A deep understanding of the biological mechanisms of lung cancer offers more precise treatment options for patients. In our study, we integrated data from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) to investigate lung adenocarcinoma. Analyzing 538 lung cancer samples and 31 normal samples, we focused on 3076 autophagy-related genes. Using Seurat, dplyr, tidyverse, and ggplot2, we conducted single-cell data analysis, assessing the quality and performing Principal Component Analysis (PCA) and t-SNE analyses. Differential analysis of TCGA data using the “Limma” package, followed by immune infiltration analysis using the CIBERSORT algorithm, led us to identify seven key genes. These genes underwent further scrutiny through consensus clustering and gene set variation analysis (GSVA). We developed a prognostic model using Lasso Cox regression and multivariable Cox analysis, which was then validated with a nomogram, predicting survival rates for lung adenocarcinoma. The model’s accuracy and universality were corroborated by ROC curves. Additionally, we explored the relationship between immune checkpoint genes and immune cell infiltration and identified two key genes, HLA-DQB1 and OLR1. This highlighted their potential as therapeutic targets. Our comprehensive approach sheds light on the molecular landscape of lung adenocarcinoma and offers insights into potential treatment strategies, emphasizing the importance of integrating single-cell and genomic data in cancer research.