Abstract
To identify clinical risk factors for diabetes mellitus (DM) in long-stay schizophrenia (LS-SCZ) patients, explore shared molecular mechanisms of schizophrenia and DM, and validate targeted interventions. Clinical data of LS-SCZ patients were analyzed via multiple logistic regression to identify DM risk factors. Differentially expressed genes (DEGs) from GSE53987/GSE161355 datasets were screened; overlapping DEGs were analyzed for co-expression and KEGG enrichment. Support vector machine (SVM) selected feature genes whose diagnostic efficacy was evaluated by ROC curves. Molecular docking verified Genistein's binding with feature gene proteins. Risk factors for diabetes in LS-SCZ patients included age ≥50 years, hospitalization >20 years, BRI >5.306, family history of diabetes, and hypertension (all p < 0.05). Molecularly, 27 overlapping DEGs from two datasets were enriched in neuroactive ligand-receptor interaction, HIF-1, and MAPK pathways. SVM identified four feature genes (NPY, MKNK2, IFITM3, and S100A8) with good diagnostic efficacy. Genistein bound strongly to their proteins (binding energies: -8.98 to -6.026 kcal/mol). In conclusion, LS-SCZ patients have high DM risk. Targeting clinical risk factors and using Genistein for feature genes may reduce DM comorbidity.