Unveiling Anti-atherosclerotic Targets of Perilla frutescens through a Multi-scale Computational Framework Integrating Network Pharmacology, Single-cell Analysis, Machine Learning, and Molecular Dynamics

Unveiling Anti-atherosclerotic Targets of Perilla frutescens through a Multi-scale Computational Framework Integrating Network Pharmacology, Single-cell Analysis, Machine Learning, and Molecular Dynamics | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Unveiling Anti-atherosclerotic Targets of Perilla frutescens through a Multi-scale Computational Framework Integrating Network Pharmacology, Single-cell Analysis, Machine Learning, and Molecular Dynamics chenchen yang, Jianrong Xing, Mengzhu Wang, Wanyi Zhou, Ying Yang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9174552/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Despite the widespread implementation of lipid-lowering therapy, the persistence of residual inflammatory risk, driven by immunometabolic network dysregulation, remains a cardinal therapeutic challenge in atherosclerosis (AS) management. While Perilla frutescens exhibits well-documented anti-inflammatory properties, the precise molecular targeting within the atherosclerotic plaque microenvironment and the regulatory mechanisms governing intercellular communication networks remain poorly elucidated. We established a multi-scale integrative computational framework synergizing network pharmacology, human atherosclerotic plaque single-cell transcriptomic (scRNA-seq) profiling, and ensemble machine learning algorithms (LASSO and random forest) for systematic identification of robust therapeutic targets. Molecular dynamics simulations validated the binding affinity and thermodynamic stability of drug–target complexes. We analyzed the cellular heterogeneity lineage of plaques were identified and a core feature set of 10 genes were identified which specifically mapped the differentiation trajectory of macrophages to foam cells. External validation in an independent cohort demonstrated superior diagnostic performance of this signature (AUC = 0.996). Cellular communication network dissection revealed the foam cell-driven SPP1–ITGB1 signaling axis as a pivotal conduit orchestrating inflammatory crosstalk. Molecular docking demonstrated pronounced binding affinity between luteolin, the principal bioactive constituent of Perilla frutescens , and ITGB1 (binding energy: −8.9 kcal/mol). Molecular dynamics simulations further corroborated the efficacy of luteolin in stabilizing ITGB1 conformation via a "conformational-locking" mechanism (RMSD equilibration within 0.10–0.20 nm), thereby abrogating pathological cell adhesion signaling transduction. Collectively, this study provides a high-resolution molecular atlas of Perilla frutescens -mediated AS intervention, systematically elucidating the mechanistic paradigm whereby luteolin attenuates vascular inflammation through targeted disruption of the SPP1–ITGB1 communication axis. Atherosclerosis Perilla frutescens Single-cell RNA sequencing Machine learning Molecular dynamics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Atherosclerosis (AS) represents a chronic progressive pathological condition fundamentally characterized by arterial intimal lipid accumulation and maladaptive immunological responses. Despite the demonstrated efficacy of lipid-lowering pharmacotherapy, particularly statins and proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors, substantial residual inflammatory risk (RIR) persists, perpetuating plaque progression and precipitating acute cardiovascular events (Di Muro et al., 2025 ). Emerging mechanistic investigations have revealed that AS is orchestrated by intricate immunometabolic networks, wherein dysregulated crosstalk between lipid metabolism and inflammatory signaling establishes a homeostatic pathological architecture exhibiting pronounced resistance to monotherapeutic interventions (Z. Wang et al., 2025 ). Terapeutic paradigms necessitate a strategic transition from linear pathway inhibition toward systemic network perturbation to enable synergistic modulation of multiple pivotal nodes within the atherosclerotic plaque microenvironment. Botanical therapeutics, by virtue of their multi-component synergistic architecture, constitute a promising reservoir of bioactive compounds for network-based modulation. Perilla frutescens frutescens seeds, a quintessential functional food with medicinal attributes, harbor an abundance of pleiotropic bioactive constituents encompassing α-linolenic acid, phytosterols, and flavonoids such as luteolin. While prior phenotypic investigations have established the therapeutic efficacy of Perilla frutescens in ameliorating dyslipidemia and suppressing vascular inflammation, the underlying molecular mechanisms remain incompletely elucidated (Wu et al., 2023 ). Specifically, the precise cellular targeting and mechanistic execution of vasculoprotective effects by these bioactive constituents within the heterogeneous cellular landscape of atherosclerotic plaques remain undefined. Conventional systems pharmacology predominantly relies on bulk tissue transcriptomic profiling, the inherent methodological constraints of which substantially impede mechanistic interrogation at cellular resolution. This aggregate approach fundamentally obscures the spatiotemporal heterogeneity intrinsic to atherosclerotic plaques, precluding effective discrimination between pathogenic cellular subsets (e.g., specific foam cell phenotypes) and bystander populations (Xiong et al., 2023 ). Moreover, plaque destabilization is orchestrated through intricate intercellular communication networks, including ligand–receptor interactions governing cell adhesion and migration, the architectural delineation of which remains intractable without single-cell resolution. Consequently, critical knowledge gaps persist regarding which cellular differentiation trajectories are targeted by Perilla frutescens and the mechanisms through which it disrupts pathological intercellular crosstalk driving AS progression. To address these knowledge deficits, we established an integrative computational framework synergizing single-cell RNA sequencing (scRNA-seq) with machine learning (ML) algorithms. We employed an ensemble feature selection strategy coupling least absolute shrinkage and selection operator (LASSO) regression with random forest (RF) algorithms to rigorously identify robust core targets from high-dimensional single-cell datasets, thereby mitigating transcriptomic noise and overfitting artifacts. Furthermore, molecular docking coupled with molecular dynamics (MD) simulations validated the thermodynamic stability and conformational dynamics of predicted drug–target complexes under quasi-physiological solvation conditions. Leveraging this strategy, we successfully deconvoluted the cellular heterogeneity landscape of human atherosclerotic plaques and identified a core genetic signature encompassing HIF1A, PPARG, and ITGB1. Critically, our investigation revealed high-affinity binding between luteolin, the principal bioactive constituent of Perilla frutescens , and integrin β1 (ITGB1), whereby conformational stabilization via a "conformational-locking" mechanism abrogates the SPP1–ITGB1 signaling axis, which serves as a pivotal conduit mediating foam cell adhesion and inflammatory activation. Collectively, these findings delineate a high-resolution molecular cartography of Perilla frutescens -mediated AS intervention, underscoring the translational potential of targeting cell adhesion receptors for mitigating residual vascular inflammation. 2. Materials and Methods 2.1. Screening of Bioactive Constituents and Target Prediction for Perilla frutescens Chemical constituents of Perilla frutescens were systematically retrieved from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database ( https://tcmsp-e.com ) using " Perilla frutescens " as the query keyword (Ru et al., 2014 ). Candidate bioactive constituents were filtered based on conventional absorption, distribution, metabolism, and excretion (ADME) criteria, with selection thresholds established at oral bioavailability (OB) ≥ 30% and drug-likeness (DL) ≥ 0.18. Redundant entries were eliminated, and compound identifiers were standardized to facilitate downstream target prediction analyses. Putative molecular targets of candidate constituents were predicted using SwissTargetPrediction, Similarity Ensemble Approach (SEA), and SuperPred, with species parameters restricted to Homo sapiens . Prediction outputs from these platforms were subsequently merged, deduplicated, and subjected to standardized mapping via the UniProt database ( https://www.uniprot.org ) for uniform conversion to gene symbol nomenclature. AS-associated disease genes were retrieved from GeneCards ( https://www.genecards.org ), Online Mendelian Inheritance in Man (OMIM; https://www.omim.org ), and DisGeNET (v7.0; https://www.disgenet.org ) using "Atherosclerosis" as the query term. Selection criteria comprised: relevance score ≥ 10 for GeneCards entries, gene-disease association score (Score_gda) ≥ 0.10 for DisGeNET entries, and explicitly annotated AS-associated genes from OMIM. Disease-associated genes retrieved from these three repositories were consolidated and deduplicated to establish a comprehensive disease gene set. Intersection analysis between the Perilla frutescens candidate target gene set and the AS disease gene set was subsequently performed to identify putative therapeutic targets. Set intersection operations and visualization were executed in R software (v4.3.2) using the ggVennDiagram package (v1.2.2) for Venn diagram construction. The resultant intersecting targets were further subjected to protein–protein interaction (PPI) network construction, functional enrichment analysis, and integrative prioritization with single-cell transcriptomic differential expression profiles. 2.2. PPI Network Construction, GO and KEGG Enrichment Analyses The PPI network of intersecting targets was constructed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (v11.5; https://string-db.org ) (Szklarczyk et al., 2021 ). Analytical parameters were configured as follows: species restriction to Homo sapiens and minimum interaction confidence threshold set to high confidence (combined score > 0.7). PPI outputs were subsequently imported into Cytoscape (v3.9.1) for network visualization and topological characterization, with isolated nodes excluded to generate a connected network for downstream analyses (Shannon et al., 2003 ). Functional enrichment analyses were conducted in the R environment utilizing the clusterProfiler package (v4.10.0), with gene annotation information sourced from the org.Hs.eg.db database (v3.18.0) (Wu et al., 2021 ). To enhance annotation consistency and analytical robustness, gene symbols were initially converted to Entrez Gene IDs, followed by Gene Ontology (GO) enrichment analysis across three ontological dimensions, including biological process (BP), cellular component (CC), and molecular function (MF), as well as Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Multiple hypothesis testing correction was implemented via the Benjamini–Hochberg (BH) procedure, with the significance threshold established at false discovery rate (FDR, p.adjust) < 0.05. Enrichment outputs were visualized as bubble plots to systematically delineate the core biological processes and pivotal signaling cascades engaged by Perilla frutescens putative therapeutic targets. 2.3. Single-Cell Transcriptomic Data Acquisition and Preprocessing The human atherosclerotic plaque single-cell transcriptomic dataset GSE159677 was retrieved from the Gene Expression Omnibus (GEO) repository, with extraction of publicly available gene–cell expression matrices and associated metadata files (Alsaigh et al., 2022 ). All computational analyses were executed in the R environment (v4.3.2), with single-cell data preprocessing and downstream analytical workflows primarily implemented using the Seurat package (v4.3.0). Seurat objects were constructed from expression matrices, followed by implementation of stringent quality control (QC) procedures to eliminate low-quality cells and mitigate potential technical noise artifacts (Hao et al., 2021 ). Subsequently, doublet artifacts were identified and excluded using DoubletFinder (v2.0.3). Parametric configurations were determined based on sample-specific cell counts and standardized protocols, with anticipated doublet rates established according to 10x Genomics empirical recommendations and optimal parameter combinations refined through pK value sweep optimization. QC-filtered data underwent normalization via the Seurat NormalizeData function, followed by identification of highly variable genes (HVGs) using the FindVariableFeatures function. Data scaling and standardization were subsequently performed using the ScaleData function, with regression of technical covariates (including sequencing depth and mitochondrial gene fraction) to mitigate batch-associated technical variation. Principal component analysis (PCA) was executed via the RunPCA function, with optimal dimensionality for downstream analyses determined through integrative assessment of ElbowPlot visualization, JackStraw permutation testing, and cumulative variance explained metrics. To mitigate potential confounding effects of batch variation on cellular clustering architecture, multi-sample integration was performed using the Seurat integration workflow, with subsequent construction of k-nearest neighbor (kNN) graphs and execution of clustering algorithms within the integrated low-dimensional embedding space. Cell type annotation was manually assigned based on canonical marker gene expression profiles, with cross-validation performed through visualization modalities including FeaturePlot and DotPlot representations. Upon completion of cell type annotation, differentially expressed genes (DEGs) were identified using the FindMarkers function (default: Wilcoxon rank-sum test), with multiple testing correction implemented via the Benjamini–Hochberg (BH) procedure. DEG selection criteria comprised FDR (p.adjust) < 0.05, with the resultant DEG signature subjected to integrative cross-validation against network pharmacology-derived candidate targets for identification of core therapeutic nodes. 2.4. Machine Learning-Based Feature Selection for Disease State Classification Feature selection was performed using two complementary machine learning algorithms, namely least absolute shrinkage and selection operator (LASSO) regression and random forest (RF). To circumvent pseudo-replication artifacts inherent to single-cell data structures, pre-modeling aggregation of single-cell expression profiles was performed using a pseudo-bulk strategy across "sample × cell type" dimensions, yielding gene expression matrices with biological samples as statistical units. Disease state labels were derived from clinical stratification metadata accompanying the GEO dataset. Training–validation partitioning was executed at the sample level, ensuring mutually exclusive sample allocation to preclude data leakage artifacts. LASSO regression modeling was implemented using the glmnet package (v4.1-8), with disease state configured as a binary outcome variable and model family specified as binomial. Following feature standardization, optimal regularization parameter λ was determined via 10-fold cross-validation, with lambda.1se adopted as the penalty coefficient. Genes exhibiting non-zero regression coefficients at the selected λ value constituted the LASSO-derived feature gene set. Random forest classification models were trained using the randomForest package (v4.7-1.1), with computation of variable importance metrics for individual features. Given potential selection bias inherent to Gini index-derived variable importance, permutation importance was adopted as the primary evaluation metric. Intersection analysis between the LASSO-derived feature set and the top 20 genes ranked by random forest variable importance yielded the core feature gene signature. In instances of insufficient intersection cardinality, the union of both gene sets was employed, with subsequent refinement through external validation and model performance assessment. Candidate feature genes were ranked in descending order of composite importance scores, with the top 20 genes prioritized for downstream analyses. 2.5. Biological Weighted Expression Score (WES) Validation To quantitatively assess the transcriptional activity of candidate genes within key cellular subsets, gene set module scores and weighted WES were computed within the Seurat analytical framework. Initially, module scores for candidate gene sets were calculated using the Seurat AddModuleScore function to quantify relative transcriptional intensity at single-cell resolution, with background gene set normalization correction. Building upon this foundation, a WES was formulated to integrate both expression abundance and detection frequency within specific cellular subsets: $$\:WES=Average\:Expression\:\times\:Detection\:Rate\:(Pct.\text{E}\text{x}\text{p})$$ Here, Average Expression denotes the mean normalized expression level within the target cellular subset (computed via the Seurat AverageExpression function from LogNormalize-standardized expression matrices), while Detection Rate represents the detection frequency (pct.exp, defined as the proportion of cells exhibiting expression values > 0). WES values were computed for each candidate gene across individual cellular subsets and ranked in descending order. Subsequently, top-ranked WES genes were subjected to intersection analysis with machine learning-derived feature signatures, with the top 10 WES-ranked intersecting genes designated as core therapeutic targets. 2.6. Spatial and Dynamic Expression Validation of Core Targets at Single-Cell Resolution To assess the spatial distribution architecture of core candidate targets across cellular subsets, gene expression was projected onto uniform manifold approximation and projection (UMAP) low-dimensional embeddings using the Seurat FeaturePlot function, with expression heterogeneity across cell types and subpopulations visualized via DotPlot and VlnPlot representations. Subsequently, myeloid lineage cells, encompassing monocytes, macrophages, and foam cell subsets, were subjected to pseudotemporal trajectory reconstruction analysis. Trajectory inference was performed using the Monocle3 package (v1.3.1) (Cao et al., 2019 ). Trajectory root cells were designated based on cellular subsets exhibiting elevated expression of early macrophage markers (e.g., LST1, FCER1G), with pseudotemporal ordering established via root cell specification through the order_cells() function. Expression dynamics of core targets along the pseudotemporal axis were visualized and subjected to trend-fitting analyses to characterize their transcriptional trajectories during macrophage-to-foam cell differentiation. 2.7. Clinical Diagnostic Performance Validation (ROC Analysis) To evaluate the clinical diagnostic performance of core candidate targets, an independent external validation dataset GSE100927 (n = 104) was retrieved from the GEO repository. Atherosclerotic pathological specimens and healthy controls were extracted based on original sample annotations to establish binary classification labels. For microarray platforms, sequential application of background correction, quantile normalization, and log₂ transformation was performed; in instances of multiple probe-to-gene mappings, probe expression values were aggregated to the gene level via median (or mean) summarization to construct gene expression matrices. Subsequently, expression profiles of core genes within the validation cohort were extracted for downstream model construction and performance assessment. Multivariable logistic regression modeling was performed based on core gene expression profiles, with receiver operating characteristic (ROC) curves generated using the pROC package (v1.18.5). Area under the curve (AUC) metrics and corresponding 95% confidence intervals (CIs) were computed via the DeLong method. To mitigate overfitting-associated performance inflation bias, 10-fold cross-validation was implemented within the external validation cohort, with computation of mean AUC values and assessment of model discriminatory capacity across clinical stratifications. 2.8. Molecular Docking Three-dimensional structures of core target proteins were retrieved from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB), with selection criteria comprising: Homo sapiens species restriction, crystallographic resolution ≤ 2.5 Å, and preferential selection of structures harboring co-crystallized ligands. In instances of missing residues or incomplete side-chain geometries, structural refinement was performed using PDBFixer (v1.9) for residue reconstruction and conformational optimization. Receptor proteins underwent processing in PyMOL (v2.5.5) for removal of redundant water molecules while preserving critical metal ions and cofactors, followed by polar hydrogen addition, Gasteiger partial charge assignment, and PDBQT format conversion using AutoDockTools (ADT, v1.5.7) (Eberhardt et al., 2021 ). Ligand molecular structures were retrieved from the PubChem database, with three-dimensional conformer generation and geometric energy minimization performed using Open Babel (v3.1.1). Molecular docking simulations were executed using AutoDock Vina (v1.2.5), with docking grid centers defined based on co-crystallized ligand coordinates or computationally predicted active site pocket topologies. Docking parameters were configured as follows: exhaustiveness = 32, num_modes = 10, energy_range = 3. Docking poses were ranked according to the Vina scoring function, with the lowest-energy conformation selected for protein–ligand interaction analysis and visualization in PyMOL. 2.9. MD Simulations MD simulations were executed using the GROMACS software package (v2024.3) (Abraham et al., 2015 ). Ligand partial charges were derived via restrained electrostatic potential (RESP/RESP2) fitting of ORCA-computed wavefunctions using the Multiwfn program, with force field topology files generated via the Sobtop utility based on the general AMBER force field (GAFF) (Lu & Chen, 2012 ). Protein parameterization employed the CHARMM36 force field, with explicit solvation maintained throughout via the TIP3P water model. Protein–ligand complexes were positioned within rhombic dodecahedral simulation boxes, with a minimum solute-to-box boundary distance of 1.2 nm. Following explicit solvation, Na⁺ and Cl⁻ counterions were introduced to achieve charge neutralization, with ionic strength adjusted to 0.15 M to recapitulate physiological salinity. Systems underwent initial energy minimization via the steepest descent algorithm. Subsequently, equilibration simulations were performed sequentially under NVT ensemble conditions (100 ps, velocity-rescaling temperature coupling) and NPT ensemble conditions (100 ps, Parrinello–Rahman pressure coupling), with positional restraints of 1000 kJ·mol⁻¹·nm⁻² applied to protein backbone and ligand heavy atoms. Production-phase simulations were conducted with removal of all positional restraints, employing a 2-fs integration timestep over 100 ns. All hydrogen-containing covalent bonds were constrained via the linear constraint solver (LINCS) algorithm, with long-range electrostatic interactions computed using the particle mesh Ewald (PME) method; van der Waals and short-range electrostatic cutoff radii were uniformly set to 1.0 nm, with neighbor list updates performed every 20 timesteps. Post-trajectory analyses encompassed temporal evolution of root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), and protein–ligand interfacial hydrogen bond occupancy, enabling systematic assessment of complex conformational stability and binding interface persistence. 2.10. Statistical Analysis Statistical analyses were performed in R. Intergroup comparisons were conducted using two-sample t-tests or Wilcoxon rank-sum tests, with multiple testing correction implemented via the Benjamini–Hochberg procedure (FDR threshold: p.adjust < 0.05). ROC analyses employed the pROC package for computation of AUC values and 95% confidence intervals via the DeLong method. Unless otherwise specified, all statistical tests were two-tailed, with P < 0.05 considered statistically significant. 3. Results 3.1. Screening of Bioactive Constituents and Identification of Putative Therapeutic Targets in Perilla frutescens A network pharmacology approach was employed to systematically dissect the pharmacodynamic material basis underlying the therapeutic efficacy of Perilla frutescens. 16 core bioactive constituents exhibiting favorable pharmacokinetic properties were identified from Perilla frutescens using ADME filtering criteria of oral bioavailability (OB) ≥ 30% and DL ≥ 0.18 within the TCMSP database framework. As illustrated in Table 1 , these constituents collectively exhibited elevated bioavailability potential and favorable drug-like structural attributes. Luteolin (MOL000006), a flavonoid constituent prioritized for downstream investigation, demonstrated balanced pharmacokinetic parameters indicative of favorable membrane permeability and intestinal absorption capacity. Phytosterol constituents including β-sitosterol (MOL000358) and stigmasterol (MOL000449) exhibited OB values exceeding 36%, with maximal values reaching 43.83%, underscoring the pivotal contribution of lipophilic compounds to Perilla frutescens pharmacological repertoire. Table 1 Physicochemical Properties and ADME Evaluation Metrics of Candidate Active Compounds in Perilla frutescens Mol ID Molecule Name MW AlogP OB (%) DL MOL000006 luteolin 286.25 2.07 36.16 0.25 MOL000358 beta-sitosterol 414.79 8.08 36.91 0.75 MOL000449 Stigmasterol 412.77 7.64 43.83 0.76 MOL000953 CLR 386.73 7.38 37.87 0.68 MOL001439 arachidonic acid 304.52 6.41 45.57 0.2 MOL002773 beta-carotene 536.96 12 37.18 0.58 MOL004355 Spinasterol 412.77 7.64 42.98 0.76 MOL005030 gondoic acid 310.58 7.75 30.7 0.2 MOL005043 campest-5-en-3beta-ol 400.76 7.63 37.58 0.71 MOL005481 2,6,10,14,18-pentamethylicosa-2,6,10,14,18-pentaene 342.67 9.51 33.4 0.24 MOL007449 24-methylidenelophenol 412.77 7.75 44.19 0.75 MOL009653 Cycloeucalenol 426.8 7.59 39.73 0.79 MOL009681 Obtusifoliol 426.8 8.15 42.55 0.76 MOL012888 citrostadienol 426.8 8.15 43.28 0.79 MOL012891 (2E,4E,6E)-icosa-2,4,6-trienoic acid 306.54 7.28 41.64 0.2 MOL012893 (E)-(4-methylbenzylidene)-(4-phenyltriazol-1-yl)amine 262.34 4.09 57.87 0.19 The integration of multiple target prediction platforms, namely SwissTargetPrediction, SEA, and SuperPred, followed by standardized annotation, yielded 695 putative protein targets potentially interacting with the aforementioned bioactive constituents. Systematic interrogation of GeneCards, OMIM, and DisGeNET repositories established an AS -associated target compendium comprising 2,444 genes. Venn diagram mapping via the ggVennDiagram package revealed substantial intersection between Perilla frutescens candidate targets and the AS disease gene set, yielding 289 overlapping genes (Fig. 1 A). These overlapping genes constituted the candidate therapeutic target set for Perilla frutescens -mediated AS intervention, providing a critical genetic foundation for elucidating molecular connectivity between botanical bioactive constituents and disease pathophysiology. To further dissect synergistic interaction architectures among candidate targets, the 289 overlapping genes were imported into the STRING database (v11.5) for high-confidence protein–protein interaction (PPI) network construction. Network topology visualization (Fig. 1 B) revealed characteristic scale-free network architecture, with dense physical and functional interconnectivity among nodes corroborating the "multi-component–multi-target–multi-pathway" synergistic regulatory paradigm of Perilla frutescens . Furthermore, quantitative topological characterization via Cytoscape, incorporating weighted ranking of topological metrics including maximal clique centrality (MCC) and node degree, identified a core functional module comprising highly connected hub nodes (Fig. 1 C). These topologically central hub genes exert pivotal regulatory roles within Perilla frutescens -mediated modulation of atherosclerotic pathological networks. 3.2. Gene Enrichment Analysis To systematically elucidate the biological functionalities of the 289 intersecting targets, multidimensional GO functional enrichment and KEGG pathway enrichment analyses were performed. GO enrichment analyses revealed the molecular and cellular functional landscape underlying the putative therapeutic effects of Perilla frutescens (Fig. 1 D). Within the biological process (BP) ontological dimension, target genes exhibited significant enrichment in immune defense responses to lipopolysaccharide (LPS) and bacterial-derived molecular patterns, with pronounced associations to wound healing and inflammatory response modulation. This enrichment architecture suggested that Perilla frutescens may ameliorate vascular endothelial chronic injury and repair dysregulation through attenuation of pathogen-associated molecular pattern (PAMP)-triggered inflammatory cascades. Additionally, pronounced enrichment of steroid and lipid metabolism-related terms suggested that Perilla frutescens may counteract atherosclerotic lipid accumulation pathology through restoration of lipid metabolic homeostasis. Within the cellular component (CC) ontology, target genes predominantly localized to subcellular structures including membrane rafts, membrane microdomains, and vesicle lumina. Given the pivotal role of membrane rafts in receptor clustering and signaling complex assembly, this enrichment profile indicated that Perilla frutescens bioactive constituents may exert vasculoprotective effects through modulation of membrane-associated signal transduction platforms. Within the molecular function (MF) dimension, target genes demonstrated significant enrichment in nuclear receptor activity, ligand-activated transcription factor activity, and eicosanoid receptor activity. This functional landscape aligned closely with the established regulatory roles of nuclear receptors (e.g., PPARs) in governing lipid metabolism and anti-inflammatory transcriptional programs, suggesting transcriptional-level intervention in AS-associated pathological networks by Perilla frutescens . KEGG pathway enrichment analysis further corroborated the mechanistic connectivity between candidate targets and atherosclerotic pathophysiology at the signaling cascade level (Fig. 1 E). The "Lipid and atherosclerosis" pathway exhibited the most pronounced enrichment score and maximal gene ratio, indicating non-random distribution of intersecting target genes with specific convergence upon core atherosclerotic pathological modules. Beyond this primary pathway, target genes demonstrated substantial enrichment in pathological processes encompassing efferocytosis, advanced glycation end product–receptor for AGE (AGE–RAGE) signaling axis, and neutrophil extracellular trap (NET) formation. This multi-pathway enrichment architecture suggested vasculoprotective efficacy of Perilla frutescens through multidimensional synergistic mechanisms: (i) enhancement of macrophage-mediated intraplaque apoptotic cell clearance, mitigating necrotic core formation; (ii) antagonism of AGE-induced oxidative injury; and (iii) suppression of NET-mediated immunothrombotic responses. Additionally, while certain enrichment terms pertained to chemical carcinogenesis and proteoglycan-associated pathways, these signaling modules likely reflect aberrant vascular smooth muscle cell (VSMC) proliferation and pathological extracellular matrix (ECM) remodeling within the atherosclerotic context. Collectively, GO and KEGG enrichment analyses indicated that Perilla frutescens bioactive constituents may attenuate atherosclerotic plaque progression and destabilization through modulation of multiple signaling networks governing cellular proliferation and vascular microenvironmental remodeling. 3.3. Single-Cell Transcriptomic Atlas Construction and Plaque Cellular Heterogeneity Delineation High-resolution single-cell transcriptomic (scRNA-seq) profiling was performed using the GSE159677 dataset, with stringent QC procedures implemented to eliminate low-quality cells and technical noise artifacts. QC metrics revealed uniform distributions of detected gene counts (nFeature_RNA) across all samples (GSM4837523–GSM4837528), with mitochondrial gene expression fractions consistently maintained below 10% (Fig. 2 A), ensuring downstream analyses were predicated upon high-fidelity transcriptomic data. Among the 2,000 HVGs identified, CCL18, APOE, SPP1, and S100A8/A9 exhibited maximal expression variability (Fig. 2 B). These HVGs predominantly encompassed chemotactic signaling, lipid trafficking, and inflammatory response mediators, indicating pronounced cellular heterogeneity associated with lipid metabolic dysregulation and robust inflammatory activation within the plaque microenvironment. Subsequently, UMAP nonlinear dimensionality reduction projected the single-cell transcriptomic landscape onto two-dimensional embeddings, with unsupervised clustering algorithms resolving 26 discrete cellular clusters (Fig. 2 C). Based on canonical marker gene expression profiles, 11 major cellular lineages were annotated (Fig. 2 D), encompassing structural vascular wall constituents, such as endothelial cells, smooth muscle cells (SMCs), and fibroblasts, alongside immune infiltrates including T cells, B cells, natural killer (NK) cells, and mast cells. Critically, higher-resolution dissection of myeloid lineage cells was achieved, enabling unambiguous discrimination of foam cells from classical macrophage populations. Foam cells occupied a discrete clustering topology within UMAP embedding space, indicating substantial transcriptional-level phenotypic reprogramming. To validate cell type annotation fidelity and characterize lineage-specific molecular signatures, expression distribution patterns of canonical marker genes were visualized via dot plot representation (Fig. 2 E). Results revealed distinct and lineage-specific marker gene expression signatures: T cells exhibited selective enrichment of IL7R and CD3D, endothelial cells demonstrated elevated VWF and PECAM1 expression, while smooth muscle cells displayed characteristic ACTA2 and MYH11 upregulation. Critically, foam cells retained myeloid lineage markers (e.g., CD68) while exhibiting pronounced upregulation of inflammation-associated genes including S100A8, S100A12, and MT1G. Given the established roles of these genes in inflammatory signal amplification, metal ion homeostasis regulation, and oxidative stress responses, this differential expression signature indicated that foam cells, following excessive lipid internalization, existed in a pathologically activated state characterized by elevated oxidative stress burden and sustained inflammatory activation. 3.4. Identification of Core Therapeutic Targets via Machine Learning and Clinical Validation To precisely identify hub genes exhibiting potential clinical diagnostic utility and disease relevance within the intersecting gene set, an integrative feature dimensionality reduction strategy synergizing statistical and machine learning methodologies was implemented. Feature selection was performed via LASSO regression, with regularization parameter λ optimized through 10-fold cross-validation. At the λ value corresponding to binomial deviance minimization, redundant variables were effectively eliminated, thereby mitigating multicollinearity-induced compromise of parameter estimation robustness (Fig. 3 A). Random forest (RF) modeling assessed individual gene contributions to disease state classification from a nonlinear perspective, with variable importance ranking identifying high-priority feature genes including PPARG, ITGB1, MMP9, and ALOX5 (Fig. 3 B). This dual-algorithm integrative framework implemented cross-validation of candidate features under divergent modeling assumptions, substantially enhancing feature selection robustness. Intersection analysis between machine learning-derived feature genes and top-ranked WES genes at single-cell resolution yielded 10 core therapeutic targets: HIF1A, ALOX5, STAT1, ITGB1, PPARG, MMP9, PIK3R1, PRKCB, CDK4, and PIK3CA (Table 2 ). To assess the translational potential of this core target signature, multivariable logistic regression diagnostic modeling was performed within the independent external validation dataset GSE100927. ROC curve analysis revealed an AUC of 0.996 (Fig. 3 C), indicating superior discriminatory capacity of the 10-gene signature for distinguishing atherosclerotic pathological specimens from healthy controls, underscoring substantial clinical diagnostic utility. Table 2 List of Core Candidate Targets Identified Based on Multidimensional Screening Strategies Symbol Full Name Function HIF1A Hypoxia Inducible Factor 1 Alpha Hypoxia response ALOX5 Arachidonate 5-Lipoxygenase Inflammatory mediator STAT1 Signal Transducer and Activator of Transcription 1 Immune signaling ITGB1 Integrin Subunit Beta 1 Cell adhesion PPARG Peroxisome Proliferator-Activated Receptor Gamma Lipid metabolism MMP9 Matrix Metallopeptidase 9 ECM degradation PIK3R1 Phosphoinositide-3-Kinase Regulatory Subunit 1 PI3K pathway regulation PRKCB Protein Kinase C Beta Signal transduction CDK4 Cyclin Dependent Kinase 4 Cell cycle control PIK3CA Phosphatidylinositol 3-Kinase Catalytic Subunit Alpha Cell growth To delineate the cellular origins and spatial distribution patterns of core targets within the plaque microenvironment, the 10 core genes were projected onto single-cell UMAP embedding space for visualization (Fig. 3 D). Results demonstrated non-uniform expression profiles across cellular compartments, with specific enrichment within macrophage and foam cell subsets, contrasting with relatively diminished expression in T cells, B cells, and smooth muscle cells. This cell type-specific enrichment pattern validated the efficacy of the WES selection strategy while revealing that Perilla frutescens likely exerts vasculoprotective effects primarily through targeted modulation of key molecular networks within intraplaque myeloid lineage cells, encompassing PPARG-mediated lipid metabolic reprogramming, MMP9-mediated extracellular matrix degradation, and HIF1A-governed hypoxic responses, thereby intercepting plaque progression and destabilization trajectories. 3.5. Pseudotemporal Dynamics and Intercellular Communication Profiling of Core Targets Single-cell pseudotemporal trajectory reconstruction of myeloid lineage cells was performed using the Monocle3 algorithm. Trajectory reconstruction revealed continuous phenotypic evolution of intraplaque macrophages from a homeostatic root state toward a terminal foam cell state (Fig. 4 A). Along this differentiation trajectory, the 10 core target genes (ALOX5, CDK4, HIF1A, ITGB1, MMP9, PIK3CA, PIK3R1, PPARG, PRKCB, STAT1) exhibited highly coordinated temporal expression dynamics, with transcriptional levels displaying sigmoidal nonlinear upregulation trajectories correlated with pseudotemporal progression. This expression architecture, exhibiting tight synchronization with cellular state transitions, indicated that these core genes not only correlate with foam cell formation but likely actively drive key pathological processes including macrophage lipid metabolic reprogramming, inflammatory transcriptional program activation, and phenotypic transformation, thereby executing pivotal regulatory functions in atherosclerotic plaque initiation and maintenance. Furthermore, global cell–cell communication networks within the plaque microenvironment were systematically reconstructed using the CellChat toolkit to delineate functional positioning of core targets within multicellular interaction architectures. Global network topological analysis revealed intensive signaling crosstalk among macrophages, foam cells, endothelial cells, and T cells, collectively establishing a cellular communication hub orchestrating the inflammatory plaque microenvironment (Fig. 4 C). Ligand–receptor interaction dissection further identified an ITGB1 (integrin β1)-centered cellular communication module (Fig. 4 B). Foam cells, functioning as principal ligand-sending cells, exhibited elevated expression of pro-inflammatory/pro-fibrotic ligands including SPP1, FN1, and VCAM1; correspondingly, the receptor ITGB1 and cognate heterodimeric complexes (e.g., α4β1, α5β1 integrins) demonstrated predominant expression in endothelial cells, smooth muscle cells (SMCs), and T cells. Ligand–receptor interaction strength quantification revealed substantial communication probabilities for SPP1–ITGB1-associated interaction pairs (including SPP1–α5β1 and SPP1–α8β1) between foam cells and smooth muscle cells/endothelial cells, implicating their involvement in mediating critical pathological processes including cell adhesion, transendothelial migration, and smooth muscle cell phenotypic switching. Collectively, these cell–cell communication network analyses suggested that Perilla frutescens bioactive constituents may disrupt pathological signaling between foam cells and stromal/immune cells through targeted modulation of ITGB1 receptor expression or activity, thereby attenuating inflammatory signal amplification cascades and intercepting SPP1–ITGB1 axis-mediated aberrant cellular migration and vascular microenvironmental remodeling. 3.6. Molecular Docking Analysis To elucidate the molecular recognition principles governing Perilla frutescens bioactive constituent–target interactions at atomic resolution, systematic molecular docking analyses were performed. Binding energy heatmap analysis (Fig. 5 A) revealed favorable binding affinities between Perilla frutescens bioactive constituents and the 10 core targets (e.g., HIF1A, ITGB1), with docking scores predominantly distributed below − 5.0 kcal/mol, indicating thermodynamically favorable ligand–receptor complex formation with propensity for stable conformational association. Notably, luteolin, the principal bioactive constituent, exhibited pronounced multi-target binding potential, demonstrating particularly favorable binding free energies with the cellular communication hub ITGB1 (− 8.9 kcal/mol), inflammatory regulatory node ALOX5 (− 8.8 kcal/mol), and extracellular matrix remodeling mediator MMP9 (− 8.6 kcal/mol). These results are indicative of robust binding affinities, suggesting pivotal contributions to the multi-target synergistic modulation of AS by Perilla frutescens . Furthermore, three-dimensional docking pose analysis unveiled mechanistic details underlying molecular recognition. Within the ALOX5–luteolin complex (Fig. 5 B), the ligand occupied the hydrophobic binding pocket of ALOX5, forming hydrogen bonds with the critical residue Gln437, thereby conferring conformational stabilization. Within the ITGB1–luteolin complex (Fig. 5 C), the ligand occupied a putative binding pocket within the receptor extracellular domain, with hydroxyl moieties forming an extensive hydrogen bonding network with the pivotal residue Glu320, complemented by synergistic van der Waals interactions and hydrophobic effects stabilizing the binding interface. Based on the spatial conformational architecture of this binding mode, luteolin may exert competitive inhibition of endogenous ligand (e.g., SPP1) engagement with ITGB1 through steric occlusion, thereby disrupting the pathological SPP1–ITGB1 communication axis identified through cell–cell communication network analyses. 3.7. MD Simulation Analysis To assess the dynamic stability and conformational evolution characteristics of the luteolin–ITGB1 complex under quasi-physiological solvation conditions, 100-ns all-atom MD simulations were executed. RMSD trajectory analysis (Fig. 5 D) revealed rapid convergence to thermodynamic equilibrium at approximately 30 ns following initial solvation relaxation. Subsequently, RMSD trajectories exhibited pronounced stability, with fluctuation amplitudes confined within a narrow 0.10–0.20 nm interval, devoid of conformational drift or ligand dissociation events, indicating preservation of protein backbone structural integrity upon ligand engagement. RMSF analysis (Fig. 5 E) quantified residue-specific local flexibility profiles. Results revealed pronounced rigidity within the ITGB1 core structural domain (RMSF < 0.20 nm), with elevated fluctuations confined to surface-exposed flexible loop regions distal to the binding site. Critically, residues within the core binding pocket exhibited minimal fluctuations, confirming that luteolin effectively stabilizes the functional active conformation of ALOX5 via an induced-fit mechanism. Rg analysis (Fig. 5 F) demonstrated stable Rg values within a narrow 2.17–2.20 nm range throughout the simulation trajectory, devoid of unfolding propensity, confirming sustained maintenance of compact globular folded architecture. Additionally, solvent accessible surface area (SASA) remained stable within the 190–205 nm² range (Fig. 5 G), indicating invariant surface exposure characteristics within the dynamic aqueous environment, with negligible structural expansion or compaction. Collectively, MD simulations furnished robust biophysical evidence corroborating the formation of persistent, specific, and thermodynamically stable luteolin–ITGB1 molecular complexes, establishing a theoretical foundation for developing luteolin as a candidate anti-atherosclerotic lead compound. 4. Discussion 4.1. Multi-component-Multi-target-Multi-pathway Synergistic Intervention Mode of Perilla frutescens from the Perspective of Systems Pharmacology AS is essentially not a linear dysregulation of a single signaling pathway, but a systemic disease driven by the deep interweaving of lipid metabolic dysregulation and immune-inflammatory responses in spatiotemporal dimensions (Ajoolabady et al., 2024 ). This pathological essence determines that AS has intrinsic resistance to single-target treatment strategies. Clinical observations reveal substantial residual cardiovascular risk even among patients achieving low-density lipoprotein cholesterol (LDL-C) targets under intensive statin regimens. This phenomenon reflects the persistent activation of compensatory inflammatory pathways alongside the intrinsic robustness of pathological networks. Specifically, the homeostatic capacity of diseased systems to sustain maladaptive functional states despite localized perturbations (Kanuri et al., 2025 ; Tang et al., 2024 ). The present investigation demonstrates that the multi-component, multi-target intervention paradigm of Perilla frutescens transcends mere additive pharmacological effects, instead achieving the systematic attenuation of topological robustness within AS pathological networks through synergistic perturbation of 289 disease-critical nodes, thereby precipitating a global phase transition from pathological to physiological homeostatic states. From a network pharmacology perspective, this multi-nodal synergistic intervention strategy exhibits enhanced resilience to compensatory mechanisms relative to the localized perturbations induced by single-target therapeutics, embodying a contemporary molecular-network interpretation of the holistic philosophy underpinning traditional Chinese medicine. The pronounced enrichment of "membrane rafts" and "membrane microdomains" within GO functional analyses provides a distinctive biophysical framework for elucidating the pharmacological mechanisms of Perilla frutescens . Membrane rafts, characterized as cholesterol- and sphingolipid-enriched liquid-ordered microdomains within plasma membranes, constitute essential spatial platforms facilitating the assembly of functional signalosomes for inflammatory pattern recognition receptors (e.g., TLR4) and lipid scavenger receptors (e.g., CD36) (Nieto-Garai et al., 2022 ). Under atherogenic conditions, oxidized low-density lipoprotein (ox-LDL) binding to raft-localized CD36 receptors precipitates macrophage lipid overload, whereas pathogen-associated molecular patterns (PAMPs) amplify inflammatory cascades through the promotion of spatial clustering of TLR4/MyD88 signaling complexes within raft microdomains (Chen et al., 2022 ). The phytosterols enriched in Perilla frutescens (e.g., β-sitosterol, stigmasterol), possessing steroid scaffolds structurally homologous to cholesterol, may intercalate into raft lipid bilayers via competitive displacement mechanisms, thereby modulating microdomain physicochemical properties encompassing membrane thickness, lipid fluidity, and membrane curvature (Zhao et al., 2025 ). Tang et al. (Tang et al., 2021 ) systematically characterized the biophysical effects of phytosterols within plant plasma membrane-mimetic lipid systems through the integration of fluorescence lifetime imaging microscopy and all-atom MD simulations. Their findings demonstrate that phytosterol incorporation induces significant reductions in membrane lipid area, elevations in bilayer thickness, enhancements in fatty acyl chain ordering, and the formation of phytosterol-enriched clusters corresponding to lipid microdomain/phase separation phenomena. This biophysical remodeling of membrane rafts disrupts the spatial clustering and functional interaction interfaces of signaling receptors, thereby imposing signal transduction blockade at the mechanistic origin, ultimately conferring membrane-targeted anti-inflammatory effects of broader spectrum than those achievable through single-receptor antagonism (Yuan et al., 2020 ). KEGG pathway enrichment analysis results show that the "Lipid and atherosclerosis" pathway presents the most significant enrichment, indicating that candidate target genes are highly enriched in the functional core modules of the AS pathological network. This pathway covers the complete lipid metabolism axis from ox-LDL uptake, cholesterol esterification to reverse transport. The synergistic targeting of multiple critical nodes within this pathway by Perilla frutescens bioactive constituents may disrupt the lipid accumulation–inflammation activation positive feedback loop inherent to foam cell formation through a bidirectional regulatory paradigm of upstream flux restriction coupled with downstream efflux enhancement (Kong et al., 2022 ). Perilla frutescens harbors an extensive repertoire of bioactive constituents (including α-linolenic acid, flavonoids, and phenolic acids) that exert synergistic actions upon multiple critical processes encompassing lipid uptake, cholesterol esterification, reverse cholesterol transport, oxidative stress mitigation, and inflammatory response attenuation (Hou et al., 2022 ). Evidence from animal models and cellular experiments demonstrates that Perilla frutescens extracts downregulate the expression of ox-LDL scavenger receptors (CD36, LOX-1), suppress foam cell formation, facilitate cholesterol efflux through the upregulation of ABCA1 and SR-B1 expression, and substantially attenuate the secretion of pro-inflammatory cytokines (e.g., IL-1β, MCP-1) (Pothinam et al., 2025 ). Impaired efferocytosis precipitates secondary necrosis of apoptotic cells and progressive expansion of necrotic cores, processes intimately associated with vulnerable plaque formation and the incidence of acute coronary syndrome (ACS) (Morrissey et al., 2020 ). Perilla frutescens may restore intraplaque efferocytic efficiency through the upregulation of phagocytic receptors (e.g., MerTK), the activation of peroxisome proliferator-activated receptor gamma (PPARγ)-dependent bridging molecule secretion, and the suppression of the inhibitory CD47–SIRPα axis. The restoration of efferocytic capacity not only diminishes necrotic core dimensions but also initiates active inflammation resolution programs through the liberation of specialized pro-resolving mediators (SPMs), including lipoxins and resolvins (Filep, 2022 ). The pronounced enrichment of the neutrophil extracellular trap (NET) formation (NETosis) pathway suggests the potential modulatory capacity of Perilla frutescens in immunothrombosis. NETs exert pathogenic actions across multiple stages of AS progression: during early phases, histones impose cytotoxic injury upon vascular endothelium; throughout disease advancement, NETs function as damage-associated molecular patterns (DAMPs) activating NLRP3 inflammasomes to amplify inflammatory responses; and during acute events, NETs serve as thrombogenic scaffolds capturing platelets and coagulation factors (Döring et al., 2017 ). Luteolin, a cardinal bioactive constituent of Perilla frutescens , demonstrates established capacity for inhibiting NADPH oxidase (NOX2) activity. Given that NOX2-mediated reactive oxygen species (ROS) generation constitutes a critical initiating determinant of NETosis, these findings suggest that Perilla frutescens may suppress NET formation at mechanistic origins (Xia et al., 2014 ). Furthermore, the substantial enrichment of the advanced glycation end product (AGE)–receptor for AGE (RAGE) signaling pathway suggests distinctive therapeutic potential for Perilla frutescens in patients with concomitant diabetes mellitus and AS. The hyperactivation of the AGE–RAGE axis perpetuates oxidative stress amplification and inflammatory signal cascades via nuclear factor-κB (NF-κB) pathways, constituting a pivotal pathogenic mechanism underlying accelerated AS progression in diabetic populations (B. Wang et al., 2025 ). 4.2. Molecular Analysis of Foam Cell Differentiation Trajectory and Core Target Identification at Single-Cell Resolution Conventional transcriptomic investigations relying upon bulk RNA sequencing (bulk RNA-seq) frequently obscure the profound cellular heterogeneity within atherosclerotic plaques through population-averaging effects, thereby precluding the precise identification of critical molecular events (Tzec-Interián et al., 2025 ). The present study reconstructed the comprehensive cellular atlas of human carotid atherosclerotic plaques through the integration of high-resolution single-cell RNA sequencing (scRNA-seq) data, thereby achieving precise discrimination between foam cells and classical macrophage subpopulations. Building upon this foundation, the implementation of a dual-algorithm feature selection framework integrating least absolute shrinkage and selection operator (LASSO) regression and random forest methodologies enabled the precise identification of a 10-gene signature module centered upon HIF1A, PPARG, and ALOX5 For instance, Xu et al. (Xu et al., 2022 ) integrated multiple Gene Expression Omnibus (GEO) microarray datasets, identifying 611 AS-associated differentially expressed genes through differential expression analyses, subsequently employing multiple machine learning algorithms (LASSO, random forest) for key gene selection, and validating findings across external human and murine specimens. Their investigation ultimately proposed a diagnostic gene pair comprising DHRS9 and PTPRJ, which exhibited robust discriminatory capacity between AS and control samples and demonstrated strong associations with diverse immune cell infiltration patterns. Ban et al. (Ban et al., 2024 ) performed weighted gene co-expression network analysis (WGCNA) and differential expression profiling on AS and ischemic stroke (IS) datasets, constructing shared differentially expressed gene networks and identifying ATF3, CCL3, CCL4, JUNB, KRAS, and ZC3H12A as shared hub genes potentially participating in the pathological processes underlying both AS and IS, with subsequent validation of expression trends via quantitative real-time PCR (qPCR) analysis of clinical specimens. This gene module represents not a stochastic assemblage but rather a constellation profoundly reflecting the immunometabolic reprogramming signature inherent to foam cell formation. Pseudotemporal trajectory reconstruction analyses confirmed that the expression abundance of this gene module exhibits highly coordinated sigmoidal upregulation dynamics along pathological differentiation trajectories, strongly implicating these core genes not as passive markers of foam cell states but rather as cardinal regulatory determinants actively orchestrating phenotypic transitions. The pronounced upregulation of HIF1A unveils the distinctive hypoxic microenvironmental signature characterizing plaque core regions. With progressive plaque expansion, oxygen diffusion limitations precipitate substantial reductions in local oxygen tension, thereby triggering HIF1A-dependent metabolic reprogramming characterized by the metabolic transition from oxidative phosphorylation to glycolysis (the Warburg effect) (Qiu et al., 2023 ). While this metabolic shift sustains cellular bioenergetic homeostasis under hypoxic conditions in the short term, it precipitates persistent lactate accumulation. Lactate, through the activation of G protein-coupled receptor 81 (GPR81), suppresses cyclic adenosine monophosphate (cAMP) signaling cascades, thereby further attenuating anti-inflammatory programs and exacerbating macrophage polarization toward M1-type pro-inflammatory phenotypes, ultimately establishing a self-perpetuating hypoxia–metabolism–inflammation vicious cycle. As a cardinal member of the lipid-sensing nuclear receptor superfamily, PPARγ facilitates reverse cholesterol transport (RCT) under physiological conditions through the transcriptional activation of lipid efflux transporters (e.g., ABCA1/ABCG1), thereby conferring atheroprotective effects (Zhang et al., 2024 ). However, under conditions of sustained intraplaque lipid overload, PPARγ function may undergo pathological "hijacking": ligand activation upregulates fatty acid translocase expression (e.g., CD36), paradoxically exacerbating dysregulated ox-LDL uptake; concurrently, elevated oxidized lipid concentrations may impair PPARγ–coactivator interactions through covalent modifications, thereby attenuating its transcriptional activation capacity. This double-edged sword phenomenon of PPARγ functionality partially explains the failure of PPARγ agonist monotherapy to substantially improve clinical outcomes in AS patients (Riccioni et al., 2009 ). LTB4 exerts potent chemotactic actions via leukotriene B4 receptor 1 (BLT1), orchestrating the recruitment of neutrophils and monocytes into atherosclerotic plaques, whereas CysLTs, acting through CysLT1 receptors, promote vascular smooth muscle cell contraction and endothelial permeability elevations. The sustained generation of ALOX5 metabolites and inflammatory cell infiltration establish a positive feedback loop, constituting a pivotal molecular mechanism underlying the recalcitrance of chronic plaque inflammation to spontaneous resolution (Kotlyarov & Kotlyarova, 2022 ). Furthermore, these 10 core gene targets demonstrated exceptional diagnostic performance (AUC = 0.996) within independent external validation cohorts. These findings not only substantiate the robustness and reproducibility of the feature selection framework but also underscore the substantial clinical translational potential of these core molecules as liquid biopsy biomarkers for AS. 4.3. Analysis of Cell-Cell Communication Network and Elucidation of ITGB1 Hub Function Atherosclerotic plaques constitute not static cellular aggregations but rather highly dynamic cellular ecosystems sustained by intricate intercellular communication networks (Raju et al., 2025 ). Diverse cellular populations within plaques establish dense information exchange networks through ligand–receptor interactions, extracellular vesicle (EV) trafficking, and metabolite signaling. These multicellular coordination patterns collectively orchestrate plaque inflammatory microenvironmental characteristics and structural integrity (Raju et al., 2024 ). The present investigation employed CellChat computational frameworks to systematically dissect the global intercellular communication landscape within plaques, thereby unveiling the molecular mechanisms through which Perilla frutescens exerts therapeutic effects via targeted disruption of the SPP1–ITGB1 signaling axis. Secreted phosphoprotein 1 (SPP1), predominantly expressed by discrete macrophage subpopulations, constitutes a pivotal signaling hub orchestrating AS progression and plaque destabilization (Li et al., 2025 ). SPP1 promotes fibrotic responses within fibro-progenitor cells through integrin receptor engagement (encompassing ITGB1, ITGAV/ITGB5). Early functional investigations demonstrate that the SPP1–integrin signaling axis drives vascular smooth muscle cell (VSMC) migration and phenotypic switching, processes intimately associated with intimal thickening and foam cell formation (Huang et al., 2024 ). Ligand–receptor interaction analyses reveal that SPP1 (osteopontin, OPN), abundantly expressed and secreted by foam cells, represents the predominant ligand activating integrin β1 (ITGB1) on vascular wall cells (endothelial cells, smooth muscle cells) and infiltrating immune cells (Yim et al., 2022 ). ITGB1, as the cardinal β subunit within the integrin superfamily, heterodimerizes with diverse α subunits (e.g., α4β1, α5β1, α8β1), exerting nodal regulatory functions in cellular adhesion, migration, and mechanotransduction through bidirectional "outside-in" and "inside-out" signaling mechanisms (Su et al., 2024 ). SPP1 engagement with endothelial α4β1/α5β1 integrins activates focal adhesion kinase (FAK)–Src signaling cascades, precipitating tyrosine phosphorylation and subsequent endocytic degradation of vascular endothelial cadherin (VE-cadherin), thereby compromising adherens junction integrity (Fei et al., 2025 ). The attenuation of endothelial barrier function precipitates elevations in vascular permeability, thereby facilitating transendothelial migration (TEM) of circulating monocytes and sustaining the replenishment of intraplaque inflammatory cell reservoirs (Dalal et al., 2019 ). Under physiological conditions, vascular smooth muscle cells (VSMCs) manifest a contractile phenotype essential for the maintenance of vascular tone and structural integrity (Cao et al., 2022 ). ITGB1, functioning as a cardinal mechanosensor, transduces ECM mechanical cues into intracellular biochemical signals. ITGB1 signaling cascades drive VSMC phenotypic switching toward synthetic states through the activation of downstream phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt) and mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathways, manifested by the downregulation of contractile proteins (α-smooth muscle actin [α-SMA], smooth muscle myosin heavy chain [SM-MHC]) alongside the upregulation of matrix synthesis proteins (collagen, proteoglycans) and matrix metalloproteinases (MMPs) (Wu et al., 2025 ). The excessive proliferation and enhanced apoptosis of synthetic VSMCs ultimately compromise plaque structural stability, thereby elevating plaque rupture risk. Predicated upon these mechanisms, Perilla frutescens bioactive constituents may interrupt the transmission of pathological signals from foam cells to the microenvironment through targeted disruption of SPP1–ITGB1 interactions at the intercellular communication level, thereby achieving: restoration of endothelial barrier functionality to attenuate inflammatory cell infiltration; preservation of VSMC contractile phenotypes to stabilize plaque architecture; and suppression of excessive MMP activation to prevent fibrous cap rupture. This intervention paradigm predicated upon the disruption of pathological intercellular communication furnishes a conceptually innovative therapeutic framework for AS that transcends conventional intracellular signaling pathway targeting. 4.4. Multi-target Binding Characteristics of Luteolin and Kinetic Stability of ITGB1 Complex Molecular docking and MD simulations furnished atomic-resolution structural and kinetic evidence characterizing interactions between Perilla frutescens bioactive constituents and core targets. Binding energy heatmap analyses reveal favorable binding affinities of 16 Perilla frutescens constituents toward the 10 core targets, indicating thermodynamic spontaneity of ligand–receptor complex formation. Within the Perilla frutescens bioactive constituent repertoire, luteolin exhibits pronounced multi-target binding potential. As a prototypical flavonoid, the molecular architecture of luteolin, encompassing A/B dual aromatic ring systems, C-ring C2–C3 unsaturation, and multiple phenolic hydroxyl moieties, confers the structural foundation for establishing stable non-covalent interactions with diverse protein targets. Docking analyses demonstrate that luteolin exhibits favorable binding free energies toward the intercellular communication hub ITGB1 (− 8.9 kcal/mol), the inflammatory rate-limiting enzyme ALOX5 (− 8.8 kcal/mol), and the extracellular matrix degradation effector MMP9 (− 8.6 kcal/mol). Three-dimensional conformational analyses unveil the molecular recognition mechanisms underlying luteolin–target interactions. Within the ALOX5–luteolin complex, the ligand intercalates into the hydrophobic binding cavity of the catalytic domain, establishing stable interactions with regions proximal to the non-heme iron active site; the A-ring phenolic hydroxyl forms a hydrogen bond with the critical residue Gln437, thereby anchoring the binding conformation. The spatial occupancy of the active site by luteolin may obstruct leukotriene biosynthetic pathways through competitive inhibition of substrate access (Ren et al., 2021 ). Within the ITGB1–luteolin complex, the ligand occupies the ligand-binding pocket of the extracellular I-like domain of ITGB1; the C4′-hydroxyl forms a hydrogen bond with the carboxyl side chain of the critical residue Glu320, while the B-ring engages in π-alkyl interactions with hydrophobic residues lining the pocket interior (De Aguiar et al., 2025 ). This binding site coincides precisely with the recognition interface for endogenous ligands such as SPP1, suggesting that luteolin may competitively abrogate pathological SPP1–ITGB1 interactions through steric occlusion (Zhou et al., 2024 ). While molecular docking furnishes static structural snapshots of ligand–receptor interactions, proteins undergo continuous conformational fluctuations under physiological conditions. The capacity of ligands to sustain stable binding within dynamic aqueous environments directly governs the duration and magnitude of pharmacological activity (Fu et al., 2018 ). This property assumes particular significance for ITGB1, given that as a mechanotransduction receptor, its functionality critically depends upon dynamic transitions between bent low-affinity and extended high-affinity conformational states. The present investigation systematically assessed the spatiotemporal dynamic evolution of the luteolin–ITGB1 complex through 100-nanosecond all-atom MD simulations. RMSD trajectory analyses reveal rapid convergence of the complex to thermodynamic equilibrium at approximately 30 nanoseconds; subsequently, fluctuation amplitudes remained within a narrow 0.10–0.20 nm range, with no observable ligand dissociation or substantial conformational drift throughout the simulation. This rapid equilibration characteristic suggests that luteolin efficiently accommodates the receptor binding pocket microenvironment, indicating potential for prompt in vivo onset of action. The local conformational stabilization effects unveiled through RMSF analyses constitute a cardinal finding of the present investigation. Core residues within the ligand-binding pocket, most notably Glu320 and adjacent hydrophobic residue clusters, exhibit substantially diminished fluctuation amplitudes (RMSF < 0.20 nm), whereas regions of elevated fluctuation remain confined to surface-exposed flexible loops distal to the binding interface. The biological significance of this local rigidification resides in luteolin capacity to effectively lock ITGB1 into a defined functional conformation through binding pocket occupancy and the establishment of multivalent non-covalent interaction networks, thereby attenuating receptor responsiveness to endogenous ligands (e.g., SPP1) or mechanical stimuli. Given that ITGB1 signal transduction relies upon ligand-induced conformational changes (outside-in signaling), this conformational locking effect may fundamentally abrogate downstream FAK/Src signaling cascades mediated by the SPP1–ITGB1 axis (Su et al., 2024 ). From a medicinal chemistry perspective, ligand-induced binding pocket rigidification diminishes conformational entropy at the binding interface, thereby elevating the energy barrier for ligand dissociation through enthalpy–entropy compensation mechanisms (Stank et al., 2016 ). This kinetics-driven pharmacological advantage assumes particular importance for chronic pathologies such as AS requiring sustained therapeutic intervention (Wang et al., 2020 ). This conformational locking mechanism significantly prolongs the receptor residence time of luteolin, thereby allowing the ligand to sustain target occupancy and functional blockade even under conditions of declining plasma concentrations. The stability of the Rg ( 2.17–2.20 nm) and solvent-accessible SASA (190–205 nm²) throughout the simulation trajectory indicates that ITGB1 maintains a compact, globular folded conformation without structural unfolding or long-range allosteric propagation. This binding mode substantiates the mechanistic hypothesis that luteolin functions as an orthosteric competitive inhibitor of the SPP1–ITGB1 axis, disrupting pathological intercellular communication through direct occupancy of the ligand recognition interface rather than indirect allosteric modulation. Integration with the aforementioned CellChat intercellular communication network analyses suggests that luteolin targeted intervention upon ITGB1 may sever pro-inflammatory and pro-remodeling signals emanating from foam cells into the vascular microenvironment at the intercellular communication level, thereby conferring multifaceted protective effects encompassing endothelial barrier restoration, VSMC contractile phenotype preservation, and extracellular matrix degradation suppression (Wu et al., 2018 ). 5. Conclusion In summary, the present investigation establishes a multiscale computational analytical framework integrating single-cell transcriptomics, machine learning algorithms, and MD simulations to systematically elucidate the multi-target synergistic mechanisms underlying the therapeutic intervention of Perilla frutescens frutescens seed in AS. The investigation identifies a 10-gene core signature module encompassing HIF1A, PPARG, and ITGB1, which not only demonstrates exceptional diagnostic performance within independent clinical validation cohorts (AUC = 0.996) but also profoundly captures the molecular signatures of immunometabolic reprogramming accompanying macrophage-to-foam cell differentiation. More critically, the study furnishes atomic-resolution biophysical evidence substantiating that luteolin, a cardinal bioactive constituent of Perilla frutescens , engages integrin β1 (ITGB1) with high specificity through conformational locking mechanisms, thereby spatially disrupting the foam cell-driven SPP1–ITGB1 pro-inflammatory communication axis. Collectively, these findings suggest that the targeted disruption of pathological cell adhesion receptors and their mediated signal crosstalk constitutes a highly promising innovative therapeutic strategy for addressing RIR in AS. While the present investigation primarily employs computational biology and molecular simulation strategies, future studies warrant systematic validation of the pharmacokinetic/pharmacodynamic (PK/PD) profiles, in vivo targeting efficiency, and long-term therapeutic efficacy of luteolin–ITGB1 interactions within animal models and clinical trials, alongside comprehensive exploration of its clinical translational potential as a precision therapeutic candidate for AS. Declarations Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Funding This work was supported by the financial support from Key Research and Development Program of Zhejiang Province, China (2020C02034). Authors' contributions Chenchen Yang: Conceptualization (assisted), data curation, formal analysis, investigation, software, visualization, writing – original draft. Acknowledgements The authors wish to acknowledge our hardworking computer, which endured long-term high-intensity workloads including bioinformatics data processing, code running, chart visualization and manuscript drafting without complaint, serving as an indispensable silent partner throughout this research. Availability of data and materials Target intersection analysis and Venn diagram visualization were performed in R (v4.3.2) with the ggVennDiagram package (v1.2.2). Subsequent PPI network construction, functional enrichment and single-cell transcriptomic analysis were carried out via standard bioinformatics pipelines. All public database resources are freely accessible via the listed URLs. Generated processed datasets are available upon reasonable request to the corresponding author for research replication. References Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B, Lindahl E (2015) GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX , 1–2 , 19–25. https://doi.org/https://doi.org/10.1016/j.softx.2015.06.001 Ajoolabady A, Pratico D, Lin L, Mantzoros CS, Bahijri S, Tuomilehto J, Ren J (2024) Inflammation in atherosclerosis: pathophysiology and mechanisms. Cell Death Dis 15(11):817. https://doi.org/10.1038/s41419-024-07166-8 Alsaigh T, Evans D, Frankel D, Torkamani A (2022) Decoding the transcriptome of calcified atherosclerotic plaque at single-cell resolution. Commun Biol 5(1):1084. https://doi.org/10.1038/s42003-022-04056-7 Ban R, Huo C, Wang J, Zhang G, Zhao X (2024) Exploration of the Shared Gene Signatures and Molecular Mechanisms Between Ischemic Stroke and Atherosclerosis. Int J Gen Med 17:2223–2239. https://doi.org/10.2147/ijgm.s454336 Cao G, Xuan X, Hu J, Zhang R, Jin H, Dong H (2022) How vascular smooth muscle cell phenotype switching contributes to vascular disease. Cell Communication Signaling: CCS 20. https://doi.org/10.1186/s12964-022-00993-2 Cao J, Spielmann M, Qiu X, Huang X, Ibrahim DM, Hill AJ, Zhang F, Mundlos S, Christiansen L, Steemers FJ, Trapnell C, Shendure J (2019) The single-cell transcriptional landscape of mammalian organogenesis. Nature 566(7745):496–502. https://doi.org/10.1038/s41586-019-0969-x Chen Y, Zhang J, Cui W, Silverstein RL (2022) CD36, a signaling receptor and fatty acid transporter that regulates immune cell metabolism and fate. J Exp Med 219(6). https://doi.org/10.1084/jem.20211314 Dalal P, Muller W, Sullivan D (2019) Endothelial Cell Calcium Signaling During Barrier Function and Inflammation. Am J Pathol. https://doi.org/10.1016/j.ajpath.2019.11.004 De Aguiar A, De Carvalho LBR, Gomes C, Castro MM, Martins F, Borges LL (2025) Computational Insights into the Antioxidant Activity of Luteolin: Density Functional Theory Analysis and Docking in Cytochrome P450 17A1. Pharmaceuticals , 18 . https://doi.org/10.3390/ph18030410 Di Muro F, Vogel B, Sartori S, Bay B, Oliva A, Feng Y, Krishnan P, Sweeny J, Gitto M, Smith K, Moreno P, Nicolas J, Krishnamoorthy P, Leone PP, Bhatt D, Dangas G, Kini A, Sharma S, Mehran R (2025) Prognostic impact of residual inflammatory and triglyceride risk in statin-treated patients with well-controlled LDL cholesterol and atherosclerotic cardiovascular disease. Eur J Prev Cardiol. https://doi.org/10.1093/eurjpc/zwaf112 Döring Y, Soehnlein O, Weber C (2017) Neutrophil Extracellular Traps in Atherosclerosis and Atherothrombosis. Circul Res 120(4):736–743. https://doi.org/doi:10.1161/CIRCRESAHA.116.309692 Eberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. J Chem Inf Model 61(8):3891–3898. https://doi.org/10.1021/acs.jcim.1c00203 Fei Z, Liu Z-T, Zhou G-W, Liang F, Wang Y-H, Chen L, Zhang W-F, Shen L, Lu Y-Q, Huo H, Shi X, Fang L, He B (2025) Integrin β3-mediated platelet extracellular vesicle adhesion facilitates vascular smooth muscle cell dysfunction in postinjury intimal hyperplasia. Int J Biol Sci 21:2380–2395. https://doi.org/10.7150/ijbs.101391 Filep JG (2022) Targeting Neutrophils for Promoting the Resolution of Inflammation. Frontiers in Immunology , 13 . https://doi.org/10.3389/fimmu.2022.866747 Fu Y, Zhao J, Chen Z (2018) Insights into the Molecular Mechanisms of Protein-Ligand Interactions by Molecular Docking and Molecular Dynamics Simulation: A Case of Oligopeptide Binding Protein. Computational and Mathematical Methods in Medicine , 2018 . https://doi.org/10.1155/2018/3502514 Hao Y, Hao S, Andersen-Nissen E, Mauck WM 3rd, Zheng S, Butler A, Lee MJ, Wilk AJ, Darby C, Zager M, Hoffman P, Stoeckius M, Papalexi E, Mimitou EP, Jain J, Srivastava A, Stuart T, Fleming LM, Yeung B, Satija R (2021) Integrated analysis of multimodal single-cell data. Cell 184(13):3573–3587e3529. https://doi.org/10.1016/j.cell.2021.04.048 Hou T, Netala VR, Zhang H, Xing Y, Li H, Zhang Z (2022) Perilla frutescens frutescens: A Rich Source of Pharmacological Active Compounds. Molecules 27. https://doi.org/10.3390/molecules27113578 Huang K, Chen S, Yu L, Wu Z, Chen QJ, Wang XQ, Li F-F, Liu J, Wang Y-X, Mao L-S, Shen W, Zhang R-Y, Shen Y, Lu L, Dai Y, Ding F (2024) Serum secreted phosphoprotein 1 level is associated with plaque vulnerability in patients with coronary artery disease. Front Immunol 15. https://doi.org/10.3389/fimmu.2024.1285813 Kanuri B, Maremanda KP, Chattopadhyay D, Essop MF, Lee MKS, Murphy AJ, Nagareddy PR (2025) Redefining Macrophage Heterogeneity in Atherosclerosis: A Focus on Possible Therapeutic Implications. Compr Physiol 15(2):e70008. https://doi.org/10.1002/cph4.70008 Kong P, Cui Z-Y, Huang X-F, Zhang D-D, Guo R-J, Han M (2022) Inflammation and atherosclerosis: signaling pathways and therapeutic intervention. Signal Transduct Target Therapy 7(1). https://doi.org/10.1038/s41392-022-00955-7 Kotlyarov S, Kotlyarova A (2022) Molecular Pharmacology of Inflammation Resolution in Atherosclerosis. Int J Mol Sci 23(9):4808 Li Y, Wang S, Zhang R, Gong Y, Che Y, Li K, Pan Z (2025) Single-cell and spatial analysis reveals the interaction between ITLN1 + foam cells and SPP1 + macrophages in atherosclerosis. Front Cardiovasc Med 12. https://doi.org/10.3389/fcvm.2025.1510082 Lu T, Chen F (2012) Multiwfn: a multifunctional wavefunction analyzer. J Comput Chem 33(5):580–592. https://doi.org/10.1002/jcc.22885 Morrissey MA, Kern N, Vale RD (2020) CD47 Ligation Repositions the Inhibitory Receptor SIRPA to Suppress Integrin Activation and Phagocytosis. Immunity 53(2):290–302e296. https://doi.org/10.1016/j.immuni.2020.07.008 Nieto-Garai JA, Lorizate M, Contreras FX (2022) Shedding light on membrane rafts structure and dynamics in living cells. Biochim et Biophys Acta (BBA) - Biomembr 1864(1):183813. https://doi.org/10.1016/j.bbamem.2021.183813 Pothinam S, Putpim C, Siriwoharn T, Jirarattanarangsri W (2025) Effects of Perilla frutescens Seed Oil on Blood Lipids, Oxidative Stress, and Inflammation in Hyperlipidemic Rats. Foods 14. https://doi.org/10.3390/foods14081380 Qiu B, Yuan P, Du X, Jin H, Du J, Huang Y (2023) Hypoxia inducible factor-1α is an important regulator of macrophage biology. Heliyon 9(6):e17167. https://doi.org/https://doi.org/10.1016/j.heliyon.2023.e17167 Raju S, Turner M, Cao C, Abdul-Samad M, Punwasi N, Blaser M, Cahalane R, Botts S, Prajapati K, Patel S, Wu R, Gustafson D, Galant N, Fiddes L, Chemaly M, Hedin U, Matic L, Seidman M, Subasri V, Howe K (2024) Multiomics unveils extracellular vesicle-driven mechanisms of endothelial communication in human carotid atherosclerosis. bioRxiv. https://doi.org/10.1101/2024.06.21.599781 Raju S, Turner M, Cao C, Abdul-Samad M, Punwasi N, Blaser M, Cahalane R, Botts S, Prajapati K, Patel S, Wu R, Gustafson D, Galant N, Fiddes L, Chemaly M, Hedin U, Matic L, Seidman M, Subasri V, Howe K (2025) Multiomic Landscape of Extracellular Vesicles in Human Carotid Atherosclerotic Plaque Reveals Endothelial Communication Networks. Arterioscler Thromb Vasc Biol 45:1277–1305. https://doi.org/10.1161/atvbaha.124.322324 Ren P, Cao J-L, Lin P-L, Cao B, Chen J, Gao K, Zhang J (2021) [Molecular mechanism of luteolin regulating lipoxygenase pathway against oxygen-glucose deprivation/reperfusion injury in H9c2 cardiomyocytes based on molecular docking]. Zhongguo Zhong yao za zhi = Zhongguo zhongyao zazhi = China journal of Chinese materia medica , 46 21 , 5665–5673. https://doi.org/10.19540/j.cnki.cjcmm.20210805.701 Riccioni G, Zanasi A, Vitulano N, Mancini B, D'Orazio N (2009) Leukotrienes in atherosclerosis: new target insights and future therapy perspectives. Mediators Inflamm , 2009 , 737282. https://doi.org/10.1155/2009/737282 Ru J, Li P, Wang J, Zhou W, Li B, Huang C, Li P, Guo Z, Tao W, Yang Y, Xu X, Li Y, Wang Y, Yang L (2014) TCMSP: a database of systems pharmacology for drug discovery from herbal medicines. J Cheminform 6:13. https://doi.org/10.1186/1758-2946-6-13 Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13(11):2498–2504. https://doi.org/10.1101/gr.1239303 Stank A, Kokh DB, Fuller JC, Wade RC (2016) Protein Binding Pocket Dynamics. Acc Chem Res 49(5):809–815. https://doi.org/10.1021/acs.accounts.5b00516 Su C, Mo J, Dong S, Liao Z, Zhang B-X, Zhu P (2024) Integrinβ-1 in disorders and cancers: molecular mechanisms and therapeutic targets. Cell Communication Signaling: CCS 22. https://doi.org/10.1186/s12964-023-01338-3 Szklarczyk D, Gable AL, Nastou KC, Lyon D, Kirsch R, Pyysalo S, Doncheva NT, Legeay M, Fang T, Bork P, Jensen LJ, von Mering C (2021) The STRING database in 2021: customizable protein-protein networks, and functional characterization of user-uploaded gene/measurement sets. Nucleic Acids Res 49(D1):D605–d612. https://doi.org/10.1093/nar/gkaa1074 Tang L, Li Y, Zhong C, Deng X, Wang X (2021) Plant Sterol Clustering Correlates with Membrane Microdomains as Revealed by Optical and Computational Microscopy. Membranes , 11 . https://doi.org/10.3390/membranes11100747 Tang S, Yang J, Xiao B, Wang Y, Lei Y, Lai D, Qiu Q (2024) Aberrant lipid metabolism and complement activation in age-related macular degeneration. Investig Ophthalmol Vis Sci 65(12):20–20. https://doi.org/10.1167/iovs.65.12.20 Tzec-Interián JA, González‐Padilla D, Góngora‐Castillo EB (2025) Bioinformatics perspectives on transcriptomics: A comprehensive review of bulk and single‐cell RNA sequencing analyses. Quant Biology 13(2):e78. https://doi.org/10.1002/qub2.78 Wang B, Jiang T, Qi Y, Luo S, Xia Y, Lang B, Zhang B, Zheng S (2025) AGE-RAGE Axis and Cardiovascular Diseases: Pathophysiologic Mechanisms and Prospects for Clinical Applications. Cardiovasc Drugs Ther 39(6):1489–1506. https://doi.org/10.1007/s10557-024-07639-0 Wang H, Zheng H, Yuheng (2020) Drug treatment of ankylosing spondylitis and related complications: an overlook review. Annals Palliat Med. https://doi.org/10.21037/apm-20-277 Wang Z, Huang Y, Guo Z, Sun J, Zheng G (2025) Interferon-Linked Lipid and Bile Acid Imbalance Uncovered in Ankylosing Spondylitis in a Sibling-Controlled Multi-Omics Study. Int J Mol Sci 26. https://doi.org/10.3390/ijms26167919 Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, Fu X, Liu S, Bo X, Yu G (2021) clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innov (Camb) 2(3):100141. https://doi.org/10.1016/j.xinn.2021.100141 Wu X, Dong S, Chen H, Guo M, Sun Z, Luo H (2023) Perilla frutescens frutescens: A traditional medicine and food homologous plant. Chin Herb Med 15(3):369–375. https://doi.org/10.1016/j.chmed.2023.03.002 Wu Y-T, Chen L, Tan Z-B, Fan H-J, Xie L-P, Zhang W-T, Chen H-M, Li J, Liu B, Zhou Y (2018) Luteolin Inhibits Vascular Smooth Muscle Cell Proliferation and Migration by Inhibiting TGFBR1 Signaling. Frontiers in Pharmacology , 9 . https://doi.org/10.3389/fphar.2018.01059 Wu Y, Wu Y, Xia S, Lian H, Lou Y, Wang L-J (2025) JMJD6-driven epigenetic activation of COL4A2 reprograms glioblastoma vascularization via integrin α1β1-dependent PI3K/MAPK signaling. Acta Neuropathol Commun 13. https://doi.org/10.1186/s40478-025-02114-9 Xia F, Wang C, Jin Y, Liu Q, Meng Q, Liu K, Sun H (2014) Luteolin protects HUVECs from TNF-α-induced oxidative stress and inflammation via its effects on the Nox4/ROS-NF-κB and MAPK pathways. J Atheroscler Thromb 21(8):768–783. https://doi.org/10.5551/jat.23697 Xiong J, Li Z, Tang H, Duan Y, Ban X, Xu K-K, Guo Y, Tu Y (2023) Bulk and single-cell characterisation of the immune heterogeneity of atherosclerosis identifies novel targets for immunotherapy. BMC Biol 21. https://doi.org/10.1186/s12915-023-01540-2 Xu J, Zhou H, Cheng Y, Xiang G (2022) Identifying potential signatures for atherosclerosis in the context of predictive, preventive, and personalized medicine using integrative bioinformatics approaches and machine-learning strategies. EPMA J 13:433–449. https://doi.org/10.1007/s13167-022-00289-y Yim A, Smith C, Brown A (2022) Osteopontin/secreted phosphoprotein-1 harnesses glial‐, immune‐, and neuronal cell ligand‐receptor interactions to sense and regulate acute and chronic neuroinflammation. Immunol Rev 311:224–233. https://doi.org/10.1111/imr.13081 Yuan L, Zhang F, Jia S, Xie J, Shen M (2020) Differences between phytosterols with different structures in regulating cholesterol synthesis, transport and metabolism in Caco-2 cells. J Funct Foods 65:103715. https://doi.org/10.1016/j.jff.2019.103715 Zhang Z, Chen Y, Fu X, Chen L, Wang J, Zheng Q, Zhang S, Zhu X (2024) Identification of PPARG as key gene to link coronary atherosclerosis disease and rheumatoid arthritis via microarray data analysis. PLoS ONE 19(4):e0300022. https://doi.org/10.1371/journal.pone.0300022 Zhao T, Li Z, Ji S, Huang Q, Sun C, Lu B (2025) Decoding the mechanism of dietary fatty acids-driven phytosterol esterification promoting intestinal absorption. Food Chem 496:146525. https://doi.org/10.1016/j.foodchem.2025.146525 Zhou J, Wu Z-Y, Zhao P (2024) Luteolin and its antidepressant properties: From mechanism of action to potential therapeutic application. J Pharm Anal 15. https://doi.org/10.1016/j.jpha.2024.101097 Supplementary Files graphicalabstract2.jpg Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 09 Apr, 2026 Reviewers invited by journal 08 Apr, 2026 Editor assigned by journal 06 Apr, 2026 First submitted to journal 31 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9174552","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":619587562,"identity":"3fc13ae8-a0ef-482f-ba0b-14fefbbb12d9","order_by":0,"name":"chenchen yang","email":"","orcid":"","institution":"Institute of Food Science","correspondingAuthor":false,"prefix":"","firstName":"chenchen","middleName":"","lastName":"yang","suffix":""},{"id":619587563,"identity":"427b1f4e-c7e1-4450-a5f3-acc296a42215","order_by":1,"name":"Jianrong Xing","email":"","orcid":"","institution":"Institute of Food Science","correspondingAuthor":false,"prefix":"","firstName":"Jianrong","middleName":"","lastName":"Xing","suffix":""},{"id":619587564,"identity":"607a2b6e-625e-40d3-8f17-fb498d062bdc","order_by":2,"name":"Mengzhu Wang","email":"","orcid":"","institution":"Institute of Food Science","correspondingAuthor":false,"prefix":"","firstName":"Mengzhu","middleName":"","lastName":"Wang","suffix":""},{"id":619587565,"identity":"ff7257bc-a39d-4c7b-b6df-c365d6c09458","order_by":3,"name":"Wanyi Zhou","email":"","orcid":"","institution":"Institute of Food Science","correspondingAuthor":false,"prefix":"","firstName":"Wanyi","middleName":"","lastName":"Zhou","suffix":""},{"id":619587566,"identity":"efe12d25-a7d5-4d4a-b60a-4a04522a598b","order_by":4,"name":"Ying Yang","email":"","orcid":"","institution":"Institute of Food Science","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Yang","suffix":""},{"id":619587567,"identity":"e5618172-25c2-44fa-8c55-459563573529","order_by":5,"name":"Wenyang Tao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYJADxgcJFTUkqOdhYGA2eHDmGGla2CQftjATVmlwvPeYxM8dtQz27IePVSQ2sDHwt3cn4Ndy5lyaZO+Z4ww8PGlpNxJ3yDBInDm7Aa8Wsxs5ZhK8bceADssxu5F4ho3BQCKXgJb7b8wk/4K08L8xK0hsYyZCyw0eM2nethoGHokcMwaitNifyTG2lm07wMBz41myRMKZYzwE/SLZfsbw5tu2Ogb2/uSDH39U1Mjxt/fi1wIELBIMDIfrG6A8HkLKQYD5AwNDHTEKR8EoGAWjYKQCAEw+Rw29jrBdAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0001-7762-8929","institution":"Zhejiang Academy of Agricultural Sciences","correspondingAuthor":true,"prefix":"","firstName":"Wenyang","middleName":"","lastName":"Tao","suffix":""}],"badges":[],"createdAt":"2026-03-20 04:01:48","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9174552/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9174552/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107483296,"identity":"3f8b0e36-2a89-4eb9-b46b-394c5516cb4e","added_by":"auto","created_at":"2026-04-22 02:27:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3109048,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork Pharmacology Analysis of Perilla frutescens Seeds Against Atherosclerosis. (A) Venn diagram of intersecting targets. (B) PPI network construction. (C) Screening of the core target sub-network. (D) KEGG pathway enrichment. (E) GO functional annotation (BP, CC, and MF).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/25afa813d906df2bf17b981f.png"},{"id":107103131,"identity":"516e71ae-3ca6-4915-865a-39d500fae8eb","added_by":"auto","created_at":"2026-04-16 19:48:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":4086758,"visible":true,"origin":"","legend":"\u003cp\u003eQuality Control, Clustering, and Cell Type Annotation of scRNA-seq Data. (A) Quality control metrics of samples. (B) Identification of highly variable genes. (C) UMAP projection of cell clusters. (D) Annotation of major cell types. (E) Dot plot of marker gene expression.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/7f66d56ffca742109c4c1255.png"},{"id":107103133,"identity":"ef2fe8fd-75e4-4264-afcb-7dff0d5e6b08","added_by":"auto","created_at":"2026-04-16 19:48:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2724634,"visible":true,"origin":"","legend":"\u003cp\u003eMachine Learning Screening and Validation of Core Targets. (A) LASSO feature selection process. (B) Variable importance scores from Random Forest analysis. (C) ROC diagnostic performance analysis in the external dataset. (D) Expression distribution of core targets in the single-cell landscape.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/98a0ce5ce901eb9e67503db0.png"},{"id":107483054,"identity":"992bbb7b-186e-4a92-b788-87e41f724c95","added_by":"auto","created_at":"2026-04-22 02:26:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3725036,"visible":true,"origin":"","legend":"\u003cp\u003ePseudotime Trajectory and Cell-Cell Communication Analysis. (A) Pseudotime expression dynamics of core targets. (B) ITGB1-based ligand-receptor interaction dot plot. (C) Overview of the intercellular communication network.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/2d2bf478d0226f3b6968ac74.png"},{"id":107103136,"identity":"a4f629ac-7417-4725-b0b4-7af6f8e542bf","added_by":"auto","created_at":"2026-04-16 19:48:57","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":5147106,"visible":true,"origin":"","legend":"\u003cp\u003eMolecular Interaction and Dynamic Stability Analysis. (A) Heatmap screening of binding affinities. (B) Molecular docking modes of Luteolin with ALOX5 and ITGB1. (C-F) 100-ns MD simulation analysis of the Luteolin-ITGB1 complex, including RMSD (C), RMSF (D), Rg (E), and SASA (F).\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/cbca0ecf8e2cfeb00b44e3df.png"},{"id":107485913,"identity":"f87a005f-f2f4-419f-9ce6-6e3e339cce31","added_by":"auto","created_at":"2026-04-22 02:36:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":19318779,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/e5700cd7-519d-46ad-879d-d0c6333ba696.pdf"},{"id":107103135,"identity":"035c2de0-b816-4e20-874f-72668ff8d601","added_by":"auto","created_at":"2026-04-16 19:48:57","extension":"jpg","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":118348,"visible":true,"origin":"","legend":"","description":"","filename":"graphicalabstract2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9174552/v1/4f1d5324545af75c5a22a073.jpg"}],"financialInterests":"","formattedTitle":"Unveiling Anti-atherosclerotic Targets of Perilla frutescens through a Multi-scale Computational Framework Integrating Network Pharmacology, Single-cell Analysis, Machine Learning, and Molecular Dynamics","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAtherosclerosis (AS) represents a chronic progressive pathological condition fundamentally characterized by arterial intimal lipid accumulation and maladaptive immunological responses. Despite the demonstrated efficacy of lipid-lowering pharmacotherapy, particularly statins and proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors, substantial residual inflammatory risk (RIR) persists, perpetuating plaque progression and precipitating acute cardiovascular events (Di Muro et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Emerging mechanistic investigations have revealed that AS is orchestrated by intricate immunometabolic networks, wherein dysregulated crosstalk between lipid metabolism and inflammatory signaling establishes a homeostatic pathological architecture exhibiting pronounced resistance to monotherapeutic interventions (Z. Wang et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Terapeutic paradigms necessitate a strategic transition from linear pathway inhibition toward systemic network perturbation to enable synergistic modulation of multiple pivotal nodes within the atherosclerotic plaque microenvironment.\u003c/p\u003e \u003cp\u003eBotanical therapeutics, by virtue of their multi-component synergistic architecture, constitute a promising reservoir of bioactive compounds for network-based modulation. \u003cem\u003ePerilla frutescens\u003c/em\u003e frutescens seeds, a quintessential functional food with medicinal attributes, harbor an abundance of pleiotropic bioactive constituents encompassing α-linolenic acid, phytosterols, and flavonoids such as luteolin. While prior phenotypic investigations have established the therapeutic efficacy of \u003cem\u003ePerilla frutescens\u003c/em\u003e in ameliorating dyslipidemia and suppressing vascular inflammation, the underlying molecular mechanisms remain incompletely elucidated (Wu et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Specifically, the precise cellular targeting and mechanistic execution of vasculoprotective effects by these bioactive constituents within the heterogeneous cellular landscape of atherosclerotic plaques remain undefined.\u003c/p\u003e \u003cp\u003eConventional systems pharmacology predominantly relies on bulk tissue transcriptomic profiling, the inherent methodological constraints of which substantially impede mechanistic interrogation at cellular resolution. This aggregate approach fundamentally obscures the spatiotemporal heterogeneity intrinsic to atherosclerotic plaques, precluding effective discrimination between pathogenic cellular subsets (e.g., specific foam cell phenotypes) and bystander populations (Xiong et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Moreover, plaque destabilization is orchestrated through intricate intercellular communication networks, including ligand\u0026ndash;receptor interactions governing cell adhesion and migration, the architectural delineation of which remains intractable without single-cell resolution. Consequently, critical knowledge gaps persist regarding which cellular differentiation trajectories are targeted by \u003cem\u003ePerilla frutescens\u003c/em\u003e and the mechanisms through which it disrupts pathological intercellular crosstalk driving AS progression.\u003c/p\u003e \u003cp\u003eTo address these knowledge deficits, we established an integrative computational framework synergizing single-cell RNA sequencing (scRNA-seq) with machine learning (ML) algorithms. We employed an ensemble feature selection strategy coupling least absolute shrinkage and selection operator (LASSO) regression with random forest (RF) algorithms to rigorously identify robust core targets from high-dimensional single-cell datasets, thereby mitigating transcriptomic noise and overfitting artifacts. Furthermore, molecular docking coupled with molecular dynamics (MD) simulations validated the thermodynamic stability and conformational dynamics of predicted drug\u0026ndash;target complexes under quasi-physiological solvation conditions.\u003c/p\u003e \u003cp\u003eLeveraging this strategy, we successfully deconvoluted the cellular heterogeneity landscape of human atherosclerotic plaques and identified a core genetic signature encompassing HIF1A, PPARG, and ITGB1. Critically, our investigation revealed high-affinity binding between luteolin, the principal bioactive constituent of \u003cem\u003ePerilla frutescens\u003c/em\u003e, and integrin β1 (ITGB1), whereby conformational stabilization via a \"conformational-locking\" mechanism abrogates the SPP1\u0026ndash;ITGB1 signaling axis, which serves as a pivotal conduit mediating foam cell adhesion and inflammatory activation. Collectively, these findings delineate a high-resolution molecular cartography of \u003cem\u003ePerilla frutescens\u003c/em\u003e-mediated AS intervention, underscoring the translational potential of targeting cell adhesion receptors for mitigating residual vascular inflammation.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Screening of Bioactive Constituents and Target Prediction for \u003cem\u003ePerilla frutescens\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eChemical constituents of \u003cem\u003ePerilla frutescens\u003c/em\u003e were systematically retrieved from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://tcmsp-e.com\u003c/span\u003e\u003cspan address=\"https://tcmsp-e.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) using \"\u003cem\u003ePerilla frutescens\u003c/em\u003e\" as the query keyword (Ru et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Candidate bioactive constituents were filtered based on conventional absorption, distribution, metabolism, and excretion (ADME) criteria, with selection thresholds established at oral bioavailability (OB)\u0026thinsp;\u0026ge;\u0026thinsp;30% and drug-likeness (DL)\u0026thinsp;\u0026ge;\u0026thinsp;0.18. Redundant entries were eliminated, and compound identifiers were standardized to facilitate downstream target prediction analyses. Putative molecular targets of candidate constituents were predicted using SwissTargetPrediction, Similarity Ensemble Approach (SEA), and SuperPred, with species parameters restricted to \u003cem\u003eHomo sapiens\u003c/em\u003e. Prediction outputs from these platforms were subsequently merged, deduplicated, and subjected to standardized mapping via the UniProt database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.uniprot.org\u003c/span\u003e\u003cspan address=\"https://www.uniprot.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for uniform conversion to gene symbol nomenclature.\u003c/p\u003e \u003cp\u003eAS-associated disease genes were retrieved from GeneCards (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.genecards.org\u003c/span\u003e\u003cspan address=\"https://www.genecards.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), Online Mendelian Inheritance in Man (OMIM; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.omim.org\u003c/span\u003e\u003cspan address=\"https://www.omim.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), and DisGeNET (v7.0; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.disgenet.org\u003c/span\u003e\u003cspan address=\"https://www.disgenet.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) using \"Atherosclerosis\" as the query term. Selection criteria comprised: relevance score\u0026thinsp;\u0026ge;\u0026thinsp;10 for GeneCards entries, gene-disease association score (Score_gda)\u0026thinsp;\u0026ge;\u0026thinsp;0.10 for DisGeNET entries, and explicitly annotated AS-associated genes from OMIM. Disease-associated genes retrieved from these three repositories were consolidated and deduplicated to establish a comprehensive disease gene set.\u003c/p\u003e \u003cp\u003eIntersection analysis between the \u003cem\u003ePerilla frutescens\u003c/em\u003e candidate target gene set and the AS disease gene set was subsequently performed to identify putative therapeutic targets. Set intersection operations and visualization were executed in R software (v4.3.2) using the ggVennDiagram package (v1.2.2) for Venn diagram construction. The resultant intersecting targets were further subjected to protein\u0026ndash;protein interaction (PPI) network construction, functional enrichment analysis, and integrative prioritization with single-cell transcriptomic differential expression profiles.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. PPI Network Construction, GO and KEGG Enrichment Analyses\u003c/h2\u003e \u003cp\u003eThe PPI network of intersecting targets was constructed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (v11.5; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://string-db.org\u003c/span\u003e\u003cspan address=\"https://string-db.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) (Szklarczyk et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Analytical parameters were configured as follows: species restriction to Homo sapiens and minimum interaction confidence threshold set to high confidence (combined score\u0026thinsp;\u0026gt;\u0026thinsp;0.7). PPI outputs were subsequently imported into Cytoscape (v3.9.1) for network visualization and topological characterization, with isolated nodes excluded to generate a connected network for downstream analyses (Shannon et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFunctional enrichment analyses were conducted in the R environment utilizing the clusterProfiler package (v4.10.0), with gene annotation information sourced from the org.Hs.eg.db database (v3.18.0) (Wu et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). To enhance annotation consistency and analytical robustness, gene symbols were initially converted to Entrez Gene IDs, followed by Gene Ontology (GO) enrichment analysis across three ontological dimensions, including biological process (BP), cellular component (CC), and molecular function (MF), as well as Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Multiple hypothesis testing correction was implemented via the Benjamini\u0026ndash;Hochberg (BH) procedure, with the significance threshold established at false discovery rate (FDR, p.adjust)\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Enrichment outputs were visualized as bubble plots to systematically delineate the core biological processes and pivotal signaling cascades engaged by \u003cem\u003ePerilla frutescens\u003c/em\u003e putative therapeutic targets.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Single-Cell Transcriptomic Data Acquisition and Preprocessing\u003c/h2\u003e \u003cp\u003eThe human atherosclerotic plaque single-cell transcriptomic dataset GSE159677 was retrieved from the Gene Expression Omnibus (GEO) repository, with extraction of publicly available gene\u0026ndash;cell expression matrices and associated metadata files (Alsaigh et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). All computational analyses were executed in the R environment (v4.3.2), with single-cell data preprocessing and downstream analytical workflows primarily implemented using the Seurat package (v4.3.0). Seurat objects were constructed from expression matrices, followed by implementation of stringent quality control (QC) procedures to eliminate low-quality cells and mitigate potential technical noise artifacts (Hao et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSubsequently, doublet artifacts were identified and excluded using DoubletFinder (v2.0.3). Parametric configurations were determined based on sample-specific cell counts and standardized protocols, with anticipated doublet rates established according to 10x Genomics empirical recommendations and optimal parameter combinations refined through pK value sweep optimization. QC-filtered data underwent normalization via the Seurat NormalizeData function, followed by identification of highly variable genes (HVGs) using the FindVariableFeatures function. Data scaling and standardization were subsequently performed using the ScaleData function, with regression of technical covariates (including sequencing depth and mitochondrial gene fraction) to mitigate batch-associated technical variation. Principal component analysis (PCA) was executed via the RunPCA function, with optimal dimensionality for downstream analyses determined through integrative assessment of ElbowPlot visualization, JackStraw permutation testing, and cumulative variance explained metrics. To mitigate potential confounding effects of batch variation on cellular clustering architecture, multi-sample integration was performed using the Seurat integration workflow, with subsequent construction of k-nearest neighbor (kNN) graphs and execution of clustering algorithms within the integrated low-dimensional embedding space.\u003c/p\u003e \u003cp\u003eCell type annotation was manually assigned based on canonical marker gene expression profiles, with cross-validation performed through visualization modalities including FeaturePlot and DotPlot representations. Upon completion of cell type annotation, differentially expressed genes (DEGs) were identified using the FindMarkers function (default: Wilcoxon rank-sum test), with multiple testing correction implemented via the Benjamini\u0026ndash;Hochberg (BH) procedure. DEG selection criteria comprised FDR (p.adjust)\u0026thinsp;\u0026lt;\u0026thinsp;0.05, with the resultant DEG signature subjected to integrative cross-validation against network pharmacology-derived candidate targets for identification of core therapeutic nodes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Machine Learning-Based Feature Selection for Disease State Classification\u003c/h2\u003e \u003cp\u003eFeature selection was performed using two complementary machine learning algorithms, namely least absolute shrinkage and selection operator (LASSO) regression and random forest (RF). To circumvent pseudo-replication artifacts inherent to single-cell data structures, pre-modeling aggregation of single-cell expression profiles was performed using a pseudo-bulk strategy across \"sample \u0026times; cell type\" dimensions, yielding gene expression matrices with biological samples as statistical units. Disease state labels were derived from clinical stratification metadata accompanying the GEO dataset. Training\u0026ndash;validation partitioning was executed at the sample level, ensuring mutually exclusive sample allocation to preclude data leakage artifacts.\u003c/p\u003e \u003cp\u003eLASSO regression modeling was implemented using the glmnet package (v4.1-8), with disease state configured as a binary outcome variable and model family specified as binomial. Following feature standardization, optimal regularization parameter λ was determined via 10-fold cross-validation, with lambda.1se adopted as the penalty coefficient. Genes exhibiting non-zero regression coefficients at the selected λ value constituted the LASSO-derived feature gene set. Random forest classification models were trained using the randomForest package (v4.7-1.1), with computation of variable importance metrics for individual features. Given potential selection bias inherent to Gini index-derived variable importance, permutation importance was adopted as the primary evaluation metric. Intersection analysis between the LASSO-derived feature set and the top 20 genes ranked by random forest variable importance yielded the core feature gene signature. In instances of insufficient intersection cardinality, the union of both gene sets was employed, with subsequent refinement through external validation and model performance assessment. Candidate feature genes were ranked in descending order of composite importance scores, with the top 20 genes prioritized for downstream analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Biological Weighted Expression Score (WES) Validation\u003c/h2\u003e \u003cp\u003eTo quantitatively assess the transcriptional activity of candidate genes within key cellular subsets, gene set module scores and weighted WES were computed within the Seurat analytical framework. Initially, module scores for candidate gene sets were calculated using the Seurat AddModuleScore function to quantify relative transcriptional intensity at single-cell resolution, with background gene set normalization correction.\u003c/p\u003e \u003cp\u003eBuilding upon this foundation, a WES was formulated to integrate both expression abundance and detection frequency within specific cellular subsets:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:WES=Average\\:Expression\\:\\times\\:Detection\\:Rate\\:(Pct.\\text{E}\\text{x}\\text{p})$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eHere, Average Expression denotes the mean normalized expression level within the target cellular subset (computed via the Seurat AverageExpression function from LogNormalize-standardized expression matrices), while Detection Rate represents the detection frequency (pct.exp, defined as the proportion of cells exhibiting expression values\u0026thinsp;\u0026gt;\u0026thinsp;0). WES values were computed for each candidate gene across individual cellular subsets and ranked in descending order. Subsequently, top-ranked WES genes were subjected to intersection analysis with machine learning-derived feature signatures, with the top 10 WES-ranked intersecting genes designated as core therapeutic targets.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Spatial and Dynamic Expression Validation of Core Targets at Single-Cell Resolution\u003c/h2\u003e \u003cp\u003eTo assess the spatial distribution architecture of core candidate targets across cellular subsets, gene expression was projected onto uniform manifold approximation and projection (UMAP) low-dimensional embeddings using the Seurat FeaturePlot function, with expression heterogeneity across cell types and subpopulations visualized via DotPlot and VlnPlot representations. Subsequently, myeloid lineage cells, encompassing monocytes, macrophages, and foam cell subsets, were subjected to pseudotemporal trajectory reconstruction analysis. Trajectory inference was performed using the Monocle3 package (v1.3.1) (Cao et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Trajectory root cells were designated based on cellular subsets exhibiting elevated expression of early macrophage markers (e.g., LST1, FCER1G), with pseudotemporal ordering established via root cell specification through the order_cells() function. Expression dynamics of core targets along the pseudotemporal axis were visualized and subjected to trend-fitting analyses to characterize their transcriptional trajectories during macrophage-to-foam cell differentiation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7. Clinical Diagnostic Performance Validation (ROC Analysis)\u003c/h2\u003e \u003cp\u003eTo evaluate the clinical diagnostic performance of core candidate targets, an independent external validation dataset GSE100927 (n\u0026thinsp;=\u0026thinsp;104) was retrieved from the GEO repository. Atherosclerotic pathological specimens and healthy controls were extracted based on original sample annotations to establish binary classification labels. For microarray platforms, sequential application of background correction, quantile normalization, and log₂ transformation was performed; in instances of multiple probe-to-gene mappings, probe expression values were aggregated to the gene level via median (or mean) summarization to construct gene expression matrices. Subsequently, expression profiles of core genes within the validation cohort were extracted for downstream model construction and performance assessment.\u003c/p\u003e \u003cp\u003eMultivariable logistic regression modeling was performed based on core gene expression profiles, with receiver operating characteristic (ROC) curves generated using the pROC package (v1.18.5). Area under the curve (AUC) metrics and corresponding 95% confidence intervals (CIs) were computed via the DeLong method. To mitigate overfitting-associated performance inflation bias, 10-fold cross-validation was implemented within the external validation cohort, with computation of mean AUC values and assessment of model discriminatory capacity across clinical stratifications.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.8. Molecular Docking\u003c/h2\u003e \u003cp\u003eThree-dimensional structures of core target proteins were retrieved from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB), with selection criteria comprising: Homo sapiens species restriction, crystallographic resolution\u0026thinsp;\u0026le;\u0026thinsp;2.5 \u0026Aring;, and preferential selection of structures harboring co-crystallized ligands. In instances of missing residues or incomplete side-chain geometries, structural refinement was performed using PDBFixer (v1.9) for residue reconstruction and conformational optimization. Receptor proteins underwent processing in PyMOL (v2.5.5) for removal of redundant water molecules while preserving critical metal ions and cofactors, followed by polar hydrogen addition, Gasteiger partial charge assignment, and PDBQT format conversion using AutoDockTools (ADT, v1.5.7) (Eberhardt et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Ligand molecular structures were retrieved from the PubChem database, with three-dimensional conformer generation and geometric energy minimization performed using Open Babel (v3.1.1).\u003c/p\u003e \u003cp\u003eMolecular docking simulations were executed using AutoDock Vina (v1.2.5), with docking grid centers defined based on co-crystallized ligand coordinates or computationally predicted active site pocket topologies. Docking parameters were configured as follows: exhaustiveness\u0026thinsp;=\u0026thinsp;32, num_modes\u0026thinsp;=\u0026thinsp;10, energy_range\u0026thinsp;=\u0026thinsp;3. Docking poses were ranked according to the Vina scoring function, with the lowest-energy conformation selected for protein\u0026ndash;ligand interaction analysis and visualization in PyMOL.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.9. MD Simulations\u003c/h2\u003e \u003cp\u003eMD simulations were executed using the GROMACS software package (v2024.3) (Abraham et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Ligand partial charges were derived via restrained electrostatic potential (RESP/RESP2) fitting of ORCA-computed wavefunctions using the Multiwfn program, with force field topology files generated via the Sobtop utility based on the general AMBER force field (GAFF) (Lu \u0026amp; Chen, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Protein parameterization employed the CHARMM36 force field, with explicit solvation maintained throughout via the TIP3P water model. Protein\u0026ndash;ligand complexes were positioned within rhombic dodecahedral simulation boxes, with a minimum solute-to-box boundary distance of 1.2 nm. Following explicit solvation, Na⁺ and Cl⁻ counterions were introduced to achieve charge neutralization, with ionic strength adjusted to 0.15 M to recapitulate physiological salinity.\u003c/p\u003e \u003cp\u003eSystems underwent initial energy minimization via the steepest descent algorithm. Subsequently, equilibration simulations were performed sequentially under NVT ensemble conditions (100 ps, velocity-rescaling temperature coupling) and NPT ensemble conditions (100 ps, Parrinello\u0026ndash;Rahman pressure coupling), with positional restraints of 1000 kJ\u0026middot;mol⁻\u0026sup1;\u0026middot;nm⁻\u0026sup2; applied to protein backbone and ligand heavy atoms. Production-phase simulations were conducted with removal of all positional restraints, employing a 2-fs integration timestep over 100 ns. All hydrogen-containing covalent bonds were constrained via the linear constraint solver (LINCS) algorithm, with long-range electrostatic interactions computed using the particle mesh Ewald (PME) method; van der Waals and short-range electrostatic cutoff radii were uniformly set to 1.0 nm, with neighbor list updates performed every 20 timesteps. Post-trajectory analyses encompassed temporal evolution of root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), and protein\u0026ndash;ligand interfacial hydrogen bond occupancy, enabling systematic assessment of complex conformational stability and binding interface persistence.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.10. Statistical Analysis\u003c/h2\u003e \u003cp\u003eStatistical analyses were performed in R. Intergroup comparisons were conducted using two-sample t-tests or Wilcoxon rank-sum tests, with multiple testing correction implemented via the Benjamini\u0026ndash;Hochberg procedure (FDR threshold: \u003cem\u003ep.adjust\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). ROC analyses employed the pROC package for computation of AUC values and 95% confidence intervals via the DeLong method. Unless otherwise specified, all statistical tests were two-tailed, with \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Screening of Bioactive Constituents and Identification of Putative Therapeutic Targets in \u003cem\u003ePerilla frutescens\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eA network pharmacology approach was employed to systematically dissect the pharmacodynamic material basis underlying the therapeutic efficacy of Perilla frutescens. 16 core bioactive constituents exhibiting favorable pharmacokinetic properties were identified from \u003cem\u003ePerilla frutescens\u003c/em\u003e using ADME filtering criteria of oral bioavailability (OB)\u0026thinsp;\u0026ge;\u0026thinsp;30% and DL\u0026thinsp;\u0026ge;\u0026thinsp;0.18 within the TCMSP database framework. As illustrated in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, these constituents collectively exhibited elevated bioavailability potential and favorable drug-like structural attributes. Luteolin (MOL000006), a flavonoid constituent prioritized for downstream investigation, demonstrated balanced pharmacokinetic parameters indicative of favorable membrane permeability and intestinal absorption capacity. Phytosterol constituents including β-sitosterol (MOL000358) and stigmasterol (MOL000449) exhibited OB values exceeding 36%, with maximal values reaching 43.83%, underscoring the pivotal contribution of lipophilic compounds to \u003cem\u003ePerilla frutescens\u003c/em\u003e pharmacological repertoire.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePhysicochemical Properties and ADME Evaluation Metrics of Candidate Active Compounds in \u003cem\u003ePerilla frutescens\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMol ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMolecule Name\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMW\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAlogP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOB (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDL\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL000006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eluteolin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e286.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e36.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL000358\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ebeta-sitosterol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e414.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e36.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL000449\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStigmasterol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e412.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e43.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL000953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e386.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e37.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL001439\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003earachidonic acid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e304.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e45.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL002773\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ebeta-carotene\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e536.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e37.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL004355\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSpinasterol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e412.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e42.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL005030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003egondoic acid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e310.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e30.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL005043\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ecampest-5-en-3beta-ol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e400.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e37.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.71\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL005481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,6,10,14,18-pentamethylicosa-2,6,10,14,18-pentaene\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e342.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e33.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.24\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL007449\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24-methylidenelophenol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e412.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e44.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL009653\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCycloeucalenol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e426.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e39.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL009681\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eObtusifoliol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e426.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e42.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL012888\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ecitrostadienol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e426.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e43.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL012891\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2E,4E,6E)-icosa-2,4,6-trienoic acid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e306.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e41.64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMOL012893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(E)-(4-methylbenzylidene)-(4-phenyltriazol-1-yl)amine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e262.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e57.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.19\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe integration of multiple target prediction platforms, namely SwissTargetPrediction, SEA, and SuperPred, followed by standardized annotation, yielded 695 putative protein targets potentially interacting with the aforementioned bioactive constituents. Systematic interrogation of GeneCards, OMIM, and DisGeNET repositories established an AS -associated target compendium comprising 2,444 genes.\u003c/p\u003e \u003cp\u003eVenn diagram mapping via the ggVennDiagram package revealed substantial intersection between \u003cem\u003ePerilla frutescens\u003c/em\u003e candidate targets and the AS disease gene set, yielding 289 overlapping genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). These overlapping genes constituted the candidate therapeutic target set for \u003cem\u003ePerilla frutescens\u003c/em\u003e-mediated AS intervention, providing a critical genetic foundation for elucidating molecular connectivity between botanical bioactive constituents and disease pathophysiology. To further dissect synergistic interaction architectures among candidate targets, the 289 overlapping genes were imported into the STRING database (v11.5) for high-confidence protein\u0026ndash;protein interaction (PPI) network construction. Network topology visualization (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB) revealed characteristic scale-free network architecture, with dense physical and functional interconnectivity among nodes corroborating the \"multi-component\u0026ndash;multi-target\u0026ndash;multi-pathway\" synergistic regulatory paradigm of \u003cem\u003ePerilla frutescens\u003c/em\u003e. Furthermore, quantitative topological characterization via Cytoscape, incorporating weighted ranking of topological metrics including maximal clique centrality (MCC) and node degree, identified a core functional module comprising highly connected hub nodes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). These topologically central hub genes exert pivotal regulatory roles within \u003cem\u003ePerilla frutescens\u003c/em\u003e-mediated modulation of atherosclerotic pathological networks.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Gene Enrichment Analysis\u003c/h2\u003e \u003cp\u003eTo systematically elucidate the biological functionalities of the 289 intersecting targets, multidimensional GO functional enrichment and KEGG pathway enrichment analyses were performed. GO enrichment analyses revealed the molecular and cellular functional landscape underlying the putative therapeutic effects of \u003cem\u003ePerilla frutescens\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD). Within the biological process (BP) ontological dimension, target genes exhibited significant enrichment in immune defense responses to lipopolysaccharide (LPS) and bacterial-derived molecular patterns, with pronounced associations to wound healing and inflammatory response modulation. This enrichment architecture suggested that \u003cem\u003ePerilla frutescens\u003c/em\u003e may ameliorate vascular endothelial chronic injury and repair dysregulation through attenuation of pathogen-associated molecular pattern (PAMP)-triggered inflammatory cascades. Additionally, pronounced enrichment of steroid and lipid metabolism-related terms suggested that \u003cem\u003ePerilla frutescens\u003c/em\u003e may counteract atherosclerotic lipid accumulation pathology through restoration of lipid metabolic homeostasis.\u003c/p\u003e \u003cp\u003eWithin the cellular component (CC) ontology, target genes predominantly localized to subcellular structures including membrane rafts, membrane microdomains, and vesicle lumina. Given the pivotal role of membrane rafts in receptor clustering and signaling complex assembly, this enrichment profile indicated that \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents may exert vasculoprotective effects through modulation of membrane-associated signal transduction platforms. Within the molecular function (MF) dimension, target genes demonstrated significant enrichment in nuclear receptor activity, ligand-activated transcription factor activity, and eicosanoid receptor activity. This functional landscape aligned closely with the established regulatory roles of nuclear receptors (e.g., PPARs) in governing lipid metabolism and anti-inflammatory transcriptional programs, suggesting transcriptional-level intervention in AS-associated pathological networks by \u003cem\u003ePerilla frutescens\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eKEGG pathway enrichment analysis further corroborated the mechanistic connectivity between candidate targets and atherosclerotic pathophysiology at the signaling cascade level (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eE). The \"Lipid and atherosclerosis\" pathway exhibited the most pronounced enrichment score and maximal gene ratio, indicating non-random distribution of intersecting target genes with specific convergence upon core atherosclerotic pathological modules. Beyond this primary pathway, target genes demonstrated substantial enrichment in pathological processes encompassing efferocytosis, advanced glycation end product\u0026ndash;receptor for AGE (AGE\u0026ndash;RAGE) signaling axis, and neutrophil extracellular trap (NET) formation. This multi-pathway enrichment architecture suggested vasculoprotective efficacy of \u003cem\u003ePerilla frutescens\u003c/em\u003e through multidimensional synergistic mechanisms: (i) enhancement of macrophage-mediated intraplaque apoptotic cell clearance, mitigating necrotic core formation; (ii) antagonism of AGE-induced oxidative injury; and (iii) suppression of NET-mediated immunothrombotic responses. Additionally, while certain enrichment terms pertained to chemical carcinogenesis and proteoglycan-associated pathways, these signaling modules likely reflect aberrant vascular smooth muscle cell (VSMC) proliferation and pathological extracellular matrix (ECM) remodeling within the atherosclerotic context. Collectively, GO and KEGG enrichment analyses indicated that \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents may attenuate atherosclerotic plaque progression and destabilization through modulation of multiple signaling networks governing cellular proliferation and vascular microenvironmental remodeling.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Single-Cell Transcriptomic Atlas Construction and Plaque Cellular Heterogeneity Delineation\u003c/h2\u003e \u003cp\u003eHigh-resolution single-cell transcriptomic (scRNA-seq) profiling was performed using the GSE159677 dataset, with stringent QC procedures implemented to eliminate low-quality cells and technical noise artifacts. QC metrics revealed uniform distributions of detected gene counts (nFeature_RNA) across all samples (GSM4837523\u0026ndash;GSM4837528), with mitochondrial gene expression fractions consistently maintained below 10% (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), ensuring downstream analyses were predicated upon high-fidelity transcriptomic data. Among the 2,000 HVGs identified, CCL18, APOE, SPP1, and S100A8/A9 exhibited maximal expression variability (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). These HVGs predominantly encompassed chemotactic signaling, lipid trafficking, and inflammatory response mediators, indicating pronounced cellular heterogeneity associated with lipid metabolic dysregulation and robust inflammatory activation within the plaque microenvironment.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSubsequently, UMAP nonlinear dimensionality reduction projected the single-cell transcriptomic landscape onto two-dimensional embeddings, with unsupervised clustering algorithms resolving 26 discrete cellular clusters (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). Based on canonical marker gene expression profiles, 11 major cellular lineages were annotated (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD), encompassing structural vascular wall constituents, such as endothelial cells, smooth muscle cells (SMCs), and fibroblasts, alongside immune infiltrates including T cells, B cells, natural killer (NK) cells, and mast cells. Critically, higher-resolution dissection of myeloid lineage cells was achieved, enabling unambiguous discrimination of foam cells from classical macrophage populations. Foam cells occupied a discrete clustering topology within UMAP embedding space, indicating substantial transcriptional-level phenotypic reprogramming.\u003c/p\u003e \u003cp\u003eTo validate cell type annotation fidelity and characterize lineage-specific molecular signatures, expression distribution patterns of canonical marker genes were visualized via dot plot representation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE). Results revealed distinct and lineage-specific marker gene expression signatures: T cells exhibited selective enrichment of IL7R and CD3D, endothelial cells demonstrated elevated VWF and PECAM1 expression, while smooth muscle cells displayed characteristic ACTA2 and MYH11 upregulation. Critically, foam cells retained myeloid lineage markers (e.g., CD68) while exhibiting pronounced upregulation of inflammation-associated genes including S100A8, S100A12, and MT1G. Given the established roles of these genes in inflammatory signal amplification, metal ion homeostasis regulation, and oxidative stress responses, this differential expression signature indicated that foam cells, following excessive lipid internalization, existed in a pathologically activated state characterized by elevated oxidative stress burden and sustained inflammatory activation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.4. Identification of Core Therapeutic Targets via Machine Learning and Clinical Validation\u003c/h2\u003e \u003cp\u003eTo precisely identify hub genes exhibiting potential clinical diagnostic utility and disease relevance within the intersecting gene set, an integrative feature dimensionality reduction strategy synergizing statistical and machine learning methodologies was implemented. Feature selection was performed via LASSO regression, with regularization parameter λ optimized through 10-fold cross-validation. At the λ value corresponding to binomial deviance minimization, redundant variables were effectively eliminated, thereby mitigating multicollinearity-induced compromise of parameter estimation robustness (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Random forest (RF) modeling assessed individual gene contributions to disease state classification from a nonlinear perspective, with variable importance ranking identifying high-priority feature genes including PPARG, ITGB1, MMP9, and ALOX5 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). This dual-algorithm integrative framework implemented cross-validation of candidate features under divergent modeling assumptions, substantially enhancing feature selection robustness.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIntersection analysis between machine learning-derived feature genes and top-ranked WES genes at single-cell resolution yielded 10 core therapeutic targets: HIF1A, ALOX5, STAT1, ITGB1, PPARG, MMP9, PIK3R1, PRKCB, CDK4, and PIK3CA (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). To assess the translational potential of this core target signature, multivariable logistic regression diagnostic modeling was performed within the independent external validation dataset GSE100927. ROC curve analysis revealed an AUC of 0.996 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC), indicating superior discriminatory capacity of the 10-gene signature for distinguishing atherosclerotic pathological specimens from healthy controls, underscoring substantial clinical diagnostic utility.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eList of Core Candidate Targets Identified Based on Multidimensional Screening Strategies\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSymbol\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFull Name\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFunction\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHIF1A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHypoxia Inducible Factor 1 Alpha\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHypoxia response\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALOX5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eArachidonate 5-Lipoxygenase\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInflammatory mediator\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSTAT1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSignal Transducer and Activator of Transcription 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eImmune signaling\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eITGB1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntegrin Subunit Beta 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCell adhesion\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePPARG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePeroxisome Proliferator-Activated Receptor Gamma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLipid metabolism\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMMP9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMatrix Metallopeptidase 9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eECM degradation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePIK3R1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePhosphoinositide-3-Kinase Regulatory Subunit 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePI3K pathway regulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePRKCB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProtein Kinase C Beta\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSignal transduction\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCDK4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCyclin Dependent Kinase 4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCell cycle control\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePIK3CA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePhosphatidylinositol 3-Kinase Catalytic Subunit Alpha\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCell growth\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTo delineate the cellular origins and spatial distribution patterns of core targets within the plaque microenvironment, the 10 core genes were projected onto single-cell UMAP embedding space for visualization (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD). Results demonstrated non-uniform expression profiles across cellular compartments, with specific enrichment within macrophage and foam cell subsets, contrasting with relatively diminished expression in T cells, B cells, and smooth muscle cells. This cell type-specific enrichment pattern validated the efficacy of the WES selection strategy while revealing that \u003cem\u003ePerilla frutescens\u003c/em\u003e likely exerts vasculoprotective effects primarily through targeted modulation of key molecular networks within intraplaque myeloid lineage cells, encompassing PPARG-mediated lipid metabolic reprogramming, MMP9-mediated extracellular matrix degradation, and HIF1A-governed hypoxic responses, thereby intercepting plaque progression and destabilization trajectories.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e3.5. Pseudotemporal Dynamics and Intercellular Communication Profiling of Core Targets\u003c/h2\u003e \u003cp\u003eSingle-cell pseudotemporal trajectory reconstruction of myeloid lineage cells was performed using the Monocle3 algorithm. Trajectory reconstruction revealed continuous phenotypic evolution of intraplaque macrophages from a homeostatic root state toward a terminal foam cell state (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). Along this differentiation trajectory, the 10 core target genes (ALOX5, CDK4, HIF1A, ITGB1, MMP9, PIK3CA, PIK3R1, PPARG, PRKCB, STAT1) exhibited highly coordinated temporal expression dynamics, with transcriptional levels displaying sigmoidal nonlinear upregulation trajectories correlated with pseudotemporal progression. This expression architecture, exhibiting tight synchronization with cellular state transitions, indicated that these core genes not only correlate with foam cell formation but likely actively drive key pathological processes including macrophage lipid metabolic reprogramming, inflammatory transcriptional program activation, and phenotypic transformation, thereby executing pivotal regulatory functions in atherosclerotic plaque initiation and maintenance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFurthermore, global cell\u0026ndash;cell communication networks within the plaque microenvironment were systematically reconstructed using the CellChat toolkit to delineate functional positioning of core targets within multicellular interaction architectures. Global network topological analysis revealed intensive signaling crosstalk among macrophages, foam cells, endothelial cells, and T cells, collectively establishing a cellular communication hub orchestrating the inflammatory plaque microenvironment (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Ligand\u0026ndash;receptor interaction dissection further identified an ITGB1 (integrin β1)-centered cellular communication module (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). Foam cells, functioning as principal ligand-sending cells, exhibited elevated expression of pro-inflammatory/pro-fibrotic ligands including SPP1, FN1, and VCAM1; correspondingly, the receptor ITGB1 and cognate heterodimeric complexes (e.g., α4β1, α5β1 integrins) demonstrated predominant expression in endothelial cells, smooth muscle cells (SMCs), and T cells.\u003c/p\u003e \u003cp\u003eLigand\u0026ndash;receptor interaction strength quantification revealed substantial communication probabilities for SPP1\u0026ndash;ITGB1-associated interaction pairs (including SPP1\u0026ndash;α5β1 and SPP1\u0026ndash;α8β1) between foam cells and smooth muscle cells/endothelial cells, implicating their involvement in mediating critical pathological processes including cell adhesion, transendothelial migration, and smooth muscle cell phenotypic switching. Collectively, these cell\u0026ndash;cell communication network analyses suggested that \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents may disrupt pathological signaling between foam cells and stromal/immune cells through targeted modulation of ITGB1 receptor expression or activity, thereby attenuating inflammatory signal amplification cascades and intercepting SPP1\u0026ndash;ITGB1 axis-mediated aberrant cellular migration and vascular microenvironmental remodeling.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e3.6. Molecular Docking Analysis\u003c/h2\u003e \u003cp\u003eTo elucidate the molecular recognition principles governing \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituent\u0026ndash;target interactions at atomic resolution, systematic molecular docking analyses were performed. Binding energy heatmap analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA) revealed favorable binding affinities between \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents and the 10 core targets (e.g., HIF1A, ITGB1), with docking scores predominantly distributed below \u0026minus;\u0026thinsp;5.0 kcal/mol, indicating thermodynamically favorable ligand\u0026ndash;receptor complex formation with propensity for stable conformational association.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eNotably, luteolin, the principal bioactive constituent, exhibited pronounced multi-target binding potential, demonstrating particularly favorable binding free energies with the cellular communication hub ITGB1 (\u0026minus;\u0026thinsp;8.9 kcal/mol), inflammatory regulatory node ALOX5 (\u0026minus;\u0026thinsp;8.8 kcal/mol), and extracellular matrix remodeling mediator MMP9 (\u0026minus;\u0026thinsp;8.6 kcal/mol). These results are indicative of robust binding affinities, suggesting pivotal contributions to the multi-target synergistic modulation of AS by \u003cem\u003ePerilla frutescens\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eFurthermore, three-dimensional docking pose analysis unveiled mechanistic details underlying molecular recognition. Within the ALOX5\u0026ndash;luteolin complex (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB), the ligand occupied the hydrophobic binding pocket of ALOX5, forming hydrogen bonds with the critical residue Gln437, thereby conferring conformational stabilization. Within the ITGB1\u0026ndash;luteolin complex (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC), the ligand occupied a putative binding pocket within the receptor extracellular domain, with hydroxyl moieties forming an extensive hydrogen bonding network with the pivotal residue Glu320, complemented by synergistic van der Waals interactions and hydrophobic effects stabilizing the binding interface. Based on the spatial conformational architecture of this binding mode, luteolin may exert competitive inhibition of endogenous ligand (e.g., SPP1) engagement with ITGB1 through steric occlusion, thereby disrupting the pathological SPP1\u0026ndash;ITGB1 communication axis identified through cell\u0026ndash;cell communication network analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.7. MD Simulation Analysis\u003c/h2\u003e \u003cp\u003eTo assess the dynamic stability and conformational evolution characteristics of the luteolin\u0026ndash;ITGB1 complex under quasi-physiological solvation conditions, 100-ns all-atom MD simulations were executed. RMSD trajectory analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD) revealed rapid convergence to thermodynamic equilibrium at approximately 30 ns following initial solvation relaxation. Subsequently, RMSD trajectories exhibited pronounced stability, with fluctuation amplitudes confined within a narrow 0.10\u0026ndash;0.20 nm interval, devoid of conformational drift or ligand dissociation events, indicating preservation of protein backbone structural integrity upon ligand engagement.\u003c/p\u003e \u003cp\u003eRMSF analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE) quantified residue-specific local flexibility profiles. Results revealed pronounced rigidity within the ITGB1 core structural domain (RMSF\u0026thinsp;\u0026lt;\u0026thinsp;0.20 nm), with elevated fluctuations confined to surface-exposed flexible loop regions distal to the binding site. Critically, residues within the core binding pocket exhibited minimal fluctuations, confirming that luteolin effectively stabilizes the functional active conformation of ALOX5 via an induced-fit mechanism.\u003c/p\u003e \u003cp\u003eRg analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF) demonstrated stable Rg values within a narrow 2.17\u0026ndash;2.20 nm range throughout the simulation trajectory, devoid of unfolding propensity, confirming sustained maintenance of compact globular folded architecture. Additionally, solvent accessible surface area (SASA) remained stable within the 190\u0026ndash;205 nm\u0026sup2; range (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG), indicating invariant surface exposure characteristics within the dynamic aqueous environment, with negligible structural expansion or compaction.\u003c/p\u003e \u003cp\u003eCollectively, MD simulations furnished robust biophysical evidence corroborating the formation of persistent, specific, and thermodynamically stable luteolin\u0026ndash;ITGB1 molecular complexes, establishing a theoretical foundation for developing luteolin as a candidate anti-atherosclerotic lead compound.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Multi-component-Multi-target-Multi-pathway Synergistic Intervention Mode of \u003cem\u003ePerilla frutescens\u003c/em\u003e from the Perspective of Systems Pharmacology\u003c/h2\u003e \u003cp\u003eAS is essentially not a linear dysregulation of a single signaling pathway, but a systemic disease driven by the deep interweaving of lipid metabolic dysregulation and immune-inflammatory responses in spatiotemporal dimensions (Ajoolabady et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This pathological essence determines that AS has intrinsic resistance to single-target treatment strategies. Clinical observations reveal substantial residual cardiovascular risk even among patients achieving low-density lipoprotein cholesterol (LDL-C) targets under intensive statin regimens. This phenomenon reflects the persistent activation of compensatory inflammatory pathways alongside the intrinsic robustness of pathological networks. Specifically, the homeostatic capacity of diseased systems to sustain maladaptive functional states despite localized perturbations (Kanuri et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Tang et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe present investigation demonstrates that the multi-component, multi-target intervention paradigm of \u003cem\u003ePerilla frutescens\u003c/em\u003e transcends mere additive pharmacological effects, instead achieving the systematic attenuation of topological robustness within AS pathological networks through synergistic perturbation of 289 disease-critical nodes, thereby precipitating a global phase transition from pathological to physiological homeostatic states. From a network pharmacology perspective, this multi-nodal synergistic intervention strategy exhibits enhanced resilience to compensatory mechanisms relative to the localized perturbations induced by single-target therapeutics, embodying a contemporary molecular-network interpretation of the holistic philosophy underpinning traditional Chinese medicine.\u003c/p\u003e \u003cp\u003eThe pronounced enrichment of \"membrane rafts\" and \"membrane microdomains\" within GO functional analyses provides a distinctive biophysical framework for elucidating the pharmacological mechanisms of \u003cem\u003ePerilla frutescens\u003c/em\u003e. Membrane rafts, characterized as cholesterol- and sphingolipid-enriched liquid-ordered microdomains within plasma membranes, constitute essential spatial platforms facilitating the assembly of functional signalosomes for inflammatory pattern recognition receptors (e.g., TLR4) and lipid scavenger receptors (e.g., CD36) (Nieto-Garai et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Under atherogenic conditions, oxidized low-density lipoprotein (ox-LDL) binding to raft-localized CD36 receptors precipitates macrophage lipid overload, whereas pathogen-associated molecular patterns (PAMPs) amplify inflammatory cascades through the promotion of spatial clustering of TLR4/MyD88 signaling complexes within raft microdomains (Chen et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe phytosterols enriched in \u003cem\u003ePerilla frutescens\u003c/em\u003e (e.g., β-sitosterol, stigmasterol), possessing steroid scaffolds structurally homologous to cholesterol, may intercalate into raft lipid bilayers via competitive displacement mechanisms, thereby modulating microdomain physicochemical properties encompassing membrane thickness, lipid fluidity, and membrane curvature (Zhao et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Tang et al. (Tang et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) systematically characterized the biophysical effects of phytosterols within plant plasma membrane-mimetic lipid systems through the integration of fluorescence lifetime imaging microscopy and all-atom MD simulations. Their findings demonstrate that phytosterol incorporation induces significant reductions in membrane lipid area, elevations in bilayer thickness, enhancements in fatty acyl chain ordering, and the formation of phytosterol-enriched clusters corresponding to lipid microdomain/phase separation phenomena. This biophysical remodeling of membrane rafts disrupts the spatial clustering and functional interaction interfaces of signaling receptors, thereby imposing signal transduction blockade at the mechanistic origin, ultimately conferring membrane-targeted anti-inflammatory effects of broader spectrum than those achievable through single-receptor antagonism (Yuan et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eKEGG pathway enrichment analysis results show that the \"Lipid and atherosclerosis\" pathway presents the most significant enrichment, indicating that candidate target genes are highly enriched in the functional core modules of the AS pathological network. This pathway covers the complete lipid metabolism axis from ox-LDL uptake, cholesterol esterification to reverse transport. The synergistic targeting of multiple critical nodes within this pathway by \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents may disrupt the lipid accumulation\u0026ndash;inflammation activation positive feedback loop inherent to foam cell formation through a bidirectional regulatory paradigm of upstream flux restriction coupled with downstream efflux enhancement (Kong et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). \u003cem\u003ePerilla frutescens\u003c/em\u003e harbors an extensive repertoire of bioactive constituents (including α-linolenic acid, flavonoids, and phenolic acids) that exert synergistic actions upon multiple critical processes encompassing lipid uptake, cholesterol esterification, reverse cholesterol transport, oxidative stress mitigation, and inflammatory response attenuation (Hou et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Evidence from animal models and cellular experiments demonstrates that \u003cem\u003ePerilla frutescens\u003c/em\u003e extracts downregulate the expression of ox-LDL scavenger receptors (CD36, LOX-1), suppress foam cell formation, facilitate cholesterol efflux through the upregulation of ABCA1 and SR-B1 expression, and substantially attenuate the secretion of pro-inflammatory cytokines (e.g., IL-1β, MCP-1) (Pothinam et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eImpaired efferocytosis precipitates secondary necrosis of apoptotic cells and progressive expansion of necrotic cores, processes intimately associated with vulnerable plaque formation and the incidence of acute coronary syndrome (ACS) (Morrissey et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). \u003cem\u003ePerilla frutescens\u003c/em\u003e may restore intraplaque efferocytic efficiency through the upregulation of phagocytic receptors (e.g., MerTK), the activation of peroxisome proliferator-activated receptor gamma (PPARγ)-dependent bridging molecule secretion, and the suppression of the inhibitory CD47\u0026ndash;SIRPα axis. The restoration of efferocytic capacity not only diminishes necrotic core dimensions but also initiates active inflammation resolution programs through the liberation of specialized pro-resolving mediators (SPMs), including lipoxins and resolvins (Filep, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe pronounced enrichment of the neutrophil extracellular trap (NET) formation (NETosis) pathway suggests the potential modulatory capacity of Perilla frutescens in immunothrombosis. NETs exert pathogenic actions across multiple stages of AS progression: during early phases, histones impose cytotoxic injury upon vascular endothelium; throughout disease advancement, NETs function as damage-associated molecular patterns (DAMPs) activating NLRP3 inflammasomes to amplify inflammatory responses; and during acute events, NETs serve as thrombogenic scaffolds capturing platelets and coagulation factors (D\u0026ouml;ring et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Luteolin, a cardinal bioactive constituent of \u003cem\u003ePerilla frutescens\u003c/em\u003e, demonstrates established capacity for inhibiting NADPH oxidase (NOX2) activity. Given that NOX2-mediated reactive oxygen species (ROS) generation constitutes a critical initiating determinant of NETosis, these findings suggest that \u003cem\u003ePerilla frutescens\u003c/em\u003e may suppress NET formation at mechanistic origins (Xia et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFurthermore, the substantial enrichment of the advanced glycation end product (AGE)\u0026ndash;receptor for AGE (RAGE) signaling pathway suggests distinctive therapeutic potential for \u003cem\u003ePerilla frutescens\u003c/em\u003e in patients with concomitant diabetes mellitus and AS. The hyperactivation of the AGE\u0026ndash;RAGE axis perpetuates oxidative stress amplification and inflammatory signal cascades via nuclear factor-κB (NF-κB) pathways, constituting a pivotal pathogenic mechanism underlying accelerated AS progression in diabetic populations (B. Wang et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Molecular Analysis of Foam Cell Differentiation Trajectory and Core Target Identification at Single-Cell Resolution\u003c/h2\u003e \u003cp\u003eConventional transcriptomic investigations relying upon bulk RNA sequencing (bulk RNA-seq) frequently obscure the profound cellular heterogeneity within atherosclerotic plaques through population-averaging effects, thereby precluding the precise identification of critical molecular events (Tzec-Interi\u0026aacute;n et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The present study reconstructed the comprehensive cellular atlas of human carotid atherosclerotic plaques through the integration of high-resolution single-cell RNA sequencing (scRNA-seq) data, thereby achieving precise discrimination between foam cells and classical macrophage subpopulations. Building upon this foundation, the implementation of a dual-algorithm feature selection framework integrating least absolute shrinkage and selection operator (LASSO) regression and random forest methodologies enabled the precise identification of a 10-gene signature module centered upon HIF1A, PPARG, and ALOX5\u003c/p\u003e \u003cp\u003eFor instance, Xu et al. (Xu et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) integrated multiple Gene Expression Omnibus (GEO) microarray datasets, identifying 611 AS-associated differentially expressed genes through differential expression analyses, subsequently employing multiple machine learning algorithms (LASSO, random forest) for key gene selection, and validating findings across external human and murine specimens. Their investigation ultimately proposed a diagnostic gene pair comprising DHRS9 and PTPRJ, which exhibited robust discriminatory capacity between AS and control samples and demonstrated strong associations with diverse immune cell infiltration patterns. Ban et al. (Ban et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) performed weighted gene co-expression network analysis (WGCNA) and differential expression profiling on AS and ischemic stroke (IS) datasets, constructing shared differentially expressed gene networks and identifying ATF3, CCL3, CCL4, JUNB, KRAS, and ZC3H12A as shared hub genes potentially participating in the pathological processes underlying both AS and IS, with subsequent validation of expression trends via quantitative real-time PCR (qPCR) analysis of clinical specimens. This gene module represents not a stochastic assemblage but rather a constellation profoundly reflecting the immunometabolic reprogramming signature inherent to foam cell formation. Pseudotemporal trajectory reconstruction analyses confirmed that the expression abundance of this gene module exhibits highly coordinated sigmoidal upregulation dynamics along pathological differentiation trajectories, strongly implicating these core genes not as passive markers of foam cell states but rather as cardinal regulatory determinants actively orchestrating phenotypic transitions.\u003c/p\u003e \u003cp\u003eThe pronounced upregulation of HIF1A unveils the distinctive hypoxic microenvironmental signature characterizing plaque core regions. With progressive plaque expansion, oxygen diffusion limitations precipitate substantial reductions in local oxygen tension, thereby triggering HIF1A-dependent metabolic reprogramming characterized by the metabolic transition from oxidative phosphorylation to glycolysis (the Warburg effect) (Qiu et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). While this metabolic shift sustains cellular bioenergetic homeostasis under hypoxic conditions in the short term, it precipitates persistent lactate accumulation. Lactate, through the activation of G protein-coupled receptor 81 (GPR81), suppresses cyclic adenosine monophosphate (cAMP) signaling cascades, thereby further attenuating anti-inflammatory programs and exacerbating macrophage polarization toward M1-type pro-inflammatory phenotypes, ultimately establishing a self-perpetuating hypoxia\u0026ndash;metabolism\u0026ndash;inflammation vicious cycle.\u003c/p\u003e \u003cp\u003eAs a cardinal member of the lipid-sensing nuclear receptor superfamily, PPARγ facilitates reverse cholesterol transport (RCT) under physiological conditions through the transcriptional activation of lipid efflux transporters (e.g., ABCA1/ABCG1), thereby conferring atheroprotective effects (Zhang et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, under conditions of sustained intraplaque lipid overload, PPARγ function may undergo pathological \"hijacking\": ligand activation upregulates fatty acid translocase expression (e.g., CD36), paradoxically exacerbating dysregulated ox-LDL uptake; concurrently, elevated oxidized lipid concentrations may impair PPARγ\u0026ndash;coactivator interactions through covalent modifications, thereby attenuating its transcriptional activation capacity. This double-edged sword phenomenon of PPARγ functionality partially explains the failure of PPARγ agonist monotherapy to substantially improve clinical outcomes in AS patients (Riccioni et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). LTB4 exerts potent chemotactic actions via leukotriene B4 receptor 1 (BLT1), orchestrating the recruitment of neutrophils and monocytes into atherosclerotic plaques, whereas CysLTs, acting through CysLT1 receptors, promote vascular smooth muscle cell contraction and endothelial permeability elevations. The sustained generation of ALOX5 metabolites and inflammatory cell infiltration establish a positive feedback loop, constituting a pivotal molecular mechanism underlying the recalcitrance of chronic plaque inflammation to spontaneous resolution (Kotlyarov \u0026amp; Kotlyarova, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFurthermore, these 10 core gene targets demonstrated exceptional diagnostic performance (AUC\u0026thinsp;=\u0026thinsp;0.996) within independent external validation cohorts. These findings not only substantiate the robustness and reproducibility of the feature selection framework but also underscore the substantial clinical translational potential of these core molecules as liquid biopsy biomarkers for AS.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Analysis of Cell-Cell Communication Network and Elucidation of ITGB1 Hub Function\u003c/h2\u003e \u003cp\u003eAtherosclerotic plaques constitute not static cellular aggregations but rather highly dynamic cellular ecosystems sustained by intricate intercellular communication networks (Raju et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Diverse cellular populations within plaques establish dense information exchange networks through ligand\u0026ndash;receptor interactions, extracellular vesicle (EV) trafficking, and metabolite signaling. These multicellular coordination patterns collectively orchestrate plaque inflammatory microenvironmental characteristics and structural integrity (Raju et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The present investigation employed CellChat computational frameworks to systematically dissect the global intercellular communication landscape within plaques, thereby unveiling the molecular mechanisms through which \u003cem\u003ePerilla frutescens\u003c/em\u003e exerts therapeutic effects via targeted disruption of the SPP1\u0026ndash;ITGB1 signaling axis. Secreted phosphoprotein 1 (SPP1), predominantly expressed by discrete macrophage subpopulations, constitutes a pivotal signaling hub orchestrating AS progression and plaque destabilization (Li et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). SPP1 promotes fibrotic responses within fibro-progenitor cells through integrin receptor engagement (encompassing ITGB1, ITGAV/ITGB5). Early functional investigations demonstrate that the SPP1\u0026ndash;integrin signaling axis drives vascular smooth muscle cell (VSMC) migration and phenotypic switching, processes intimately associated with intimal thickening and foam cell formation (Huang et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLigand\u0026ndash;receptor interaction analyses reveal that SPP1 (osteopontin, OPN), abundantly expressed and secreted by foam cells, represents the predominant ligand activating integrin β1 (ITGB1) on vascular wall cells (endothelial cells, smooth muscle cells) and infiltrating immune cells (Yim et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). ITGB1, as the cardinal β subunit within the integrin superfamily, heterodimerizes with diverse α subunits (e.g., α4β1, α5β1, α8β1), exerting nodal regulatory functions in cellular adhesion, migration, and mechanotransduction through bidirectional \"outside-in\" and \"inside-out\" signaling mechanisms (Su et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSPP1 engagement with endothelial α4β1/α5β1 integrins activates focal adhesion kinase (FAK)\u0026ndash;Src signaling cascades, precipitating tyrosine phosphorylation and subsequent endocytic degradation of vascular endothelial cadherin (VE-cadherin), thereby compromising adherens junction integrity (Fei et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The attenuation of endothelial barrier function precipitates elevations in vascular permeability, thereby facilitating transendothelial migration (TEM) of circulating monocytes and sustaining the replenishment of intraplaque inflammatory cell reservoirs (Dalal et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eUnder physiological conditions, vascular smooth muscle cells (VSMCs) manifest a contractile phenotype essential for the maintenance of vascular tone and structural integrity (Cao et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). ITGB1, functioning as a cardinal mechanosensor, transduces ECM mechanical cues into intracellular biochemical signals. ITGB1 signaling cascades drive VSMC phenotypic switching toward synthetic states through the activation of downstream phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt) and mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathways, manifested by the downregulation of contractile proteins (α-smooth muscle actin [α-SMA], smooth muscle myosin heavy chain [SM-MHC]) alongside the upregulation of matrix synthesis proteins (collagen, proteoglycans) and matrix metalloproteinases (MMPs) (Wu et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The excessive proliferation and enhanced apoptosis of synthetic VSMCs ultimately compromise plaque structural stability, thereby elevating plaque rupture risk.\u003c/p\u003e \u003cp\u003ePredicated upon these mechanisms, \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents may interrupt the transmission of pathological signals from foam cells to the microenvironment through targeted disruption of SPP1\u0026ndash;ITGB1 interactions at the intercellular communication level, thereby achieving: restoration of endothelial barrier functionality to attenuate inflammatory cell infiltration; preservation of VSMC contractile phenotypes to stabilize plaque architecture; and suppression of excessive MMP activation to prevent fibrous cap rupture. This intervention paradigm predicated upon the disruption of pathological intercellular communication furnishes a conceptually innovative therapeutic framework for AS that transcends conventional intracellular signaling pathway targeting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Multi-target Binding Characteristics of Luteolin and Kinetic Stability of ITGB1 Complex\u003c/h2\u003e \u003cp\u003eMolecular docking and MD simulations furnished atomic-resolution structural and kinetic evidence characterizing interactions between \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituents and core targets. Binding energy heatmap analyses reveal favorable binding affinities of 16 \u003cem\u003ePerilla frutescens\u003c/em\u003e constituents toward the 10 core targets, indicating thermodynamic spontaneity of ligand\u0026ndash;receptor complex formation.\u003c/p\u003e \u003cp\u003eWithin the \u003cem\u003ePerilla frutescens\u003c/em\u003e bioactive constituent repertoire, luteolin exhibits pronounced multi-target binding potential. As a prototypical flavonoid, the molecular architecture of luteolin, encompassing A/B dual aromatic ring systems, C-ring C2\u0026ndash;C3 unsaturation, and multiple phenolic hydroxyl moieties, confers the structural foundation for establishing stable non-covalent interactions with diverse protein targets. Docking analyses demonstrate that luteolin exhibits favorable binding free energies toward the intercellular communication hub ITGB1 (\u0026minus;\u0026thinsp;8.9 kcal/mol), the inflammatory rate-limiting enzyme ALOX5 (\u0026minus;\u0026thinsp;8.8 kcal/mol), and the extracellular matrix degradation effector MMP9 (\u0026minus;\u0026thinsp;8.6 kcal/mol).\u003c/p\u003e \u003cp\u003eThree-dimensional conformational analyses unveil the molecular recognition mechanisms underlying luteolin\u0026ndash;target interactions. Within the ALOX5\u0026ndash;luteolin complex, the ligand intercalates into the hydrophobic binding cavity of the catalytic domain, establishing stable interactions with regions proximal to the non-heme iron active site; the A-ring phenolic hydroxyl forms a hydrogen bond with the critical residue Gln437, thereby anchoring the binding conformation. The spatial occupancy of the active site by luteolin may obstruct leukotriene biosynthetic pathways through competitive inhibition of substrate access (Ren et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Within the ITGB1\u0026ndash;luteolin complex, the ligand occupies the ligand-binding pocket of the extracellular I-like domain of ITGB1; the C4\u0026prime;-hydroxyl forms a hydrogen bond with the carboxyl side chain of the critical residue Glu320, while the B-ring engages in π-alkyl interactions with hydrophobic residues lining the pocket interior (De Aguiar et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This binding site coincides precisely with the recognition interface for endogenous ligands such as SPP1, suggesting that luteolin may competitively abrogate pathological SPP1\u0026ndash;ITGB1 interactions through steric occlusion (Zhou et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWhile molecular docking furnishes static structural snapshots of ligand\u0026ndash;receptor interactions, proteins undergo continuous conformational fluctuations under physiological conditions. The capacity of ligands to sustain stable binding within dynamic aqueous environments directly governs the duration and magnitude of pharmacological activity (Fu et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This property assumes particular significance for ITGB1, given that as a mechanotransduction receptor, its functionality critically depends upon dynamic transitions between bent low-affinity and extended high-affinity conformational states. The present investigation systematically assessed the spatiotemporal dynamic evolution of the luteolin\u0026ndash;ITGB1 complex through 100-nanosecond all-atom MD simulations.\u003c/p\u003e \u003cp\u003eRMSD trajectory analyses reveal rapid convergence of the complex to thermodynamic equilibrium at approximately 30 nanoseconds; subsequently, fluctuation amplitudes remained within a narrow 0.10\u0026ndash;0.20 nm range, with no observable ligand dissociation or substantial conformational drift throughout the simulation. This rapid equilibration characteristic suggests that luteolin efficiently accommodates the receptor binding pocket microenvironment, indicating potential for prompt in vivo onset of action.\u003c/p\u003e \u003cp\u003eThe local conformational stabilization effects unveiled through RMSF analyses constitute a cardinal finding of the present investigation. Core residues within the ligand-binding pocket, most notably Glu320 and adjacent hydrophobic residue clusters, exhibit substantially diminished fluctuation amplitudes (RMSF\u0026thinsp;\u0026lt;\u0026thinsp;0.20 nm), whereas regions of elevated fluctuation remain confined to surface-exposed flexible loops distal to the binding interface. The biological significance of this local rigidification resides in luteolin capacity to effectively lock ITGB1 into a defined functional conformation through binding pocket occupancy and the establishment of multivalent non-covalent interaction networks, thereby attenuating receptor responsiveness to endogenous ligands (e.g., SPP1) or mechanical stimuli. Given that ITGB1 signal transduction relies upon ligand-induced conformational changes (outside-in signaling), this conformational locking effect may fundamentally abrogate downstream FAK/Src signaling cascades mediated by the SPP1\u0026ndash;ITGB1 axis (Su et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFrom a medicinal chemistry perspective, ligand-induced binding pocket rigidification diminishes conformational entropy at the binding interface, thereby elevating the energy barrier for ligand dissociation through enthalpy\u0026ndash;entropy compensation mechanisms (Stank et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). This kinetics-driven pharmacological advantage assumes particular importance for chronic pathologies such as AS requiring sustained therapeutic intervention (Wang et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This conformational locking mechanism significantly prolongs the receptor residence time of luteolin, thereby allowing the ligand to sustain target occupancy and functional blockade even under conditions of declining plasma concentrations.\u003c/p\u003e \u003cp\u003eThe stability of the Rg ( 2.17\u0026ndash;2.20 nm) and solvent-accessible SASA (190\u0026ndash;205 nm\u0026sup2;) throughout the simulation trajectory indicates that ITGB1 maintains a compact, globular folded conformation without structural unfolding or long-range allosteric propagation. This binding mode substantiates the mechanistic hypothesis that luteolin functions as an orthosteric competitive inhibitor of the SPP1\u0026ndash;ITGB1 axis, disrupting pathological intercellular communication through direct occupancy of the ligand recognition interface rather than indirect allosteric modulation. Integration with the aforementioned CellChat intercellular communication network analyses suggests that luteolin targeted intervention upon ITGB1 may sever pro-inflammatory and pro-remodeling signals emanating from foam cells into the vascular microenvironment at the intercellular communication level, thereby conferring multifaceted protective effects encompassing endothelial barrier restoration, VSMC contractile phenotype preservation, and extracellular matrix degradation suppression (Wu et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn summary, the present investigation establishes a multiscale computational analytical framework integrating single-cell transcriptomics, machine learning algorithms, and MD simulations to systematically elucidate the multi-target synergistic mechanisms underlying the therapeutic intervention of \u003cem\u003ePerilla frutescens\u003c/em\u003e frutescens seed in AS. The investigation identifies a 10-gene core signature module encompassing HIF1A, PPARG, and ITGB1, which not only demonstrates exceptional diagnostic performance within independent clinical validation cohorts (AUC\u0026thinsp;=\u0026thinsp;0.996) but also profoundly captures the molecular signatures of immunometabolic reprogramming accompanying macrophage-to-foam cell differentiation. More critically, the study furnishes atomic-resolution biophysical evidence substantiating that luteolin, a cardinal bioactive constituent of \u003cem\u003ePerilla frutescens\u003c/em\u003e, engages integrin β1 (ITGB1) with high specificity through conformational locking mechanisms, thereby spatially disrupting the foam cell-driven SPP1\u0026ndash;ITGB1 pro-inflammatory communication axis. Collectively, these findings suggest that the targeted disruption of pathological cell adhesion receptors and their mediated signal crosstalk constitutes a highly promising innovative therapeutic strategy for addressing RIR in AS. While the present investigation primarily employs computational biology and molecular simulation strategies, future studies warrant systematic validation of the pharmacokinetic/pharmacodynamic (PK/PD) profiles, in vivo targeting efficiency, and long-term therapeutic efficacy of luteolin\u0026ndash;ITGB1 interactions within animal models and clinical trials, alongside comprehensive exploration of its clinical translational potential as a precision therapeutic candidate for AS.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eDeclaration of competing interest\u003c/h2\u003e \u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis work was supported by the financial support from Key Research and Development Program of Zhejiang Province, China (2020C02034).\u003c/p\u003e\u003ch2\u003eAuthors' contributions\u003c/h2\u003e \u003cp\u003eChenchen Yang: Conceptualization (assisted), data curation, formal analysis, investigation, software, visualization, writing \u0026ndash; original draft.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eThe authors wish to acknowledge our hardworking computer, which endured long-term high-intensity workloads including bioinformatics data processing, code running, chart visualization and manuscript drafting without complaint, serving as an indispensable silent partner throughout this research.\u003c/p\u003e\u003ch2\u003eAvailability of data and materials\u003c/h2\u003e \u003cp\u003eTarget intersection analysis and Venn diagram visualization were performed in R (v4.3.2) with the ggVennDiagram package (v1.2.2). Subsequent PPI network construction, functional enrichment and single-cell transcriptomic analysis were carried out via standard bioinformatics pipelines.\u003c/p\u003e \u003cp\u003eAll public database resources are freely accessible via the listed URLs. Generated processed datasets are available upon reasonable request to the corresponding author for research replication.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbraham MJ, Murtola T, Schulz R, P\u0026aacute;ll S, Smith JC, Hess B, Lindahl E (2015) GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. \u003cem\u003eSoftwareX\u003c/em\u003e, \u003cem\u003e1\u0026ndash;2\u003c/em\u003e, 19\u0026ndash;25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1016/j.softx.2015.06.001\u003c/span\u003e\u003cspan address=\"10.1016/j.softx.2015.06.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAjoolabady A, Pratico D, Lin L, Mantzoros CS, Bahijri S, Tuomilehto J, Ren J (2024) Inflammation in atherosclerosis: pathophysiology and mechanisms. Cell Death Dis 15(11):817. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41419-024-07166-8\u003c/span\u003e\u003cspan address=\"10.1038/s41419-024-07166-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlsaigh T, Evans D, Frankel D, Torkamani A (2022) Decoding the transcriptome of calcified atherosclerotic plaque at single-cell resolution. Commun Biol 5(1):1084. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s42003-022-04056-7\u003c/span\u003e\u003cspan address=\"10.1038/s42003-022-04056-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBan R, Huo C, Wang J, Zhang G, Zhao X (2024) Exploration of the Shared Gene Signatures and Molecular Mechanisms Between Ischemic Stroke and Atherosclerosis. Int J Gen Med 17:2223\u0026ndash;2239. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2147/ijgm.s454336\u003c/span\u003e\u003cspan address=\"10.2147/ijgm.s454336\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao G, Xuan X, Hu J, Zhang R, Jin H, Dong H (2022) How vascular smooth muscle cell phenotype switching contributes to vascular disease. Cell Communication Signaling: CCS 20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12964-022-00993-2\u003c/span\u003e\u003cspan address=\"10.1186/s12964-022-00993-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao J, Spielmann M, Qiu X, Huang X, Ibrahim DM, Hill AJ, Zhang F, Mundlos S, Christiansen L, Steemers FJ, Trapnell C, Shendure J (2019) The single-cell transcriptional landscape of mammalian organogenesis. Nature 566(7745):496\u0026ndash;502. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-019-0969-x\u003c/span\u003e\u003cspan address=\"10.1038/s41586-019-0969-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Y, Zhang J, Cui W, Silverstein RL (2022) CD36, a signaling receptor and fatty acid transporter that regulates immune cell metabolism and fate. J Exp Med 219(6). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1084/jem.20211314\u003c/span\u003e\u003cspan address=\"10.1084/jem.20211314\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDalal P, Muller W, Sullivan D (2019) Endothelial Cell Calcium Signaling During Barrier Function and Inflammation. Am J Pathol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ajpath.2019.11.004\u003c/span\u003e\u003cspan address=\"10.1016/j.ajpath.2019.11.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Aguiar A, De Carvalho LBR, Gomes C, Castro MM, Martins F, Borges LL (2025) Computational Insights into the Antioxidant Activity of Luteolin: Density Functional Theory Analysis and Docking in Cytochrome P450 17A1. \u003cem\u003ePharmaceuticals\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ph18030410\u003c/span\u003e\u003cspan address=\"10.3390/ph18030410\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Muro F, Vogel B, Sartori S, Bay B, Oliva A, Feng Y, Krishnan P, Sweeny J, Gitto M, Smith K, Moreno P, Nicolas J, Krishnamoorthy P, Leone PP, Bhatt D, Dangas G, Kini A, Sharma S, Mehran R (2025) Prognostic impact of residual inflammatory and triglyceride risk in statin-treated patients with well-controlled LDL cholesterol and atherosclerotic cardiovascular disease. Eur J Prev Cardiol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/eurjpc/zwaf112\u003c/span\u003e\u003cspan address=\"10.1093/eurjpc/zwaf112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD\u0026ouml;ring Y, Soehnlein O, Weber C (2017) Neutrophil Extracellular Traps in Atherosclerosis and Atherothrombosis. Circul Res 120(4):736\u0026ndash;743. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1161/CIRCRESAHA.116.309692\u003c/span\u003e\u003cspan address=\"doi:10.1161/CIRCRESAHA.116.309692\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. J Chem Inf Model 61(8):3891\u0026ndash;3898. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jcim.1c00203\u003c/span\u003e\u003cspan address=\"10.1021/acs.jcim.1c00203\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFei Z, Liu Z-T, Zhou G-W, Liang F, Wang Y-H, Chen L, Zhang W-F, Shen L, Lu Y-Q, Huo H, Shi X, Fang L, He B (2025) Integrin β3-mediated platelet extracellular vesicle adhesion facilitates vascular smooth muscle cell dysfunction in postinjury intimal hyperplasia. Int J Biol Sci 21:2380\u0026ndash;2395. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7150/ijbs.101391\u003c/span\u003e\u003cspan address=\"10.7150/ijbs.101391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFilep JG (2022) Targeting Neutrophils for Promoting the Resolution of Inflammation. \u003cem\u003eFrontiers in Immunology\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fimmu.2022.866747\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2022.866747\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu Y, Zhao J, Chen Z (2018) Insights into the Molecular Mechanisms of Protein-Ligand Interactions by Molecular Docking and Molecular Dynamics Simulation: A Case of Oligopeptide Binding Protein. \u003cem\u003eComputational and Mathematical Methods in Medicine\u003c/em\u003e, \u003cem\u003e2018\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2018/3502514\u003c/span\u003e\u003cspan address=\"10.1155/2018/3502514\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHao Y, Hao S, Andersen-Nissen E, Mauck WM 3rd, Zheng S, Butler A, Lee MJ, Wilk AJ, Darby C, Zager M, Hoffman P, Stoeckius M, Papalexi E, Mimitou EP, Jain J, Srivastava A, Stuart T, Fleming LM, Yeung B, Satija R (2021) Integrated analysis of multimodal single-cell data. Cell 184(13):3573\u0026ndash;3587e3529. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2021.04.048\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2021.04.048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou T, Netala VR, Zhang H, Xing Y, Li H, Zhang Z (2022) \u003cem\u003ePerilla frutescens\u003c/em\u003e frutescens: A Rich Source of Pharmacological Active Compounds. Molecules 27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/molecules27113578\u003c/span\u003e\u003cspan address=\"10.3390/molecules27113578\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang K, Chen S, Yu L, Wu Z, Chen QJ, Wang XQ, Li F-F, Liu J, Wang Y-X, Mao L-S, Shen W, Zhang R-Y, Shen Y, Lu L, Dai Y, Ding F (2024) Serum secreted phosphoprotein 1 level is associated with plaque vulnerability in patients with coronary artery disease. Front Immunol 15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fimmu.2024.1285813\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2024.1285813\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKanuri B, Maremanda KP, Chattopadhyay D, Essop MF, Lee MKS, Murphy AJ, Nagareddy PR (2025) Redefining Macrophage Heterogeneity in Atherosclerosis: A Focus on Possible Therapeutic Implications. Compr Physiol 15(2):e70008. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/cph4.70008\u003c/span\u003e\u003cspan address=\"10.1002/cph4.70008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKong P, Cui Z-Y, Huang X-F, Zhang D-D, Guo R-J, Han M (2022) Inflammation and atherosclerosis: signaling pathways and therapeutic intervention. Signal Transduct Target Therapy 7(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41392-022-00955-7\u003c/span\u003e\u003cspan address=\"10.1038/s41392-022-00955-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKotlyarov S, Kotlyarova A (2022) Molecular Pharmacology of Inflammation Resolution in Atherosclerosis. Int J Mol Sci 23(9):4808\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Y, Wang S, Zhang R, Gong Y, Che Y, Li K, Pan Z (2025) Single-cell and spatial analysis reveals the interaction between ITLN1\u0026thinsp;+\u0026thinsp;foam cells and SPP1\u0026thinsp;+\u0026thinsp;macrophages in atherosclerosis. Front Cardiovasc Med 12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fcvm.2025.1510082\u003c/span\u003e\u003cspan address=\"10.3389/fcvm.2025.1510082\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu T, Chen F (2012) Multiwfn: a multifunctional wavefunction analyzer. J Comput Chem 33(5):580\u0026ndash;592. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jcc.22885\u003c/span\u003e\u003cspan address=\"10.1002/jcc.22885\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorrissey MA, Kern N, Vale RD (2020) CD47 Ligation Repositions the Inhibitory Receptor SIRPA to Suppress Integrin Activation and Phagocytosis. Immunity 53(2):290\u0026ndash;302e296. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.immuni.2020.07.008\u003c/span\u003e\u003cspan address=\"10.1016/j.immuni.2020.07.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNieto-Garai JA, Lorizate M, Contreras FX (2022) Shedding light on membrane rafts structure and dynamics in living cells. Biochim et Biophys Acta (BBA) - Biomembr 1864(1):183813. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bbamem.2021.183813\u003c/span\u003e\u003cspan address=\"10.1016/j.bbamem.2021.183813\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePothinam S, Putpim C, Siriwoharn T, Jirarattanarangsri W (2025) Effects of \u003cem\u003ePerilla frutescens\u003c/em\u003e Seed Oil on Blood Lipids, Oxidative Stress, and Inflammation in Hyperlipidemic Rats. Foods 14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/foods14081380\u003c/span\u003e\u003cspan address=\"10.3390/foods14081380\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu B, Yuan P, Du X, Jin H, Du J, Huang Y (2023) Hypoxia inducible factor-1α is an important regulator of macrophage biology. Heliyon 9(6):e17167. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1016/j.heliyon.2023.e17167\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2023.e17167\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaju S, Turner M, Cao C, Abdul-Samad M, Punwasi N, Blaser M, Cahalane R, Botts S, Prajapati K, Patel S, Wu R, Gustafson D, Galant N, Fiddes L, Chemaly M, Hedin U, Matic L, Seidman M, Subasri V, Howe K (2024) Multiomics unveils extracellular vesicle-driven mechanisms of endothelial communication in human carotid atherosclerosis. bioRxiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1101/2024.06.21.599781\u003c/span\u003e\u003cspan address=\"10.1101/2024.06.21.599781\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaju S, Turner M, Cao C, Abdul-Samad M, Punwasi N, Blaser M, Cahalane R, Botts S, Prajapati K, Patel S, Wu R, Gustafson D, Galant N, Fiddes L, Chemaly M, Hedin U, Matic L, Seidman M, Subasri V, Howe K (2025) Multiomic Landscape of Extracellular Vesicles in Human Carotid Atherosclerotic Plaque Reveals Endothelial Communication Networks. Arterioscler Thromb Vasc Biol 45:1277\u0026ndash;1305. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1161/atvbaha.124.322324\u003c/span\u003e\u003cspan address=\"10.1161/atvbaha.124.322324\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen P, Cao J-L, Lin P-L, Cao B, Chen J, Gao K, Zhang J (2021) [Molecular mechanism of luteolin regulating lipoxygenase pathway against oxygen-glucose deprivation/reperfusion injury in H9c2 cardiomyocytes based on molecular docking]. \u003cem\u003eZhongguo Zhong yao za zhi\u0026thinsp;=\u0026thinsp;Zhongguo zhongyao zazhi\u0026thinsp;=\u0026thinsp;China journal of Chinese materia medica\u003c/em\u003e, \u003cem\u003e46 21\u003c/em\u003e, 5665\u0026ndash;5673. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.19540/j.cnki.cjcmm.20210805.701\u003c/span\u003e\u003cspan address=\"10.19540/j.cnki.cjcmm.20210805.701\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiccioni G, Zanasi A, Vitulano N, Mancini B, D'Orazio N (2009) Leukotrienes in atherosclerosis: new target insights and future therapy perspectives. \u003cem\u003eMediators Inflamm\u003c/em\u003e, \u003cem\u003e2009\u003c/em\u003e, 737282. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2009/737282\u003c/span\u003e\u003cspan address=\"10.1155/2009/737282\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRu J, Li P, Wang J, Zhou W, Li B, Huang C, Li P, Guo Z, Tao W, Yang Y, Xu X, Li Y, Wang Y, Yang L (2014) TCMSP: a database of systems pharmacology for drug discovery from herbal medicines. J Cheminform 6:13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/1758-2946-6-13\u003c/span\u003e\u003cspan address=\"10.1186/1758-2946-6-13\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13(11):2498\u0026ndash;2504. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1101/gr.1239303\u003c/span\u003e\u003cspan address=\"10.1101/gr.1239303\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStank A, Kokh DB, Fuller JC, Wade RC (2016) Protein Binding Pocket Dynamics. Acc Chem Res 49(5):809\u0026ndash;815. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.accounts.5b00516\u003c/span\u003e\u003cspan address=\"10.1021/acs.accounts.5b00516\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSu C, Mo J, Dong S, Liao Z, Zhang B-X, Zhu P (2024) Integrinβ-1 in disorders and cancers: molecular mechanisms and therapeutic targets. Cell Communication Signaling: CCS 22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12964-023-01338-3\u003c/span\u003e\u003cspan address=\"10.1186/s12964-023-01338-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSzklarczyk D, Gable AL, Nastou KC, Lyon D, Kirsch R, Pyysalo S, Doncheva NT, Legeay M, Fang T, Bork P, Jensen LJ, von Mering C (2021) The STRING database in 2021: customizable protein-protein networks, and functional characterization of user-uploaded gene/measurement sets. Nucleic Acids Res 49(D1):D605\u0026ndash;d612. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/nar/gkaa1074\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkaa1074\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang L, Li Y, Zhong C, Deng X, Wang X (2021) Plant Sterol Clustering Correlates with Membrane Microdomains as Revealed by Optical and Computational Microscopy. \u003cem\u003eMembranes\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/membranes11100747\u003c/span\u003e\u003cspan address=\"10.3390/membranes11100747\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang S, Yang J, Xiao B, Wang Y, Lei Y, Lai D, Qiu Q (2024) Aberrant lipid metabolism and complement activation in age-related macular degeneration. Investig Ophthalmol Vis Sci 65(12):20\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1167/iovs.65.12.20\u003c/span\u003e\u003cspan address=\"10.1167/iovs.65.12.20\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTzec-Interi\u0026aacute;n JA, Gonz\u0026aacute;lez‐Padilla D, G\u0026oacute;ngora‐Castillo EB (2025) Bioinformatics perspectives on transcriptomics: A comprehensive review of bulk and single‐cell RNA sequencing analyses. Quant Biology 13(2):e78. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/qub2.78\u003c/span\u003e\u003cspan address=\"10.1002/qub2.78\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang B, Jiang T, Qi Y, Luo S, Xia Y, Lang B, Zhang B, Zheng S (2025) AGE-RAGE Axis and Cardiovascular Diseases: Pathophysiologic Mechanisms and Prospects for Clinical Applications. Cardiovasc Drugs Ther 39(6):1489\u0026ndash;1506. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10557-024-07639-0\u003c/span\u003e\u003cspan address=\"10.1007/s10557-024-07639-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang H, Zheng H, Yuheng (2020) Drug treatment of ankylosing spondylitis and related complications: an overlook review. Annals Palliat Med. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21037/apm-20-277\u003c/span\u003e\u003cspan address=\"10.21037/apm-20-277\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Z, Huang Y, Guo Z, Sun J, Zheng G (2025) Interferon-Linked Lipid and Bile Acid Imbalance Uncovered in Ankylosing Spondylitis in a Sibling-Controlled Multi-Omics Study. Int J Mol Sci 26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijms26167919\u003c/span\u003e\u003cspan address=\"10.3390/ijms26167919\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, Fu X, Liu S, Bo X, Yu G (2021) clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innov (Camb) 2(3):100141. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.xinn.2021.100141\u003c/span\u003e\u003cspan address=\"10.1016/j.xinn.2021.100141\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu X, Dong S, Chen H, Guo M, Sun Z, Luo H (2023) \u003cem\u003ePerilla frutescens\u003c/em\u003e frutescens: A traditional medicine and food homologous plant. Chin Herb Med 15(3):369\u0026ndash;375. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chmed.2023.03.002\u003c/span\u003e\u003cspan address=\"10.1016/j.chmed.2023.03.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu Y-T, Chen L, Tan Z-B, Fan H-J, Xie L-P, Zhang W-T, Chen H-M, Li J, Liu B, Zhou Y (2018) Luteolin Inhibits Vascular Smooth Muscle Cell Proliferation and Migration by Inhibiting TGFBR1 Signaling. \u003cem\u003eFrontiers in Pharmacology\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fphar.2018.01059\u003c/span\u003e\u003cspan address=\"10.3389/fphar.2018.01059\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu Y, Wu Y, Xia S, Lian H, Lou Y, Wang L-J (2025) JMJD6-driven epigenetic activation of COL4A2 reprograms glioblastoma vascularization via integrin α1β1-dependent PI3K/MAPK signaling. Acta Neuropathol Commun 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40478-025-02114-9\u003c/span\u003e\u003cspan address=\"10.1186/s40478-025-02114-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia F, Wang C, Jin Y, Liu Q, Meng Q, Liu K, Sun H (2014) Luteolin protects HUVECs from TNF-α-induced oxidative stress and inflammation via its effects on the Nox4/ROS-NF-κB and MAPK pathways. J Atheroscler Thromb 21(8):768\u0026ndash;783. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5551/jat.23697\u003c/span\u003e\u003cspan address=\"10.5551/jat.23697\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong J, Li Z, Tang H, Duan Y, Ban X, Xu K-K, Guo Y, Tu Y (2023) Bulk and single-cell characterisation of the immune heterogeneity of atherosclerosis identifies novel targets for immunotherapy. BMC Biol 21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12915-023-01540-2\u003c/span\u003e\u003cspan address=\"10.1186/s12915-023-01540-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu J, Zhou H, Cheng Y, Xiang G (2022) Identifying potential signatures for atherosclerosis in the context of predictive, preventive, and personalized medicine using integrative bioinformatics approaches and machine-learning strategies. EPMA J 13:433\u0026ndash;449. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s13167-022-00289-y\u003c/span\u003e\u003cspan address=\"10.1007/s13167-022-00289-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYim A, Smith C, Brown A (2022) Osteopontin/secreted phosphoprotein-1 harnesses glial‐, immune‐, and neuronal cell ligand‐receptor interactions to sense and regulate acute and chronic neuroinflammation. Immunol Rev 311:224\u0026ndash;233. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/imr.13081\u003c/span\u003e\u003cspan address=\"10.1111/imr.13081\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan L, Zhang F, Jia S, Xie J, Shen M (2020) Differences between phytosterols with different structures in regulating cholesterol synthesis, transport and metabolism in Caco-2 cells. J Funct Foods 65:103715. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jff.2019.103715\u003c/span\u003e\u003cspan address=\"10.1016/j.jff.2019.103715\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Z, Chen Y, Fu X, Chen L, Wang J, Zheng Q, Zhang S, Zhu X (2024) Identification of PPARG as key gene to link coronary atherosclerosis disease and rheumatoid arthritis via microarray data analysis. PLoS ONE 19(4):e0300022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0300022\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0300022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao T, Li Z, Ji S, Huang Q, Sun C, Lu B (2025) Decoding the mechanism of dietary fatty acids-driven phytosterol esterification promoting intestinal absorption. Food Chem 496:146525. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.foodchem.2025.146525\u003c/span\u003e\u003cspan address=\"10.1016/j.foodchem.2025.146525\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou J, Wu Z-Y, Zhao P (2024) Luteolin and its antidepressant properties: From mechanism of action to potential therapeutic application. J Pharm Anal 15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jpha.2024.101097\u003c/span\u003e\u003cspan address=\"10.1016/j.jpha.2024.101097\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bioresources-and-bioprocessing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"biob","sideBox":"Learn more about [Bioresources and Bioprocessing](http://bioresourcesbioprocessing.springeropen.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/biob/default.aspx","title":"Bioresources and Bioprocessing","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Atherosclerosis, Perilla frutescens, Single-cell RNA sequencing, Machine learning, Molecular dynamics","lastPublishedDoi":"10.21203/rs.3.rs-9174552/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9174552/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDespite the widespread implementation of lipid-lowering therapy, the persistence of residual inflammatory risk, driven by immunometabolic network dysregulation, remains a cardinal therapeutic challenge in atherosclerosis (AS) management. While \u003cem\u003ePerilla frutescens\u003c/em\u003e exhibits well-documented anti-inflammatory properties, the precise molecular targeting within the atherosclerotic plaque microenvironment and the regulatory mechanisms governing intercellular communication networks remain poorly elucidated. We established a multi-scale integrative computational framework synergizing network pharmacology, human atherosclerotic plaque single-cell transcriptomic (scRNA-seq) profiling, and ensemble machine learning algorithms (LASSO and random forest) for systematic identification of robust therapeutic targets. Molecular dynamics simulations validated the binding affinity and thermodynamic stability of drug\u0026ndash;target complexes. We analyzed the cellular heterogeneity lineage of plaques were identified and a core feature set of 10 genes were identified which specifically mapped the differentiation trajectory of macrophages to foam cells. External validation in an independent cohort demonstrated superior diagnostic performance of this signature (AUC\u0026thinsp;=\u0026thinsp;0.996). Cellular communication network dissection revealed the foam cell-driven SPP1\u0026ndash;ITGB1 signaling axis as a pivotal conduit orchestrating inflammatory crosstalk. Molecular docking demonstrated pronounced binding affinity between luteolin, the principal bioactive constituent of \u003cem\u003ePerilla frutescens\u003c/em\u003e, and ITGB1 (binding energy: \u0026minus;8.9 kcal/mol). Molecular dynamics simulations further corroborated the efficacy of luteolin in stabilizing ITGB1 conformation via a \"conformational-locking\" mechanism (RMSD equilibration within 0.10\u0026ndash;0.20 nm), thereby abrogating pathological cell adhesion signaling transduction. Collectively, this study provides a high-resolution molecular atlas of \u003cem\u003ePerilla frutescens\u003c/em\u003e-mediated AS intervention, systematically elucidating the mechanistic paradigm whereby luteolin attenuates vascular inflammation through targeted disruption of the SPP1\u0026ndash;ITGB1 communication axis.\u003c/p\u003e","manuscriptTitle":"Unveiling Anti-atherosclerotic Targets of Perilla frutescens through a Multi-scale Computational Framework Integrating Network Pharmacology, Single-cell Analysis, Machine Learning, and Molecular Dynamics","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-16 19:48:52","doi":"10.21203/rs.3.rs-9174552/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2026-04-09T07:27:17+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-08T10:55:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-06T10:12:23+00:00","index":"","fulltext":""},{"type":"submitted","content":"Bioresources and Bioprocessing","date":"2026-03-31T21:16:00+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bioresources-and-bioprocessing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"biob","sideBox":"Learn more about [Bioresources and Bioprocessing](http://bioresourcesbioprocessing.springeropen.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/biob/default.aspx","title":"Bioresources and Bioprocessing","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6f6585f0-6869-4f97-a667-af6588a4cf40","owner":[],"postedDate":"April 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-18T02:56:15+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-16 19:48:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9174552","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9174552","identity":"rs-9174552","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}