The mechanisms of the effect of propionate metabolism and immunity on nonalcoholic fatty liver disease (NAFLD) have not been adequately studied. Firstly, differentially expressed-propionate metabolism-related genes (DE-PMRGs) were selected by overlapping PMRGs and differentially expressed genes (DEGs) between the simple steatosis (SS) and health control (HC) groups. Then, common genes were selected by overlapping DE-PMRGs and key module genes obtained from weighted gene co-expression network analysis (WGCNA). Subsequently, the biomarkers were screened out by machine learning algorithms. The expression of the biomarkers was validated by quantitative Real-time PCR. In total, 5 biomarkers (JUN, LDLR, CXCR4, NNMT, and ANXA1) were acquired. The nomogram constructed based on 5 biomarkers had good predictive power for the risk of SS. Next, 5 biomarkers, 11 miRNAs, and 149 lncRNAs were encompassed in the ceRNA regulatory network. The expression of biomarkers was significantly higher in the HC group than in the SS group, which was consistent with the results in the GSE89632 and GSE126848 datasets. In this study, 5 immune and propionate metabolism-related biomarkers (JUN, LDLR, CXCR4, NNMT, and ANXA1) were screened out to provide a basis for exploring the prediction of diagnosis of NAFLD.