Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal malignancies worldwide, characterized by early metastasis and challenges in early diagnosis. Early metastatic dissemination deprives the majority of patients of surgical opportunities, making it the most critical factor limiting overall survival. The initial step of metastasis involves tumor cells breaching the basement membrane. Therefore, identifying therapeutic targets from basement membrane-associated genes to inhibit early tumor metastasis represents a crucial molecular-targeted treatment strategy. By integrating transcriptomic and single-cell data with a basement membrane-associated gene set, we identified a series of potential metastasis-related genes. Their relevance to tumor metastasis was validated across multiple cancer types through GSEA enrichment analysis and pathway correlation analysis. Further investigations, including unsupervised clustering, GSVA, Cox regression, and various immune infiltration analyses, highlighted the critical role of these genes in PDAC metastasis. Finally, multiple machine learning algorithms identified ITGA3 as a potential therapeutic target in PDAC metastasis. Its pro-metastatic function was experimentally validated using Western blotting, RT-qPCR, CCK-8 assay, colony formation, and Transwell assays. All in all, we identified multiple genes highly associated with PDAC metastasis and used them to classify two metastatic subtypes with distinct survival outcomes, mutation frequencies, and immune characteristics. Feature selection was performed using multiple machine learning algorithms, and the model was interpreted using SHAP values. Four key genes, including ITGB3, ITGA2, ITGA3, and SPOCK1, were identified as critical players in PDAC metastasis. Among them, ITGA3 emerged as the most crucial gene, potentially serving as a therapeutic target. This study presents a novel approach for identifying metastasis-associated target genes and highlights the critical role of basement membrane-related genes in PDAC metastasis.
Zheng et al. (Sun,) studied this question.