Development of a diagnostic model for MASLD and identification of daidzein as the potential drug using bioinformatics analysis and experiments.

Background: Metabolic dysfunction-associated steatotic liver disease (MASLD) is now the predominant chronic liver disease globally, yet effective therapeutic strategies remain elusive. Methods: MASLD-related datasets were download from GEO. Subsequently, genes associated with MASLD were found through the intersection of differentially expressed genes and WGCNA. Then, key candidate genes were further screened using 113 machine learning algorithms and their diagnostic value was evaluated using ROC curve analysis across multiple datasets. Genes are then screened by Shapley Additive exPlanations (SHAP) analysis. Molecular docking (MD) and molecular dynamics simulations (MDS) were employed to validate the interaction between Daidzein and Enolase 3 (ENO3). Finally, an in vitro fatty liver cell model was constructed to validate the "Enrichr" platform to identify poteitial drugs for MASLD. Results: 62 MASLD-DEGs were finally identified. The optimal predictive model for MASLD was the 17-gene signature (IGFBP1, ENO3, SOCS2, GADD45G, NR4A2, RTP4, RAB26, CRYAA, PPP1R3C,MCAM, IL6, IER3, RTP3, NR4A1, CCL5, FOS, JUNB) selected through combined glmBoost+GBM algorithms, which was demonstrated robust predictive performance. SHAP analysis suggested that ENO3 may be the most prominent genes associated with MASLD severity. More importantly, we measured the effect of daidzein on improving lipid accumulation in vitro model. Conclusion: We developed a predictive model for MASLD and identified ENO3 as a key predictive gene. Furthermore, we discovered that daidzein may serve as a potential therapeutic agent for MASLD. Through in vitro studies, we further confirmed that daidzein alleviates lipid deposition and improves MASLD by modulating the ENO3/PPAR signaling pathway.