IntroductionPost-transcriptional gene regulation (PTGR) includes RNA maturation (splicing, editing, polyadenylation, etc.), localization, stability, translation, and degradation, which is the core of gene expression1,2,3. Increasing evidence has demonstrated that PTGR disorders, such as aberrant alternative splicing and changes in RNA stability, are key to cancer development4,5. Meanwhile, these disorders often result in a reduced therapeutic effect and poor prognosis in cancer patients. RNA-binding proteins (RBPs), as special functional proteins, have a crucial regulatory role in various RNA metabolism processes6,7. Because RBP dysregulation is the key condition for PTGR disorder occurrence8, conducting systematic research on RBP expression in cancer patients is extremely necessary.Lung cancer is the leading cause of cancer-related deaths globally9. Lung adenocarcinoma (LUAD) is the most common pathological type of lung cancer, with a 5-year survival rate of only approximately 15%10,11. Considerable progress has been made in chemoradiation, targeted therapy, and immunotherapy, but effective treatments for LUAD patients remain lacking12,13. In recent years, RBP dysregulation in LUAD patients has gradually attracted the attention of researchers, but it has been explored to a limited extent. Therefore, developing RBP-related prognostic biomarkers for LUAD treatment is exceptionally promising and necessary.The big data screening ability and accuracy of machine learning can be used to discover more promising therapeutic targets for LUAD14. However, LUAD-related research is lacking. Therefore, we conducted a comprehensive study on RBP-related genes involved in LUAD and the potential mechanisms through which they affect tumor progression, to discover more efficacious therapeutic targets and better guide clinical treatment. In this study, machine learning was employed for the systematic screening of the data from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) based on RBP-related genes. This technique calculated the C-index of 72 prediction model combinations of five public datasets and screened the genes with greater prognostic prediction values for the follow-up analysis. Subsequently, a promising prognostic risk model was established. Four LUAD GEO cohorts were used to validate the prediction performance of the model for survival. The risk model was significantly related to the TNM stage, tumor microenvironment (TME) differences, and response to immunotherapy. Finally, the levels of the hub gene DEAD-box helicase 56 (DDX56), an RNA helicase family member, were significantly upregulated in LUAD tissues compared with normal tissues, which predicted poor prognosis. The high expression of this gene was associated with the disorder of the immune microenvironment and tumor immune escape in LUAD. DDX56 promotes LUAD progression and reduces patients’ sensitivity to platinum-based drugs by regulating Bcl-2 and NF-kB pathways. Therefore, DDX56 can serve as a LUAD oncogene and is a promising therapeutic target.ResultsIdentification and analysis of prognosis-associated differentially expressed RBPsWe determined differentially expressed RBPs in LUAD tissues by analyzing differentially expressed genes (DEGs) (|logFC| > 5, adjusted p-value