Potential Therapeutic Mechanism of Radix Angelicae Biseratae and Dipsaci Radix Herb Pair against Osteoarthritis: Based on Network Pharmacology and Molecular Docking

Background A major contributor to older disability is osteoarthritis. Radix Angelicae Biseratae (known as Duhuo in China, DH, the dried rhizome of Angelica pubescens) and Dipsaci Radix (known as Xuduan in China, XD, the dried rhizome of Dipsacus asper Wall) herb pair (DXHP) is widely used to treat osteoarthritis, but the underlying molecular mechanisms still have not been revealed. This research aimed to illustrate the therapeutic mechanism of DXHP against osteoarthritis through the techniques of network pharmacology and molecular docking. Methods Gene targets for osteoarthritis and active ingredients for DXHP were screened based on the pharmacology public database and the gene-disease target database. The software program Cytoscape was used to visualize the active chemical target-disease gene network. The STRING biological information website was used to investigate protein interactions. On the Metascape bioinformatics website, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were carried out. The molecular docking of the important chemicals and primary targets identified by the aforementioned screening was performed using Autodock software. Results Twenty-six active substances from the DXHP that had strong connections to 138 osteoarthritis-related targets were screened out. According to network analysis, TNF, GAPDH, IL-6, AKT-1, IL-1B, and VEGFA are prospective therapeutic targets, while osthole, cauloside A, ammidin, angelicone, beta-sitosterol, and asperosaponin VI may be significant active components. 1705 biological processes (BP), 155 molecular functions (MF), and 89 cellular components (CC) were identified by GO analysis. KEGG analysis indicated that IL-17, NF-kappa B, HIF-1, MAPK, and AGE-RAGE signaling pathways are potentially involved. Molecular docking showed that cauloside A, osthole, and β-sitosterol have excellent binding activity with main targets. Conclusions This study comprehensively illuminated the active ingredients, potential targets, primary pharmacological effects, and relevant mechanisms of the DXHP in the treatment of OA. These findings provide fresh thoughts into the therapeutic mechanisms of the main active ingredients of DXHP and provide a reference for further exploration and clinical applications of DXHP.


Introduction
Destruction of cartilage, remodeling of the subchondral bone, the development of osteophytes, and synovial infammation are all symptoms of osteoarthritis (OA), a degenerative joint disease [1][2][3]. OA is directly associated to heredity, age, obesity, injury, and chronic infammation [4][5][6][7]. As populations of aging and obese individuals have grown in size, the prevalence of OA has also increased and is now considered a major public health problem worldwide [8]. According to recent studies, the global prevalence of OA grew by 48% between 1990 and 2019 [3,9], and that OA afects more than 500 million persons, occurring most commonly in postmenopausal women over age 50 [10]. Some guidelines indicate that nonsteroidal antiinfammatory drugs (NSAIDs) are the most crucial medications for OA [11,12]. Tis class of drugs achieves its anti-infammatory and analgesic efects by inhibiting cyclooxygenase (COX), a key enzyme in the metabolism of arachidonic acid (AA) [13]. However, the use of COX inhibitors alone is not ideal. Studies have shown that AA can also be metabolized by lipoxygenase (LOX) to produce infammatory substances such as leukotrienes [14]. Terefore, single introduction of COX inhibitors cannot limit production of infammatory mediators but instead stimulates and increases the release of infammatory mediators in the LOX pathway, resulting in adverse reactions [15]. Te current mainstream treatment options for OA have limited efects on delaying the development of joint infammation. In the long term, such drugs cause serious damage to the functions of the digestive tract, liver, and kidneys [16,17]. In addition, long-term use accelerates the progression of arthritis [18]. Terefore, research into alternative therapies that can contribute to the development of new drugs could improve treatment options for OA.
Traditional Chinese medicine (TCM), as a supplementary therapy, has accumulated rich theoretical knowledge that can potentially help advance clinical prevention and treatment strategies for OA [19]. Te earliest research on OA in TCM can be traced back thousands of years to the ancient text "Huangdi Neijing [20]." In recent years, TCM has been suggested to have anti-infammatory, anti-apoptotic, antioxidative, anti-metabolic, and proliferative efects in the treatment of OA diseases [21]. Herb pair refers to the combination of drugs that can be used simultaneously in TCM clinical to enhance efcacy or reduce adverse reactions. Compared with TCM formulas, herb pairs have a more defned synergistic mechanism [22]. Radix Angelicae Biseratae (known as Duhuo (DH), the dried rhizome of Angelica pubescens) and Dipsaci Radix (known as Xuduan (XD), the dried rhizome of Dipsacus asper Wall) are a classic herbal drug combination that has been widely used in TCM. Clinical statistical studies have shown that DXHP is one of the most frequently used Chinese medicine pairs in the treatment of OA [23]. In modern pharmacological research, the active ingredients in DH are linked to antioxidant, antiinfammatory, and analgesic efects [24], while the active components of XD purportedly protect against bone loss, promote cartilage formation, and improve bone metabolism [25]. In addition, the pharmacological compatibility mechanism of DXHP is similarly refected in the TCM theory. DXHP originated from Duhuo Xuduan decoction in the ancient Chinese medical book "Waitai Miyao." Chinese Medicine Pharmacopoeia points out that DH helps relieve symptoms of rheumatism, particularly symptoms associated with dampness, swelling, and pain. XD supposedly tonifes the liver and kidneys, nourishes the muscles and bones, and repair fracture [26]. Hence, it is highly essential to research the pharmacological mechanism of DXHP therapy as a possible means of improving the treatment of OA. However, the molecular mechanism of DXHP in the treatment of OA remains unclear.
In recent years, network pharmacology, which combines multidisciplinary knowledge and methods, has the characteristics of being systematic and holistic. Molecular docking can calculate the binding energies between ligands and receptors and predict reasonable binding mode and has been used to explore the molecular process of drug active ingredients acting on the human body. Terefore, the aim of this research is to clarify the molecular underpinnings behind the efects of DXHP on OA using network pharmacology and molecular docking technologies. Te specifc technical route adopted in this study is shown in Figure 1.

Acquisition of Bioactive Compounds of DXHP.
To obtain more comprehensive compositional data, the active ingredients of DH and XD were retrieved from three herbal pharmacological databases, including the TCMSP data platform (https://tcmsp-e.com/) [27], the ETCM data platform (https://www.tcmip.cn/ETCM/index.html) [28], and the SymMap data platform (https://www.symmap.org/) [29]. According to previous studies and recommended standards [30], oral bioavailability (OB) ≥30% is regarded as having an excellent absorption. As the selection criteria for "drug-like" compounds in traditional Chinese medicinal materials, DL ≥0.18 is thought to be appropriate for drug design. Terefore, OB and DL criteria were chosen to investigate compounds that meet basic pharmacokinetic criteria for efective drug development [31]. Finally, we searched PubMed and CNKI databases for compounds that did not meet the criteria but had been proved to have signifcant therapeutic efects on OA.

Target Prediction for Compounds.
Te active ingredients obtained after the aforementioned screening were queried by PubChem (database https://pubchem.ncbi.nlm.nih.gov/) for SMILES codes of their potential ingredients, which were entered into Swiss Target Prediction (https://www. swisstargetprediction.ch/) for further prediction of active ingredient targets, setting the screening parameter criteria to the probability value >0.1. To normalize gene nomenclature and species and avoid over-annotation of homologous proteins, these target genes were transformed into matching gene symbols that corresponded to the "Homo sapiens" species using the UniProt information system (https://www. uniprot.org/) [32].

Collecting Herb-Disease Genes.
We retrieved OA-related genes from the GEO data platform (https://www.ncbi.nlm. nih.gov/geo/) [33] database using "osteoarthritis" as a keyword search. Te datasets were processed using the robust multiarray average algorithm for background correction and matrix data normalization. Te screening conditions for signifcantly diferential genes were P < 0.05 and |log FC| >1.5. Te diferentially expressed genes for OA were obtained. Ten, we search the OA-related genes in the following databases: GeneCards (https://www.genecards.org) [34], the TTD information system (https://db.idrblab.net/ ttd) [35], and the OMIM information system (https://www. omim.org) [36]. For these searches, the species source was limited to "Homo sapiens." Te above-obtained genes were merged, and the duplicated genes were deleted to obtain the defned disease targets of osteoarthritis. Subsequently, the OA-related targets were matched with the active compound targets of DXHP to obtain the target of DXHP in the treatment of OA. Te Venn diagram was drawn using the Venny2.1 plotter (https://bioinfogp.cnb.csic.es/tools/venny/ index.html).

Protein-Protein Interaction (PPI) Network.
Based on the STRING Bioinformatics system (https://www.string-db.org/), a protein-protein interaction (PPI) network was crafted for the common targets of compounds and diseases [37]. Te choice of species was human, and the confdence level was set to >0.4. Others parameters keep the default settings. To identify the primary target genes for OA therapy that have a strong correlation with DXHP, the topological properties of common targets were analyzed using the CytoNCA topology analyzer (A plugin for Cytoscape). Te network analyzer tool was applied for topology analysis, referring to three parameters, namely the degree centrality value (DC), the betweenness centrality value (BC), and the closeness centrality value (CC). Te greater the DC of the node, the higher the importance of the node in a PPI network. Targets with all three parameters above average were selected and ordered by the degree centrality value. Evidence-Based Complementary and Alternative Medicine

Protein-Protein Interaction Enrichment MCODE
Analysis. MCODE, as a clustering algorithm, helps to capture modules with high-quality biological processes in a large PPI spherical network and helps to discover a subset of targets that are closely related to that functional module. MCODE subcluster enrichment analysis was performed on the PPI network. PPI enrichment analysis used the following databases: STRING, InWebIM, OmniPath, and BioGRID. Only the STRING Bioinformatics system (Physical Score >0.132) and BioGRID data are used in the Physical Interaction module. A subset of the proteins in the resultant network physically interacts with at least one other item on the list. Te MCODE methodology was used to fnd components of highly linked networks in subnetworks with 3 to 500 proteins.

Functional Annotation from Gene Ontology (GO) and Pathway Enrichment Analysis from the Kyoto Encyclopedia of Genes and Genomes (KEGG).
Te Metascape database (https://metascape.org/) is a powerful broad-coverage and fast-updating gene function annotation analysis tool that can analyze a large number of gene or protein functions. In order to efectively research the biological ontology of DXHP in the regulation of OA and to clarify the biological process of each core target protein and its function in signaling pathway transduction, GO functional annotation and KEGG pathway enrichment analysis were performed based on the Metascape biological system. Te common targets of DXHP and OA obtained in the above screening were imported into the target gene list, with confned to just human species and correction of all target genes to their recognized gene symbols. Multiple comparisons were performed using Benjamini-Hochberg's FDR correction to avoid false positives. Bonferroni-corrected P values <0.01 for GO and KEGG terms were considered signifcant. Te minimal count for the KEGG analysis was 3, and the enrichment factor was more than 1.5. Finally, the top 20 items were picked, and the annotated chart was formed on the Bioinformatics (https:// www.bioinformatics.com/) platform for visualization.

Molecular Docking Verifcation of Compound Target.
In network pharmacology, molecular docking is a critical tool for verifying compound-target interactions. It works by combining proteins of known targets obtained from network pharmacology with small compounds of active ingredients in natural drugs and then evaluating the strength and activity of the binding [38]. Te mol2 format fles of the main active components of DXHP were searched and obtained from the PubChem database in advance. Next, the selected protein's best resolution 3D structure was acquired in the PDB database (https://www.rcsb.org/) and their PDBID fles were downloaded. For these 3D structures, the species source was limited to "Homo sapiens" and contains the crystal structure of the complete pocket. MGLtool 1.5.7 was used to process the protein by adding hydrogen, calculating the charge, merging the nonpolar hydrogen, and saved as a PDBQT format fle as a docking ligand. Gridbox coordinates and docking box sizes were set, and molecular docking was performed using Autodock Vina1.1.2. Te lowest binding energy score conformation was selected, with lower binding energy scores indicating better docking activity and strength. Te molecular docking structures were demonstrated using Pymol 2.3.

Screening of DXHP Active Compounds.
We obtained 106 active components of DH and 63 active components of XD. After screening by OB and DL standards, 20 active ingredients with good ADME properties were obtained for the subsequent study. In addition to this, DH and XD were found to contain some pharmacologically active ingredients, which were excluded because their OB and DL values were less than the screening criteria. Tus, according to previous literature reports, we included a total of six active ingredients such as osthole, columbianadin, umbelliferone, asperosaponin VI, ursolic acid, and loganin [39,40]. A total of 26 DXHP active compounds and their corresponding 463 targets were screened (Table 1). Te results show that a single compound can regulate multiple targets, indicating that DXHP has multicomponent and multitarget components. Te active compounds of DXHP and corresponding genes are shown in Figure 2.

Drug-Disease Core Target Acquisition.
In the GEO database, we identifed the dataset that met the criteria as GSE169077. Te chip contained six osteoarthritis cartilage tissue samples as the experimental group and fve normal knee cartilage tissue samples as the control group. Te results of diferential expression gene analysis of osteoarthritis showed that 792 genes were found to be diferentially expressed. Te heatmap and Limma package in the R language were used to draw the dataset volcano map of differential genes ( Figure 3). Te GeneCards database yielded 3647 OA-related genes in total, and 885 OA-related genes were obtained using a relevance score >1 as the screening criterion. Te TTD database attained 33 OA-related genes in total, whereas the OMIM database attained 30 OA-related genes. Te above gene sets were combined, and duplicate values were removed to obtain a total of 1778 OA-related genes. Te screened drug targets and OA-related genes were input into Venny 2.1.0 plotter, a Venn fgure was constructed, and a total of 138 common targets were obtained ( Figure 4).

PPI Network Construction and Centiscape Analysis.
Te herb-disease common targets were imported into the STRING Bioinformatics system. Species option was selected as "Homo sapiens," the PPI network was constructed (Figure 5), and the CSV fle was entered into Cytoscape software 3.9.1 for network topology analysis using the network Stats tool and the centiscape plugin for the construction of the three parameter correlations such as DC, BC, and CC. 138 nodes and 1599 edges made up the network, which had an average connectedness of 23.70. A total of 25 core targets were selected through the centiscape plugin by selecting  Evidence-Based Complementary and Alternative Medicine genes with greater than average DC, BC, and CC parameters as key targets and ranking them by the degree centrality value. Te 10 highest ranked core targets are listed in Table 2. DH and XD have 155 overlapping targets, 69 of which are associated with OA ( Figure 6).

GO Functional
Annotation. GO functional annotation analysis showed that the enrichment results included 1705 biological processes (BPs), 155 molecular functions (MFs), and 89 cellular components (CCs). Most GO annotations of BP were contained to regulation of infammatory response, response to hormone, infammatory response, positive regulation of locomotion, cellular response to lipid, cellular response to organic cyclic compound, positive regulation of cell death, protein phosphorylation, positive regulation of MAPK cascade, and cell activation. MF annotation was mainly involved in protein kinase activity, endopeptidase activity, transcription factor binding, nuclear receptor activity, cytokine receptor binding, protease binding, protein homodimerization activity, transmembrane receptor protein kinase activity, oxidoreductase activity, and phosphatase binding. Te majority of CC annotations were included to the extracellular matrix, membrane raft, receptor complex, cell body, organelle outer membrane, fcolin-1-rich granule lumen, endolysosome, protein kinase complex, side of the membrane, and endoplasmic reticulum lumen. Based on their Q-value (the Q-value was used for multiple testing, it was calculated using the Benjamini-Hochberg procedure, and higher Q-values showed the greater GO term enrichment), the top 10 terms of BP, CC, and MF were rated. Te terms are presented in Figure 7.  Figure 8. Te fgures of the two infammation-related signaling pathways with the highest number of core genes enriched are shown in Figures 9 and 10.

PPI Enrichment MCODE Analysis.
Core genes were identifed through network construction and MCODE analysis using the complete datasets for the independent enrichment analysis of gene clusters. Te enrichment analysis of biological processes was used for each MCODE component ( Figure 11 and Table 3). MCODE analysis showed that the core genes of AKT1, IL1B, TLR9, VEGFA, MMP2, PTGS1,        (Table 4).

Molecular Docking Verifcation.
To further validate the reliability of the binding of key targets and components screened by the network analysis, molecular docking verifcation of some key targets and important active components was carried out using Autodock Vina1.2.0 software. It is largely accepted that the steadier the binding structure, the smaller the binding energy of the receptor-ligand docking    Evidence-Based Complementary and Alternative Medicine and therefore the higher the likelihood of interaction between the two, with a binding energy < −5.0 kcal/mol as the screening criterion. Binding energies < −5.0 kcal/mol indicate potential activity, and docking with binding energies < −7.0 kcal/mol is extremely stable [41]. Te top six active ingredients in the component-target network were selected for molecular docking with the top six targets in the PPI network. Te results showed that about 83% of targets and active components exhibited binding ability and 16% exhibited extremely strong binding ability. Tese molecular docking fndings align with earlier network screening conclusions, which indirectly validate the treatment ability of DXHP on OA and demonstrate the reliability of network pharmacology applied to this study. Te docking results of binding afnity and detailed compound-target interactions are presented in Table 5 and Figure 12.

Discussion
OA has a high prevalence in the elderly population, and modern medical treatment options are primarily symptomatic, with no options available to curb disease progression [42]. Te advantages of DXHP in relieving OA symptoms as well as delaying OA disease progression have been demonstrated in previous studies [43], but its mechanism of action remains incompletely elucidated. Terefore, we investigated the mechanism of action of DSHP on OA in a more systematic way using a network pharmacology and the molecular docking approach. According to this investigation, DXHP's primary therapeutic ingredients for OA include substances such as osthole, asperosaponin VI, angelicone, beta-sitosterol, ammidin, and cauloside A, among which β-sitosterol is the overlapping component of DH and XD. In recent years, the pharmacological efects of asperosaponin VI in antiinfammatory analgesia, prevention of osteoporosis, neuroprotection, and anti-apoptosis have attracted the attention of many scholars [44]. Many in vivo and cell experiments have shown that asperosaponin VI can phosphorylate ERK/2 protein to promote the expression of osteogenic genes such as ALP, OCN, COL1 and RUNX2, which is closely related to the PI3K/Akt signaling pathway and the MAPK signaling pathway [45]. In addition, asperosaponin VI could signifcantly reduce MDA, TNF-α, IL6, and IL10 by signifcantly inhibiting oxidative stress and infammatory response in tissues [46]. Osthole, one of the main active components of DH, is thought to improve bone metabolism and promote osteoblast activation [47]. Recent studies on the antiinfammatory mechanism of osthole have found that   Evidence-Based Complementary and Alternative Medicine osthole can inhibit the formation and resorption activity of osteoclasts by inhibiting the activation of NF-κB and NFATc1 and reducing the expression of osteoclast-specifc genes such as CTSK, MMP-9, TRAP, integrin β3, C-SRC, and NFATc1 [48]. β-sitosterol has a strong down-regulate efect on proinfammatory factors, such as IL-1β, IL-6, and TNF-α and quenches ROS produced by the human body through antioxidant activity. It can also alleviate infammation through eosinophil percolation [49]. Chen et al. assessed the knee joints of OA rabbits by morphological and histological methods and found that β-sitosterol could signifcantly inhibit the secretion of matrix metalloproteinases (MMPs) and inhibit the degradation of cartilage [50]. Te observations of the coincide target PPI analysis demonstrated that the core targets of DXHP in the healing of OA encompassed multiple targets such as TNF, GAPDH, IL6, AKT1, IL1B, VEGFA, CASP3, STAT3, MMP9, and PTGS2. TNF-α has been demonstrated in many diseases, including osteoarthritis, autoimmune diseases, ankylosing spondylitis, insulin resistance, psoriasis, nephropathy, and cancer [51]. TNF-α is associated with many cytokines, and it has been found that TNF-α implicated in angiogenesis by synergistic induction of VEGF production with IL-1β and IL-6 [52]. Nie et al. found that TNF-α could regulate the harmony between Treg cells and TH17 and TH1 in the joint synovium through FOXP3 dephosphorylation [51]. When cells are stimulated with NO, GAPDH will be nitrosylated, bound to E3 ubiquitin ligase Siah1, and undergo nuclear translocation and apoptosis [53]. Te results of molecular docking suggest that DXHP may achieve the purpose of treating OA by binding to GAPDH. Studies have shown that [54] knockdown of the IL-6 gene in an OA rat model can lead to inhibition of MMP13 expression and secretion, which may be related to the inhibition of c-fos/ap-1-mediated infammatory stimulation in OA chondrocytes. IL-6 can also cause an increase in the expression of MMP9, prevents the formation of type II collagen and proteoglycans, accelerates the degradation of extracellular matrix, and infuences bone resorption by activating osteoclasts [55]. Shahine and Elhadidi found that the expression amount of the IL-1β gene in OA organisms was positively correlated with the pain index [56]. Correspondingly, high expression of IL-1β was detected in the synovial membrane and fuid of OA organisms and was positively correlated with OA disease [57]. Apoptotic chondrocytes are essential for the progression of OA. AKT1 is a key down-regulate gene kinase involved in the PI3K pathway. Researchers have demonstrated that the PI3K/AKT pathway that involves AKT1 prevents chondrocyte apoptosis [58]. TP53 inhibits DNA replication, induces apoptosis, and accelerates cartilage degradation [59]. In addition, researchers have demonstrated that the VEGFA is closely associated with many pathological responses, such as osteophyte formation and cartilage degeneration in OA [60]. Te expression of VEGFA stimulates the division of vascular endothelial cells to promote angiogenesis, accelerate the exchange of nutrients in the knee joint and the metabolism of infammatory products, and promote the growth of cartilage synovial cells.
Te KEGG enrichment analysis showed that the main enriched infammation-related signaling pathways included AGE-RAGE, IL-17, NF-kappa B, MAPK, TLR, HIF-1, PI3K-Akt, and the VEGF signaling pathway. Te other signaling pathways enriched included pathways in cancer, lipid and atherosclerosis, hepatitis B, prostate cancer, microRNAs in cancer, apoptosis, fuid shear stress and atherosclerosis, chagas disease, human cytomegalovirus infection, FoxO signaling pathway, and osteoclast diferentiation. Te MCODE analysis revealed that the regulatory response to infammation was the most signifcant biological process in treating OA by DXHP. Te NF-κb pathway is closely related to cartilage destruction in OA and targeted therapy of OA [61]. Previous research has showed that the secretion level of NF-κb in joint synovial fuid and peripheral blood of OA patients is increased, and the degree of up-regulation of the NF-κb pathway is positively correlated with the degree of cartilage erosion and destruction [62]. After phosphorylation, IκBα dissociates from NF-κB, and NF-κB is activated and enters the nucleus, promoting the synthesis and secretion of TNF-α and IL-6 [63,64] and eventually leading to the degeneration of articular cartilage [65,66]. In addition, IL-17 can trigger the release of chemokines, cytokines, antimicrobial peptides, and matrix metalloproteinases from mesenchymal and bone marrow cells [67]. Bai et al. [68] found that the level of IL-17 in the serum of OA patients increased and was positively correlated with the severity of OA. Due to the lack of capillaries in articular cartilage, the cartilage microenvironment is essentially a hypoxic environment. Te main hypoxia-inducible factor (HIF) of articular chondrocytes is a transcription factor, which is also the main mediator of homeostatic response that enables cells to survive under hypoxic conditions [69]. In OA, the expression of HIF-1α decreases [70], and the loss of HIF-1α can up-regulate the expression of MMP13, degrade col2A1 and ACAN, promote chondrocyte degradation, and promote the development of OA [71]. Te advanced glycation end product receptor (RAGE) is secreted in a variety of cells, including macrophages and mast cells [72]. AGE is a ligand of RAGE and is the end product of glycosylation of proteins and sugar [73]. RAGE is essential for the induction of several infammatory genes as well as important signaling pathways linked to proinfammatory responses. Although AGE induces infammation by exciting NF-κB and MAPK in a variety of cells including osteocytes [74], it can also upregulate the expression of PGE2 and NO via the MAPK pathway and induces the chondrocytes infammatory [75].
As an herb pair, DH and XD have many overlapping targets and active ingredients, and the compatibility mechanism of DXHP in treating OA may be related to this. β-sitosterol, as its overlapping active ingredient, is associated with anti-infammatory and antioxidant efects, immune regulation, and bone metabolic balance [76]. Furthermore, the aforementioned KEGG and MCODE analyses revealed that the overlapping targets such as IL6, IL-1β, AKT1, VEGFA, MMP13, and STAT3 are vital links in infammatory signalling pathways such as AGE-RAGE, IL-17, MAPK, and NF-kappa B and are extensively involved in the biological processes of regulating the infammatory response in OA. 12 Evidence-Based Complementary and Alternative Medicine Tis fnding may help us to better understand the compatibility mechanism of DXHP from a molecular perspective. Molecular docking tests were carried to further confrm the molecular mechanism of DXHP in the treatment of OA. Te results revealed that osthole, angelicone, cauloside A, and β-sitosterol have excellent binding activity with multiple key targets, such as TNF, GAPDH, IL-6, AKT1, IL-1β, and VEGFA. Tis also explains the therapeutic mechanism of DXHP from another perspective. Tese results ofer insight for the subsequent application of network pharmacology methods to improve the efciency of natural medicine ingredient development, which could facilitate the development of new high-efciency, low-toxicity, multi-target OA drugs capable of improving symptoms and delaying disease progression.
Compared to recently published network pharmacology studies of similar diseases [77], we have expanded the source of the drug and disease database and introduced diferential genes from the GEO microarray, making the source more comprehensive. In order to capture modules with highquality biological processes in large PPI networks, we apply the MCODE clustering algorithm, which allows for a further in-depth interpretation of the target network. In addition, we have investigated the overlapping active ingredients and targets of this herb pair to elucidate the compatibility mechanisms of DXHP from a new perspective.
Despite the study's advantages, several limitations should be mentioned. Firstly, the sources of component and target data are scattered and vary widely between databases, resulting in insufciently accurate data sources [78]. Secondly, the validation of the fndings is only based on computer simulations of molecular docking, which has limitations in terms of convincingness [79]. Terefore, the key pathways and targets identifed in this study need to be validated in more in-depth in vivo and in vitro experiments.

Conclusion
In this study, we found that DXHP is a valuable TCM herb pair for treating patients with OA through multiple components, targets, and pathways. Its pharmacological mechanism could be through the AGE-RAGE, IL-17, NF-kappa B, MAPK, TLR, HIF-1, PI3K-Akt, and VEGF signaling pathways to alleviate OA. Hopefully, our research may provide a scientifc basis for the prodrug discovery of its natural ingredients and the identifcation of therapeutic targets in the future.

Data Availability
Te original contributions presented in the study are included within the article/supplementary material. Te data supporting the fndings of the current study are available from the corresponding authors upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.