNAQ/SLC18A2	1.45E-09
hsa05031	Amphetamine addiction	11	PPP3CA/GRIN2A/DDC/TH/MAOA/CACNA1C/PDYN/GRIN2B/SLC6A3/SLC18A2/GRIN1	2.89E-09
hsa04024	cAMP signaling pathway	16	GABBR2/CREBBP/GABBR1/BDNF/HTR1A/CACNA1C/GRIN2B/GNAI1/GRIN1/MAPK10/POMC/PAK1/GRIN2A/PDE10A/PIK3CA/CRH	1.74E-08
hsa04726	Serotonergic synapse	12	GNAO1/CYP2C9/APP/GABRB1/KCNJ6/DDC/MAOA/GNAQ/HTR1A/CACNA1C/GNAI1/SLC18A2	3.96E-08
hsa05032	Morphine addiction	11	GNAO1/GABRA2/GABBR2/GABRA1/GABRB1/KCNJ6/GABBR1/PDE10A/SLC32A1/GABRA5/GNAI1	4.50E-08
hsa04723	Retrograde endocannabinoid signaling	13	GABRA2/GABRA1/GABRB1/KCNJ6/SLC32A1/GABRA5/CACNA1C/GNAI1/GRM1/MAPK10/GNAO1/GNAQ/NDUFS1	6.36E-08
hsa05034	Alcoholism	14	NTRK2/DDC/MAOA/BDNF/PDYN/GRIN2B/SLC6A3/GNAI1/GRIN1/GNAO1/GRIN2A/TH/CRH/SLC18A2	1.06E-07

Disease-pathway-target-drug network construction

A disease–pathway–target–drug network was constructed to understand the potential mechanisms by which clozapine influences seizures. The network comprised 116 nodes (1drug node, 10 pathway nodes, 104 target nodes, and 1 disease node) and 245 edges (Fig. 4). Node types are distinguished by visual attributes (color and shape). The edges represent predicted associations between components.

Fig. 4.

Disease–pathway–target–drug network diagram illustrating the effects of clozapine on seizure.

Molecular docking

Based on the PPI topology and KEGG enrichment analysis, four potential core targets, BDNF, GRIN2B, GRIN2A, and GAD1, were selected for molecular docking with clozapine. The following crystal structures were used for docking: BDNF (PDB ID: 5FTO), GAD1 (PDB ID: 2OKJ), GRIN2A (PDB ID: 9MUL), and GRIN2B (PDB ID: 5EWJ). The highest docking affinity scores (AutoDock Vina) for each target were as follows: BDNF, −9.0 kcal/mol; GAD1, −8.5 kcal/mol; GRIN2A, −8.2 kcal/mol; and GRIN2B, −9.8 kcal/mol (Table 2). Lower binding energies indicate stronger predicted ligand–target affinity; the observed binding affinities ranged from −8.2 to −9.8 kcal/mol, indicating favorable clozapine binding to these targets. Clozapine established hydrogen‑bonding interactions with LEU1122(A) of BDNF (PDB: 5FTO); ASN256(A), ASP373(A), and ASN402(A) of GAD1 (PDB: 2OKJ); LYS534(A) of GRIN2A (PDB: 9MUL); and THR110(C) of GRIN2B (PDB: 5EWJ). The simulated binding interactions with these core targets are shown in Fig. 5A-D.

Table 2.

Binding energy of Clozapine with core targets.

Targets	PDB identifier	Docking score ((kcal/Mol))
BDNF	5FTO	−9.0
GRIN2A	9MUL	−8.2
GRIN2B	5EWJ	−9.8
GAD1	2OKJ	−8.5

Fig. 5. Molecular docking results of clozapine with core targets.

A Clozapine and BDNF. B Clozapine and GAD1. C Clozapine and GRIN2A. D Clozapine and GRIN2B.

Discussion

We integrated network pharmacology and molecular docking to identify the potential molecular mechanisms mediating clozapine-induced seizures. Our results suggest that clozapine induces seizures by disrupting the neuronal excitatory-inhibitory balance via modulation of multiple molecular targets and signaling pathways. These findings provide a new theoretical basis for clinical risk mitigation and provide the basis for the development of individualized treatment strategies. Although several of the identified targets have been previously associated with both clozapine action and seizure susceptibility, the present study integrates these dispersed findings into a coherent multi-target, multi-pathway model that may help explain the polypharmacological basis of CIS.

Twelve core targets were identified by screening the databases. Based on favorable molecular-docking affinities, which had the most direct and well-characterized mechanistic links to seizure, GRIN2A, GRIN2B, GAD1, and BDNF, were selected for detailed discussion. Molecular docking simulations showed that clozapine bound to GRIN2A, GRIN2B, GAD1, and BDNF with predicted binding energies ranging from −8.2 to −9.8 kcal/mol, suggesting favorable predicted interactions. The remaining core targets imply that additional, potentially synergistic mechanisms of clozapine action may also be involved in clozapine-induced seizures.

Insulin (INS) modulates neuronal excitability through ATP-sensitive potassium channels and GABAergic transmission³⁰, and both hyperinsulinemic hypoglycemia and insulin-resistance states are recognized as seizure-related conditions^31,32. Clozapine is also associated with metabolic abnormalities, including impaired glucose regulation and insulin resistance³³, and INS-related signaling may represent an indirect contributor to CIS. Amyloid precursor protein (APP), beyond its canonical role in synaptic structure³⁴, has been implicated in the regulation of neuronal excitability and linked to epileptiform activity in neurodegenerative conditions^35,36. GFAP reflects astrocytic activation and points toward a possible role of glial dysfunction and neuroinflammation³⁷. TH and POMC, which regulate the synthesis of dopamine and endogenous opioid peptides, respectively, contribute to upstream neuromodulation^38,39. SNAP25 is essential for synaptic vesicle release⁴⁰ CTNNB1 may be involved in synaptic function and structural stabilization^41,42. Therefore, our model does not exclude these targets but proposes that they may act as modulatory factors that interact with the more direct excitatory–inhibitory mechanisms mediated by glutamatergic and GABAergic pathways.

The NMDA receptor is a tetrameric complex composed of two GluN1 and two GluN2 subunits⁴³. GRIN2A and GRIN2B encode the GluN2A and GluN2B subunits, respectively. GluN2A and GluN2B regulate glutamatergic synaptic transmission and maintain the excitatory–inhibitory balance in neural networks^44–46. These two subunits differ in their spatial and temporal expression and functional roles. GluN2B is widely expressed during early embryonic development and is essential for neuronal development. GluN2A is primarily expressed postnatally and contributes to the functional maintenance of mature neurons^47,48. Pathogenic variants of GRIN2A and GRIN2B, causing gain or loss of function, can precipitate seizures by disrupting the function of NMDA receptors^49–54.

Based on this background, clozapine-induced seizures may be related to the dysregulation of GRIN2A and GRIN2B. Clozapine has been reported to alter the expression of these NMDA receptor subunits. For example, chronic clozapine treatment upregulates GRIN2B⁵⁵ and downregulates GRIN2A expression in the hippocampus⁵⁶. Given that NMDA receptors containing GluN2A and GluN2B critically regulate glutamatergic synaptic transmission, Ca²⁺ influx, and neuronal excitability, clozapine-induced alterations in GRIN2A/GRIN2B expression may indirectly disrupt NMDA receptor-mediated excitatory signaling. Such perturbation may increase network excitability and potentially increase seizure susceptibility^43,44,57,58.

Clozapine may also impair inhibitory systems by disrupting excitatory signaling. GAD1 encodes glutamic acid decarboxylase 67 (GAD67), the major rate-limiting enzyme for 90% of basal GABA synthesis in the central nervous system⁵⁹. GABA is the principal inhibitory neurotransmitter. Thus, reduced GABAergic function contributes to the generation of seizures⁶⁰. Functional defects in GAD1 trigger seizures. For example, GAD1 knockout rats exhibit abundant spike-wave discharges in the EEG at an early age⁶¹, and biallelic pathogenic variants of GAD1 cause infantile-onset epilepsy in humans⁶². We identified a potential high-affinity interaction between clozapine and GAD1, suggesting that clozapine may inhibit GAD1 activity, reduce GABA synthesis, and weaken global inhibitory control.

In clinical studies, the clozapine dose positively correlated with serum BDNF levels⁶³, and serum BDNF levels were higher in clozapine-treated patients compared with patients treated with typical antipsychotics⁶⁴. Animal studies also demonstrated that chronic clozapine treatment increases BDNF expression in the rat hippocampus⁶⁵. BDNF plays an important role in the onset and development of epilepsy. Elevated serum BDNF levels correlate with seizure severity, enhanced excitatory synaptic transmission via TrkB receptor activation and downstream signaling, and a reduced seizure threshold^66–68. Thus, clozapine-induced elevation of BDNF levels may synergize with the excitatory-inhibitory imbalance to augment neuronal excitability and facilitate circuit remodeling, increasing seizure risk. Taken together, these findings suggest that dysregulation of BDNF, GRIN2A, GRIN2B, and GAD1 may collectively disturb excitatory–inhibitory homeostasis and thereby increase seizure susceptibility.

BPs identified in the GO enrichment analysis included chemical synaptic transmission, neurotransmitter transport, and the GABAergic signaling pathway, response to xenobiotic stimulus, and the neuron apoptotic process. Based on these enriched BPs, clozapine may perturb neural signaling by affecting neurotransmitter release, recycling, and signal transduction. In addition, the enrichment of response to xenobiotic stimulus may reflect cellular adaptation to clozapine exposure, including drug metabolism, oxidative stress, and stress-response signaling^69,70. These processes may alter neuronal vulnerability or modify the effects of clozapine on neurotransmission. Similarly, the enrichment of the neuron apoptotic process may contribute to long-term network instability, particularly under chronic drug exposure or in patients with pre-existing neurological vulnerability^71,72. We therefore believe that these likely reflect general mechanisms related to drug metabolism and neurotoxicity, rather than mechanisms specific to CIS. Enriched CCs included the dendrite, the postsynaptic membrane, the presynaptic membrane, and axons, which are essential for neuronal signaling. The CC enrichment results, indicating that the main targets of clozapine are the regions involved in signal transmission and reception, support the BP results. The enriched MFs focused on functions of ligand-gated monoatomic ion channel activity and glutamate binding, in agreement with our GRIN analysis. These results suggest that the modulation of glutamate‑gated ion channels is a common feature of clozapine‑related targets.

KEGG pathway enrichment analysis supported the core-target results. Two functionally opposed pathways, the glutamatergic and GABAergic synapses, were enriched. GAD1 acts upstream in GABA synthesis⁵⁹ and GRINs are downstream glutamate receptors⁴³. The combined impact of clozapine on these pathways likely disrupts the excitatory–inhibitory balance, providing a theoretical model that may help inform future experimental investigation of the mechanisms underlying CIS^73,74.

Pathway enrichment analyses also implicated the dopaminergic synapses, providing a higher-level perspective on the excitatory-inhibitory imbalance. Clozapine is a dopamine receptor antagonist modulating the dopamine system⁷⁵. Thus, clozapine regulates the activity of glutamatergic neurons⁷⁶. Based on these results, we propose the following integrated, epileptogenic hypothesis: clozapine alters dopaminergic signaling, perturbs glutamatergic and GABAergic synaptic functions, and disrupts the excitatory-inhibitory balance by affecting key molecules, including the GAD1 and GRIN subunits. These effects are amplified by clozapine‑induced upregulation of BDNF–TrkB signaling. Altogether, these effects increase neuronal excitability and seizure risk. This systems‑level model offers a coherent framework for the effects of clozapine, which can be experimentally validated.

Several additional mechanisms beyond those identified in the present analysis have been proposed to contribute to CIS and deserve consideration. First, clozapine has prominent antimuscarinic/anticholinergic activity, which may contribute to seizure susceptibility by disrupting cholinergic modulation of cortical and hippocampal circuits^77,78. Cholinergic signaling regulates neuronal excitability, attention-related cortical activation, and interactions between excitatory–inhibitory interneuron networks^79,80. Therefore, anticholinergic effects may indirectly amplify glutamatergic excitation or weaken GABAergic inhibitory control⁸¹, thereby converging with the excitatory–inhibitory imbalance proposed in our model. Second, clozapine substantially alters sleep architecture, and sleep disruption is a recognized facilitator of epileptiform activity, which can lower the seizure threshold and may further amplify underlying network instability^82,83. Among the proposed mechanisms, clinical factors such as rapid dose titration, elevated plasma concentrations, and concomitant medications remain the most consistently reported determinants of CIS^84,85. Thus, CIS may arise from the convergence of direct molecular effects, neurotransmitter receptor pharmacology, metabolic factors, anticholinergic activity, clinical factors, and sleep architecture changes, our molecular model, and is therefore best interpreted as a complement to these established pharmacological and clinical considerations, rather than as a substitute for them.

Our results provide evidence for a novel systems-level framework for clozapine‑associated seizures. However, this study has several limitations. First, our work employed network pharmacology and molecular docking, which are database and algorithm-driven predictive approaches. Although this approach is useful for prioritizing candidate targets and pathways, these findings require in vitro and in vivo validation. Second, we did not account for interindividual genetic variability. Susceptibility to epilepsy and the risk of clozapine‑induced seizures vary, and the interplay between drug effects and host genetic variations should be investigated using clinical genomics data. In summary, our computational results generated a valuable hypothesis to guide experimental validation and individualized risk assessment, but the exact mechanism should be elucidated in future studies.

Conclusions

This is the first study to systematically delineate the multi‑target, multi‑pathway mechanisms underlying clozapine‑induced seizures using network pharmacology and molecular docking. Our results suggest that clozapine may disrupt opposing glutamatergic and GABAergic synaptic systems via interactions with the core targets, GRIN2A, GRIN2B, GAD1, and BDNF. These effects may be mediated upstream by dopaminergic signaling, disrupting the excitatory-inhibitory balance of neurons to increase the risk of seizures. These results provide a new framework for developing potential interventions, predictive biomarkers, and individualized dosing strategies for preventing clozapine-induced seizures.

Author contributions

Jian Li, Bo Han, Xia Liu, Ranli Li, Xiaoyan Ma: Writing—original draft, Visualization, Software, Data curation, Methodology. Hongjun Tian, Fuqiang Mao, and Chuanjun Zhuo: Writing—review & editing, Visualization, Supervision.

Funding

This work was supported by grants from the National Natural Science Foundation of China (Nos. 82171503 and 81871052) to Chuanjun Zhuo; Tianjin Anding Hospital Talent Support Grant (30,0000RMB to Chuanjun Zhuo). The funders had no further role in the study design; collection, analysis, and interpretation of data; writing of the report; or decision to submit the paper for publication.

Competing interests

The authors declare no competing interests.