整合 117 种机器学习算法与单细胞转录组学识别巨噬细胞极化和 ER 应激特征用于癌症预后和精准治疗

Macrophage polarization and endoplasmic reticulum (ER) stress play critical yet incompletely understood roles in cancer progression and therapeutic resistance.

Here, we conduct a systematic pan-cancer analysis of macrophage polarization and ER stress-related genes (MPERSRGs) by integrating multi-omics data from The Cancer Genome Atlas (TCGA), Genotype-Tissue Expression (GTEx), Cancer Cell Line Encyclopedia (CCLE), and single-cell RNA sequencing datasets across 33 cancer types.

We identify distinct expression patterns of seven core MPERSRGs (CEBPB, NUPR1, ATF3, CASP3, TNFSF10, BRSK2, NOD2) that correlate significantly with tumor stage, immune infiltration, and patient prognosis. Employing 117 machine learning algorithm combinations, we develop a robust five-gene prognostic signature (FAM83A, RHOV, CPS1, STRIP2, SLC2A1) for lung adenocarcinoma (LUAD) with area under the curve values of 0.692, 0.688, and 0.614 for 1-, 3-, and 5-year overall survival, respectively. Single-cell transcriptomic analysis of 86,378 cells reveals three functionally distinct fibroblast subpopulations (MFAP5+, MATK+, HP+) with differential MPERSRG expression profiles, with MFAP5 + fibroblasts showing the highest enrichment in epithelial-mesenchymal transition and angiogenesis pathways. Cell-cell communication analysis identifies fibroblasts and epithelial cells as the most interactive populations, with the CLEC2C-KLRB1 ligand-receptor pair mediating the strongest signaling between mast cells and NK cells. Drug sensitivity predictions across multiple databases identify vorinostat, nilotinib, olaparib, and paclitaxel as potential therapeutic agents showing differential efficacy based on MPERSRG expression stratification.

These findings establish MPERSRGs as key determinants of tumor-immune interactions and provide actionable biomarkers for risk stratification and precision therapy selection in cancer.