Deciphering bisphenol A (BPA)-elicited osteoarthritis mechanisms through network toxicology and molecular docking, then de novo generation of novel therapeutic candidates Authors: Sumei Xu, Liping Jiang, Zhuo Zhang, Xin Luo, Wei Wu, Huilan Wu, Zhichun Huang, Zackary Falls, Ram Samudrala, Zhirong Tan Published: 2025 ABSTRACT AbstractObjectiveBisphenol A (BPA), a pervasive environmental pollutant, is increasingly associated with osteoarthritis (OA) development, yet its molecular mechanisms remain unknown. Currently, there is no definitive cure for OA.MethodsBPA targets were predicted using STITCH and Swiss Target Prediction, while OA-related targets were collected from GeneCards, OMIM, and the Therapeutic Target Database (TTD). Protein-protein interaction (PPI) networks were constructed using STRING and visualized in Cytoscape to identify hub targets. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed, and molecular docking with AutoDock evaluated BPA-core target interactions. We employed our Computational Analysis of Novel Drug Opportunities (CANDO) platform for de novo drug prediction.ResultsSystematic bioinformatics analysis identified 26 candidate targets, with ESR1, PTGS2, CCL2, FLNA, and TRPV1 as key hubs. Pathway analysis revealed involvement in calcium ion transport, muscle contraction, IL-17 signaling, and estrogen signaling. Molecular docking confirmed strong BPA-target binding affinities. CANDO predicted 14 potential OA treatments, including glucosamine, ibuprofen, celecoxib, indomethacin, palmitic acid, and linoleic acid. Notably, qRT-PCR validation revealed that ESR1, PTGS2, CCL2, and TRPV1 were highly expressed, whereas FLNA was expressed at lower levels in the osteoarthritis blood samples.ConclusionsThis study elucidates BPA’s molecular mechanisms in OA and identifies promising therapeutic candidates. The integration of network toxicology, molecular docking, and computational drug discovery provides a robust framework for understanding environmental toxicants and advancing OA therapies.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40360-025-01069-w.Keywords: Osteoarthritis, Bisphenol A, Network toxicology, Machine learning, Computational biology Introduction Osteoarthritis (OA), the most prevalent joint disorder worldwide, is characterized by progressive cartilage degradation and structural alterations in associated tissues, including ligaments, synovium, meniscus, and subchondral bone. These changes lead to chronic pain, functional impairment, and increased mortality [1–3]. Globally, symptomatic OA affects approximately 10% of men and 18% of women aged 60 and above [4, 5]. The etiology of OA is multifactorial, involving genetic, biological, biomechanical, and environmental influences [6–8]. The widespread use of plastics in modern life has resulted in significant human exposure to endocrine-disrupting chemicals (EDCs), such as bisphenol A (BPA), a key component in plastic production [9, 10]. BPA is predominantly manufactured in 12 countries, with the USA, Germany, and Japan being major contributors [11]. It is commonly found in epoxy resins lining food and beverage containers and in polycarbonate plastics used in various consumer products. Human exposure occurs primarily through diet, as well as drinking water, dental sealants, skin contact, and inhalation of household dust, with detectable levels present in over 90% of the US population [12, 13]. Studies have demonstrated BPA’s endocrine-disrupting effects in experimental animals [11, 14], and growing evidence suggests its adverse impacts on human health, particularly in reproductive and neurological systems [15], immune functions [16], and other areas [17, 18]. BPA interacts with hormone receptors, disrupting testosterone [19, 20] and thyroid hormone [21] signaling, both of which play critical roles in maintaining joint and cartilage health. Disruptions in these hormonal pathways have been identified as potential risk factors for OA [22, 23]. Furthermore, BPA has been shown to antagonize estrogen’s protective effects on chondrocytes in a concentration-dependent manner, potentially contributing to OA progression [24]. Despite these findings, research exploring the link between BPA and OA remains limited, with most studies focusing on BPA’s estrogenic effects, such as alterations in reproductive system development and prolactin modulation. While these studies hint at BPA’s role in OA pathogenesis, the specific molecular mechanisms remain poorly understood. To address this gap, this study employs network toxicology and molecular docking to investigate the molecular mechanisms underlying BPA-induced OA. Network toxicology integrates bioinformatics with high-throughput data from genomics, proteomics, and metabolomics, enabling a systematic assessment of multi-target toxicants [25]. Molecular docking complements this approach by simulating ligand-target interactions and predicting binding affinities [26]. Together, these methods provide a robust framework for elucidating the molecular interactions between BPA and OA, advancing our understanding of its toxicological effects and potential role in OA development. Currently, there is no cure for OA. To address this challenge, we developed the Computational Analysis of Novel Drug Repurposing Opportunities (CANDO) platform, which leverages multitarget drug discovery and repurposing strategies to overcome the limitations of traditional single-target approaches [27–42]. CANDO constructs interaction signatures that reflect drug/compound behaviors across multiple biological scales, enabling the prediction of putative drug candidates for various diseases. By comparing and ranking these signatures, CANDO identifies compounds with similar biological behaviors, offering a novel approach to drug discovery. In this study, we utilize CANDO’s de novo predictive pipeline to identify potential therapeutic candidates for OA, providing new insights into its treatment and prevention. Methods Identification of BPA-associated targets The chemical structure and SMILES notation of BPA were obtained from the PubChem database [43] using the keyword “Bisphenol A.” Molecular targets for BPA were then retrieved from the ChEMBL database [44], focusing on Homo sapiens. To broaden the search, the SMILES notation was queried in the STITCH (http://stitch.embl.de/) and Swiss Target Prediction (http://www.swisstargetprediction.ch/) databases, applying a prediction probability threshold >0. Results from all sources were merged to create a comprehensive library of potential BPA-associated targets. Compilation of OA-related targets Potential molecular targets linked to OA were collected from three key bioinformatics resources: GeneCards (https://www.genecards.org/), OMIM (https://omim.org/), and the Therapeutic Target Database (TTD) (https://db.idrblab.net/ttd/). To refine the GeneCards data, only targets with a relevance score > 5 were included. The relevance score in GeneCards represents the strength of association between a gene and a disease, based on integrated evidence from multiple sources including publications, genetic studies, and functional annotations. A cutoff of > 5 was selected to include genes with moderate to strong disease relevance while excluding weakly associated targets. This threshold ensures a comprehensive yet reliable set of candidate genes, consistent with previous studies utilizing GeneCards-based target prioritization. The resulting data were combined, and duplicate entries were removed to generate a unified list of non-redundant OA-related targets. A Venn diagram was created to identify overlapping targets between BPA and OA, with these shared targets representing potential candidates for BPA-induced OA. Additionally, Cytoscape (version 3.10.3) was employed to construct a compound-target-disease network. Construction of protein-protein interaction (PPI) network and screening of core targets Overlapping targets between BPA and OA were identified by intersecting their respective target lists, highlighting potential mediators of BPA-induced OA. These targets were analyzed using the STRING database (http://string-db.org), focusing on Homo sapiens. Analysis settings included a minimum interaction confidence score > 0.4 and a “High” false discovery rate (FDR) stringency value to ensure reliable protein interaction predictions. The interaction data from STRING were imported into Cytoscape (version 3.10.3) for visualization and network analysis [45]. Key network metrics such as degree centrality, betweenness centrality, and closeness centrality were calculated to evaluate the importance of individual proteins within the PPI network. To further refine the analysis, the CytoHubba plugin was used [46], employing the Maximal Clique Centrality (MCC) algorithm to identify hub proteins in highly interconnected regions. The MCC algorithm measures the centrality of a node based on the number and size of maximal cliques to which it belongs. A clique is a subset of nodes in which every two nodes are directly connected; the larger and more numerous the cliques a node participates in, the higher its MCC score. Nodes with higher MCC scores are considered more essential for maintaining the network’s structure and biological function. The top-ranked genes according to MCC were defined as hub genes for subsequent analyses. The MCODE plugin was also used for cross-verification, ensuring robust identification of core targets [47]. These prioritized hub proteins were selected for deeper exploration through molecular docking studies and functional investigations. Functional and pathway enrichment analysis To understand the biological significance of the potential targets involved in BPA-induced OA, functional enrichment analyses were performed using the clusterProfiler R package [48]. Gene Ontology (GO) analysis classified targets into categories of biological processes (BP), molecular functions (MF), and cellular components (CC) [49]. Circos plots were produced, which utilizes the ComplexHeatmap [50] and Circlize [51] R packages for generating Circos plots to compare data from association analyses. Concurrently, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis identified critical pathways associated with BPA-related OA mechanisms. A stringent P-value cutoff of 0.05 was applied, and enriched terms required a minimum overlap of three genes with the input set to ensure statistically meaningful results. Molecular docking analysis of BPA and core targets Molecular docking studies were conducted to investigate interactions between BPA and core targets. The three-dimensional crystal structures of the target proteins were retrieved from the RCSB Protein Data Bank (PDB). Selection criteria for PDB structures included: (1) the protein originated from Homo sapiens, (2) a high-quality crystal structure with a resolution better than 3.0 Å was available, (3) the structure contained a complete or nearly complete amino acid sequence, and (4) no major mutations or missing residues were present in the active site. When multiple entries were available for the same protein, the structure with the highest resolution and best overall quality was selected. The selected protein structures were preprocessed using PyMOL, which involved not only the removal of water molecules and ligands but also a cleaning step to add missing amino acid residues and correct mismatched sequences, ensuring the structural integrity of the proteins. Hydrogen atoms were added and partial charges were assigned using AutoDock Tools 1.5.7. Docking simulations were performed with AutoDock Vina, and the resulting binding affinities and interaction modes were analyzed and visualized using Discovery Studio and PyMOL. To validate the docking procedure, a standard redocking approach was applied by re-docking the co-crystallized ligands into their corresponding protein active sites. The reproduced binding conformations were consistent with the experimental poses, confirming the reliability of the docking protocol. Furthermore, molecular dynamics (MD) simulations were conducted to assess the stability of the docked protein–ligand complexes, providing additional validation and dynamic insights into the binding interactions. Osteoarthritis drug candidate generation using CANDO To identify potential drug candidates for OA, we employed a de novo pipeline within the CANDO platform. This pipeline ranked compounds based on their predicted interaction scores against five hub proteins implicated in OA. This approach predicts potential therapeutics for OA by summing or counting the number of interactions above a specified threshold, with a threshold of 0.0 (no threshold) applied in this study. Interaction scores were calculated using our in-house bioanalytic docking protocol (BANDOCK), which estimates binding strength or probability [37, 38, 40, 42]. The de novo protocol aggregated the interaction scores of each compound across all hub proteins and ranked them from highest to lowest. The top-ranked compounds were further assessed for their therapeutic potential in OA. To validate the predictions, we compared the results with known OA drugs and experimental evidence from the literature, ensuring the reliability and relevance of the identified candidates. Quantitative real-time PCR (qRT-PCR) analysis To further validate the expression levels of the hub genes identified in OA (as defined in Sect. “Construction of protein-protein interaction (PPI) network and screening of core targets”), peripheral blood samples were collected from external participants. A total of five samples were obtained from patients with OA at Xiangya Hospital, Central South University. All participants provided written informed consent prior to sample collection. Patients with other inflammatory joint diseases, autoimmune disorders, infections, or malignancies were excluded. The study protocol was reviewed and approved by the Institutional Review Board of Xiangya Hospital, Central South University (No: 2025101788), and all procedures were conducted in accordance with the institution’s ethical guidelines. The control blood samples used for qPCR analysis were obtained from healthy individuals through the Biobank of Xiangya Hospital, Central South University, which has been accredited by China National Accreditation Service for Conformity Assessment (CNAS). The collection, storage, and use of samples were in accordance with ISO 20,387: 2018 Biotechnology-Biobanking-General Requirements for Biobanking (CNAS-CL10 Accreditation Criteria for the Quality and Competence of Biobank). Total RNA was extracted from the blood samples using a commercial extraction kit (Solarbio, China) according to the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized from the isolated RNA using the SynScript® III RT SuperMix with genomic DNA removal reagent (Tsingke, China). Quantitative real-time PCR (qRT-PCR) was performed on a ViiA™ 7 Real-Time PCR System (Applied Biosystems, USA) using the SYBR Green Pro Taq HS kit (Accurate Biology, China). GAPDH was used as an internal reference gene for normalization. The amplification program consisted of an initial denaturation at 95 °C for 1 min, followed by 40 cycles of 95 °C for 10 s, 60 °C for 20 s, and 72 °C for 10 s. Relative mRNA expression levels were calculated using the 2−ΔΔCt method. All primers were synthesized by Tsingke Biotechnology Co., Ltd. (Beijing, China), and their sequences are listed in Table S1. Results Identification of potential targets for BPA-elicited OA In this study, we initially identified 2,988 potential molecular targets of BPA using the ChEMBL, STITCH, and Swiss Target Prediction databases. These targets were cross-referenced with 297 OA-related targets compiled from GeneCards, OMIM, and the Therapeutic Target Database (TTD). After integration and removal of duplicates, 26 overlapping targets were identified as potential mediators of BPA-induced OA toxicity (Fig. 1A). A compound-target-disease network was constructed to visualize the relationships between BPA, its targets, and OA (Fig. 1B). Identification of potential targets associated with bisphenol A (BPA)-induced osteoarthritis (OA). (A) Venn diagram showing the overlap between BPA-related targets predicted from the STITCH and SwissTargetPrediction databases and OA-related targets retrieved from GeneCards, OMIM, and the Therapeutic Target Database (TTD). (B) The intersecting region represents 26 common targets that are potentially involved in BPA-induced OA pathogenesis. These overlapping genes were selected for subsequent network and enrichment analyses. Abbreviations: BPA, bisphenol A; OA, osteoarthritis PPI network and hub targets acquisition The 26 intersecting targets were mapped onto a PPI network using the STRING database, resulting in a network comprising 26 nodes and 38 edges. The network exhibited an average node degree of 2.92 and a highly significant PPI enrichment p-value (< 1.13e-09), indicating robust interactions among the targets (Fig. 2). Using the MCC algorithm, five hub targets were identified: Estrogen Receptor 1 (ESR1), Prostaglandin-Endoperoxide Synthase 2 (PTGS2), C-C Motif Chemokine Ligand 2 (CCL2), Filamin A (FLNA), and Transient Receptor Potential Vanilloid 1 (TRPV1). These hub targets are represented as small red circles in the network visualization, with other targets depicted as larger circles, where redder colors indicate higher degree centrality values (Fig. 3). Protein-protein interaction (PPI) network of the 26 potential targets. The PPI network was generated using the STRING database and visualized in Cytoscape software. Each node represents a protein, and each edge denotes a predicted interaction. The dense interconnectivity indicates strong functional relationships among the targets, suggesting coordinated roles in OA-related biological processes. Abbreviation: PPI, protein–protein interaction Hub target identification in the PPI network. The PPI network was further analyzed using the CytoHubba plugin in Cytoscape to determine key hub proteins based on degree centrality. The top five hub targets—ESR1, PTGS2, CCL2, FLNA, and TRPV1—are highlighted in red, indicating their central roles in the BPA-induced OA network. These targets are likely to serve as crucial regulators in the disease mechanism. Abbreviations: ESR1, estrogen receptor 1; PTGS2, prostaglandin-endoperoxide synthase 2; CCL2, C-C motif chemokine ligand 2; FLNA, filamin A; TRPV1, transient receptor potential vanilloid 1 Functional and pathway enrichment analysis GO enrichment analysis of the 26 targets was performed using the clusterProfiler R package. The top ten enriched terms for biological processes (BP), cellular components (CC), and molecular functions (MF) were identified and visualized as bar plots and Circos plots (Fig. 4). Key BP terms included calcium ion transport and muscle contraction, while CC terms highlighted cortical cytoskeleton, focal adhesion, and cell-substrate junction. MF terms such as nuclear receptor activity, ligand-activated transcription factor activity, steroid binding, and endopeptidase activity were also significantly enriched. Gene Ontology (GO) enrichment analysis of the 26 potential targets. The top enriched GO terms in biological process (BP), cellular component (CC), and molecular function (MF) categories are shown in bar and Circos plots. Abbreviations: GO, Gene Ontology; BP, biological process; CC, cellular component; MF, molecular function KEGG pathway analysis revealed the top signaling pathways associated with BPA-induced OA, ranked by p-value and visualized as bubble charts and histograms (Fig. 5). Key pathways included the IL-17 signaling pathway, estrogen signaling pathway, and TNF signaling pathway, among others. These findings suggest that BPA may influence OA progression through these critical pathways. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. The bubble chart and histogram display the top significantly enriched signaling pathways obtained from the KEGG database. Pathways such as TNF signaling, IL-17 signaling, and estrogen signaling were among the most prominent, suggesting that BPA exposure may exacerbate OA through inflammation and hormonal regulation mechanisms. Abbreviation: KEGG, Kyoto Encyclopedia of Genes and Genomes Molecular docking for BPA with core targets Molecular docking simulations were conducted to evaluate the binding interactions between BPA and the five hub targets (ESR1, PTGS2, CCL2, FLNA, and TRPV1) using AutoDock. The corresponding PDB IDs and resolution values of these proteins are listed in Table 1. All docking interactions were successful, with vina scores ≤ -5.0 (Table 1). The most stable interaction was observed between BPA and ESR1, exhibiting a binding energy of -8.3 kcal/mol. Detailed visualization of hydrogen bond interactions and key amino acid residues involved in the binding is provided in Fig. 6. Intermolecular distances between BPA and the interacting residues were measured using PyMOL, and all hydrogen bonds and close contacts within 3.5 Å were considered significant, confirming the plausibility of the binding conformations. These results underscore the strong binding affinity of BPA to ESR1, suggesting its potential role in mediating BPA-induced OA pathology. Table 1Molecular docking results between BPA and each core target proteinIngredientTargetPDB IDVina scoreBisphenol AESR16PSJ-8.3Bisphenol APTGS25F19-7.9Bisphenol ACCL22NZ1-7.0Bisphenol AFLNA4M9P-6.1Bisphenol ATRPV18GF8-7.6 Molecular docking results between BPA and each core target protein Molecular docking results of bisphenol A (BPA) with the core target proteins. The three-dimensional docking conformations of BPA with the five hub targets—(A) ESR1, (B) PTGS2, (C) CCL2, (D) FLNA, and (E) TRPV1—are illustrated. Molecular docking was performed using AutoDock Vina to predict the binding affinity and potential interaction sites. For each target, the lowest binding energy conformation was selected to represent the most stable binding pose. Key interacting residues are labeled, and hydrogen bonds are shown as yellow dashed lines. Intermolecular distances were measured using PyMOL, and all hydrogen bonds and close contacts within 3.5 Å were considered significant. The docking results suggest that BPA forms stable interactions with residues critical for target protein function, indicating potential modulation of these targets by BPA exposure. Abbreviations: BPA, bisphenol A; ESR1, estrogen receptor 1; PTGS2, prostaglandin-endoperoxide synthase 2; CCL2, C-C motif chemokine ligand 2; FLNA, filamin A; TRPV1, transient receptor potential vanilloid 1; Å, angstrom Computational prediction of osteoarthritis drug candidates Using the CANDO platform, we generated a ranked list of the top 50 drug candidates for OA (Table S2). Among these, 14 high-priority candidates were selected based on ranking metrics and corroboration with existing literature (Table 2). Notably, several well-known OA treatments, such as glucosamine (rank 1) and ibuprofen (rank 16), were identified, validating the predictive accuracy of the platform. Other candidates, including celecoxib (rank 7) and indomethacin (rank 20), are established nonsteroidal anti-inflammatory drugs (NSAIDs) used to manage OA symptoms. Table 2Predicted drug candidates for osteoarthritis (OA) using the CANDO platform, corroborated through literature analysisDrug rankCount scoreSum scoreDifferenceDrug name151.7331.733Glucosamine251.7191.719Palmitic acid351.6701.670Zinc651.6431.643Angiotensin Ⅱ751.6401.640Celecoxib1051.5941.594Linoleic acid1151.5921.592Naproxen1651.5641.564Ibuprofen2051.5331.533Indomethacin2351.5251.525Oleic acid2951.5021.502Arbutin3251.4661.466Flurbiprofen3951.4291.429Tolfenamic acid4851.3821.382Polydatin Predicted drug candidates for osteoarthritis (OA) using the CANDO platform, corroborated through literature analysis In addition to known therapeutics, the analysis identified novel candidates with potential OA-modulating properties. For example, palmitic acid (rank 2) and linoleic acid (rank 10) have demonstrated anti-inflammatory and chondroprotective effects in preclinical studies. Zinc (rank 3), known for its role in cartilage repair, also emerged as a promising candidate. These findings highlight the utility of the CANDO platform in identifying both established and novel therapeutic agents for OA, offering new avenues for drug development and repurposing. Validation of hub genes expression in OA Notably, qRT-PCR validation confirmed differential expression patterns of the five hub genes in peripheral blood samples from OA patients. As shown in Fig. 7, ESR1, PTGS2, CCL2, and TRPV1 exhibited significantly higher expression levels, indicating their potential upregulation in OA. In contrast, FLNA displayed relatively lower expression levels in OA samples, suggesting possible downregulation or reduced transcriptional activity. These results are consistent with the network analysis findings, further supporting the involvement of these hub genes in the molecular mechanisms underlying BPA-induced OA pathogenesis. Experimental validation of hub gene expression in osteoarthritis. Quantitative real-time PCR (qRT-PCR) analysis was performed to determine the mRNA expression levels of ESR1, PTGS2, CCL2, FLNA, and TRPV1 in peripheral blood samples from osteoarthritis (OA) patients and healthy controls. Bars represent mean ± SD. *** P < 0.001 vs. control group Discussion In this study, we employed a comprehensive approach to identify 26 potential targets associated with BPA-induced OA by integrating data from ChEMBL, STITCH, GeneCards, OMIM, and the TTD. Using the STRING database and Cytoscape software, we constructed a protein-protein interaction (PPI) network and identified five hub targets—ESR1, PTGS2, CCL2, FLNA, and TRPV1—as central players in BPA-induced OA toxicity. GO analysis revealed significant associations with calcium ion transport and muscle contraction, while KEGG pathway enrichment highlighted key signaling pathways, including the IL-17 signaling pathway, estrogen signaling pathway, and TNF signaling pathway. Molecular docking studies demonstrated strong binding affinities between BPA and these core targets, suggesting a mechanistic link to OA pathogenesis. Furthermore, we used our CANDO platform predicted 14 potential OA treatments, such as glucosamine, ibuprofen, celecoxib, indomethacin, palmitic acid, and linoleic acid, all supported by existing literature evidence. Among the identified hub targets, PTGS2 and ESR1 have been extensively studied in the context of OA, providing valuable insights into their roles in disease progression. Previous studies have established PTGS2 as a critical inflammatory marker in OA, primarily due to its role in the production of prostanoids (TXA2, PGE2, PGD2, and PGI2) [52]. Elevated PTGS2 levels exacerbate inflammatory responses and oxidative stress [53], while its downregulation has been associated with reduced OA progression through decreased COX2 production [54]. However, Tu et al. reported lower PTGS2 expression in OA synovium [55], underscoring the complexity of the immune microenvironment in OA. These findings may explain the limited efficacy of COX2 inhibitors in advanced OA patients. Similarly, the estrogen receptor α (ERα), encoded by the ESR1 gene, plays a pivotal role in maintaining articular cartilage homeostasis via the estrogen signaling pathway. Estrogen protects articular cartilage by promoting chondrocyte autophagy, thereby mitigating OA progression [56]. Estrogen deficiency has been linked to the onset and progression of OA, affecting multiple joint tissues, including articular cartilage. In vivo studies using ERα knockout mice have demonstrated severe cartilage degradation, increased osteophyte formation, and joint capsule fibrosis [57], highlighting the critical role of ERα in OA development. These insights further support the significance of ESR1 as a key target in BPA-induced OA pathology. In addition to ESR1 and PTGS2, the hub targets CCL2, FLNA, and TRPV1 also play critical roles in OA pathogenesis, particularly in inflammation, structural integrity, and pain signaling. In osteoarthritis, elevated levels of inflammatory chemokines, such as CCL2 and CCL5, have been observed in both gene expression profiles and serum samples compared to healthy individuals [58–60]. CCL2, along with its high-affinity receptor CCR2, plays a central role in mediating pain associated with knee osteoarthritis, highlighting its significance in OA-related inflammation and nociception [58]. Filamins, a family of large actin-binding proteins, are critical for stabilizing three-dimensional actin networks and linking them to cellular membranes. This integration of structural and signaling functions is essential for processes such as cell locomotion. The filamin family comprises three homologous proteins: FLNA, FLNB, and FLNC. These proteins interact with a wide range of functionally diverse partners and are implicated in various genetic disorders affecting the skeleton, brain, and heart [61]. Despite their known roles in cellular architecture and signaling, the specific contribution of filamins, particularly FLNA, to OA pathogenesis remains poorly understood and warrants further investigation. Transient receptor potential vanilloid 1 (TRPV1), a nonspecific cation channel, is activated by thermal, mechanical, and chemical stimuli [62]. Its role in pain perception has been extensively studied across multiple disease models [63–65]. Notably, intra-articular administration of TRPV1 agonists has been shown to significantly alleviate pain in both OA animal models and human patients, suggesting its potential as a therapeutic target for OA-related pain management [66, 67]. These findings collectively underscore the complex interplay of inflammatory mediators, structural proteins, and pain-sensing mechanisms in OA, providing a foundation for further exploration of targeted therapeutic strategies. In addition to the roles of inflammatory chemokines, structural proteins, and pain-sensing mechanisms in OA, the cytokine IL-17 has emerged as a key regulator of autophagy, a process critical for maintaining cartilage homeostasis. However, the specific mechanisms by which IL-17 regulates autophagy in OA remain poorly characterized. A study observed significantly elevated IL-17 expression in an IL-1β-induced OA chondrocyte inflammation model, which exacerbated autophagy defects and promoted cartilage degeneration by mediating the PI3K/AKT/mTOR autophagy-related pathways [68]. Furthermore, stimulation of mouse osteoblasts with IL-17 was found to suppress autophagy levels by activating the JAK2/STAT3 signaling pathway and downregulating the expression of key autophagy markers such as Beclin 1, LC3, and Atg7 [69]. In summary, IL-17 contributes to OA progression primarily by inhibiting autophagy, a process closely linked to the PI3K/AKT/mTOR and JAK2/STAT3 signaling pathways. Targeting these pathways or blocking IL-17 activity may represent promising therapeutic strategies for OA. However, further research is needed to clarify whether IL-17 acts as an independent upstream regulator of autophagy and to fully elucidate its role in OA pathogenesis. Among the predicted compounds, palmitic acid and linoleic acid have garnered attention for their potential anti-inflammatory and chondroprotective properties, which may offer new therapeutic avenues for OA management. Palmitic acid, a saturated fatty acid, has been shown to modulate inflammatory responses by regulating the production of pro-inflammatory cytokines and oxidative stress markers. Studies suggest that palmitic acid can inhibit the activation of nuclear factor-kappa B (NF-κB), a key regulator of inflammation, thereby reducing cartilage degradation in OA [70, 71]. Additionally, palmitic acid has been reported to enhance the expression of antioxidant enzymes, which may protect chondrocytes from oxidative damage [72]. Linoleic acid, an essential polyunsaturated fatty acid, has also demonstrated promising effects in OA. It is a precursor to anti-inflammatory mediators such as prostaglandin E1 (PGE1) and lipoxins, which can counteract the pro-inflammatory effects of arachidonic acid derivatives [73]. Furthermore, linoleic acid has been shown to suppress the expression of matrix metalloproteinases (MMPs), enzymes that degrade extracellular matrix components in cartilage, thereby preserving cartilage integrity [74]. These properties make linoleic acid a potential candidate for mitigating OA progression. The inclusion of palmitic acid and linoleic acid in our predicted drug candidates highlights the potential of targeting lipid metabolism and inflammatory pathways in OA treatment. Further experimental validation is needed to confirm their efficacy and elucidate their mechanisms of action in OA models. Conclusions This study employed an integrated approach combining network toxicology and molecular docking to investigate the molecular mechanisms underlying BPA-induced OA. Through systematic analysis, we identified 26 potential targets and five hub—ESR1, PTGS2, CCL2, FLNA, and TRPV1—that are central to OA pathogenesis, with enriched pathways including IL-17 signaling, estrogen signaling, and TNF signaling. Using the CANDO platform, we predicted 14 potential therapeutic candidates, such as glucosamine, ibuprofen, and linoleic acid. These findings provide critical insights into the molecular interplay between BPA exposure and OA development, highlighting potential therapeutic targets and drug candidates. However, the computational predictions in this study require further experimental validation to confirm the molecular mechanisms of BPA-induced OA and the therapeutic efficacy of the identified candidates. By bridging predictive modeling with biological validation, this study provides a foundation for future research aimed at developing targeted interventions to mitigate the adverse effects of environmental toxins on joint health. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors would like to acknowledge the Center for Computational Research (CCR) at University at Buffalo for computational support. We would also like to thank all members of the Samudrala Computational Biology Group.The authors sincerely thank related investigators for sharing the statistics included in this study. Abbreviations Bisphenol A Osteoarthritis Therapeutic Target Database Protein-protein interaction Gene ontology Kyoto Encyclopedia of Genes and Genomes Computational Analysis of Novel Drug Opportunities Endocrine-disrupting chemicals False discovery rate Maximal Clique Centrality Biological processes Molecular functions cellular components Protein Data Bank Quantitative real-time PCR Author contributions SM. X. helped conceive the research project and pipelines, conducted all analyses, and drafted the manuscript. LP. J., Z.Z., X.L., W.W., HL. W., and Z.F. helped data analyses. R.S. and ZR. T. conceived of the research design, approach, and methods, supervised the overall project, and edited the manuscript. All authors have read and agreed to the published version of the manuscript. Funding This work was supported by the Key Project Funded by Hunan Provincial Health Commission [grant number A202313018918] and the Natural Science Foundation of Hunan Province [grant numbers 2024JJ7630]. Data availability Data is provided within the manuscript or supplementary information files. Declarations Ethical approval Approved by the Institutional Review Board of Xiangya Hospital, Central South University (No: 2025101788). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Ram Samudrala, Email: ram@compbio.org. Zhirong Tan, Email: 803119@csu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data is provided within the manuscript or supplementary information files.