Macrophage-Related Gene Signatures for Predicting Prognosis and Immunotherapy of Lung Adenocarcinoma by Machine Learning and Bioinformatics

Abstract

Background: In recent years, the immunotherapy of lung adenocarcinoma has developed rapidly, but the good therapeutic effect only exists in some patients, and most of the current predictors cannot predict it very well. Tumor-infiltrating macrophages have been reported to play a crucial role in lung adenocarcinoma (LUAD). Thus, we want to build novel molecular markers based on macrophages. Methods: By non-negative matrix factorization (NMF) algorithm and Cox regression analysis, we constructed macrophage-related subtypes of LUAD patients and built a novel gene signature consisting of 12 differentially expressed genes between two subtypes. The gene signature was further validated in Gene-Expression Omnibus (GEO) datasets. Its predictive effect on prognosis and immunotherapy outcome was further evaluated with rounded analyses. We finally explore the role of TRIM28 in LUAD with a series of in vitro experiments. Results: Our research indicated that a higher LMS score was significantly correlated with tumor staging, pathological grade, tumor node metastasis stage, and survival. LMS was identified as an independent risk factor for OS in LUAD patients and verified in GEO datasets. Clinical response to immunotherapy was better in patients with low LMS score compared to those with high LMS score. TRIM28, a key gene in the gene signature, was shown to promote the proliferation, invasion and migration of LUAD cell. Conclusion: Our study highlights the significant role of gene signature in predicting the prognosis and immunotherapy efficacy of LUAD patients, and identifies TRIM28 as a potential biomarker for the treatment of LUAD.