Machine learning integration with multi-omics data constructs a robust prognostic model and identifies PTGES3 as a therapeutic target for precision oncology in lung adenocarcinoma

Background Lung adenocarcinoma is the most prevalent lung cancer type, with a 5-year survival rate for advanced patients below 20%. This study aims to develop a risk model to guide treatment for these patients. Materials and methods RNA-seq data from TCGA and GEO were analyzed using Cox regression and 10 machine learning algorithms to identify prognostic genes and stratify patients. Single-cell datasets were integrated to examine PTGES3’s role in tumor progression, with SCENIC and ATAC-seq revealing its transcriptional regulators. PTGES3 expression was evaluated via tissue microarray immunohistochemistry. Functional assays (CCK-8, colony formation, flow cytometry, Western blot) after lentiviral knockdown in lung cancer cells assessed its effects on proliferation, apoptosis, and cell cycle. ZBTB7A was validated as a transcriptional regulator of PTGES3 by dual-luciferase reporter assay, and xenograft models in nude mice evaluated tumor growth in vivo . Results Our analysis identified 28 key genes, classifying lung adenocarcinoma samples into high-score and low-score groups. Patients in the high-score group showed worse prognoses, linked to clinical stage progression and phenotypes like angiogenesis and epithelial-mesenchymal transition. PTGES3 knockdown inhibited tumor growth, leading to cell cycle arrest and increased apoptosis. ZBTB7A was identified as a key regulator of PTGES3, while interactions among LGALS9, P4HB, and CD44 significantly impacted signaling pathways influencing the tumor microenvironment’s immune status. Conclusions Our findings highlight the potential of LS score-based molecular subtyping to improve treatment strategies for lung adenocarcinoma and emphasize PTGES3’s role in new therapeutic development.