Unveiling anti-atherosclerotic targets of Perilla frutescens through a multi-scale computational framework integrating network pharmacology, single-cell analysis, machine learning, and molecular dynamics.

Medicinal plants have long served as important sources of therapeutic agents owing to their diverse bioactive constituents and multi-target pharmacological properties. In particular, plant-derived compounds have attracted increasing attention for the management of chronic inflammatory and metabolic diseases, including atherosclerosis. However, the molecular mechanisms by which medicinal plants modulate the cellular heterogeneity and intercellular communication networks within atherosclerotic plaques remain insufficiently understood. 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. To address this gap, 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. Subsequently, molecular docking coupled with 100-ns all-atom molecular dynamics (MD) simulations validated the binding affinity and thermodynamic stability of drug-target complexes. The study successfully analyzed the cellular heterogeneity lineage of plaques and identified a core feature set of 10 genes including HIF1A, PPARG and ITGB1, 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). MD 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. These findings underscore the therapeutic targeting of cell adhesion receptors as a translationally promising strategy for mitigating residual inflammatory risk in AS.